2 M A NAG E M E N T & R E S E A RC H The need for us to quantify skin tones e n i c i d e M f o l o o h c S d r o f n a t S © Making AI in dermatology w to standard file formats and colour fidelity, there is a wide variety on how to store images with uncom- pressed or compressed JPG, and the compression level used is not always encoded.’ Once image capture challenges are resolved, what should be done with the images? Research is under- Roxana Daneshjou, MD PhD, is a clini- cal scholar in dermatology and a post- doctoral research fellow in biomedical data sciences at Stanford University School of Medicine in Redwood City, California. She is interested in bridging new technologies such as genomics and machine learning with clinical medicine and in the use of Twitter for scientific communication and medical education. way in the fields of segmentation, wound measurement, and diagno- sis using machine learning and AI, Rotemberg said, adding that a 2020 SIIM-ISIC Challenge for melanoma classification outperformed derma- tologists, with a sensitivity of 79.3% for AI compared to 71.9% for der- matologists. However, issues remain for clini- cal implementation of AI for skin cancer screening, such as identify- ing the appropriate evaluation crite- ria and appropriate users. stumbling block, Image storage requirements for dermatology data can be anoth- er suggests Konstantinos Liopyris, MD PhD, of the University of Athens Andreas Syggros Hospital of Skin and Venereal Diseases. Dermatology data is massive, including patient demographics, anatomic locations challenge to the way medical sci- ence has worked so far. In dec- ades of evidence-based medicine and care, large, randomised studies have become the gold standard of medical knowledge acquisition. Their goal is to evaluate the effects of a therapy (or other intervention) with optimum reliability and preci- sion, eliminating bias, confounding factors and just plain coincidence as far as possible. In the age of randomised con- trolled trials (RCTs), an intervention is only considered effective if there is a statistically significant differ- ence in outcome still visible after averaging hundreds or thousands of patients. In this way, the RCT does not only eliminate coincidence but also the influence of individual genetic pre- dispositions, unless the study popu- lation is chosen explicitly for a cer- tain genetic trait. This, however, is possible only for the most frequent genetic variations. For example, breast and ovarian cancer patients are likely to profit from treatment with a PARP inhibitor only if they carry a BRCA-1 or BRCA-2 muta- tion. Such mutations are sufficiently frequent to include their carriers in large clinical studies. However, there are many more genes more subtly involved in can- cer risk, and once we start consider- ing not only single gene mutations but also combinations of mutations, things get infinitely more complex. As we know, there are hardly two people in the world sharing the exact same genetic make-up. Thus, the RCT might even be considered EUROPEAN HOSPITAL Vol 30 Issue 3/21 Although artificial intelligence (AI) tools and smartphone apps that help identify suspicious moles and potential skin cancers are starting to proliferate, dermatology informatics has far to go before becom- ing a clinically adoptable technology. Many challenges need to be resolved, not least of which is the need for more image data repre- senting people of colour. racy and fidelity are huge issues. In addition to lighting affecting colour, there is much variability in image sources from smartphones. Each acquisition device may code RGB values differently when capturing and storing images. With respect Report: Cynthia E. Keen Recently, three dermatologists work- ing to advance informatics and the use of AI for dermatology dis- cussed current challenges at the 2021 Society of Imaging Informatics in Medicine (SIIM) annual meeting. The scientific session raised more questions than answers, making it apparent that much more collabo- rative research needs to be under- taken as a global initiative. The scope of dermatology pho- tographs is broad, encompassing photos taken by patients and their primary care physicians, clinical photos, wound care photos, and his- topathology CT scans. All of these need to be managed, labelled and stored appropriately, or else they cannot be used clinically or as refer- ence data, said Veronica Rotemberg, MD PhD, of the Memorial Sloan Kettering Cancer Center in New York City. ‘Standards for image acquisi- tion are challenging. Many imaging standards need to be established, such as setting photography stand- ards, standards for special consider- ations for sensitive sites, and stand- ards for skin of colour.’ The use of DICOM, the interna- tional standard for medical imaging, for dermatological imaging has been limited. The ability of DICOM to encode relevant metadata and ena- ble interoperability with electronic health records and clinical research software has driven the interest in its application to dermatology. A DICOM dermatology working group has developed dermatology-specif- ic Information Object Definitions (IODs) for inclusion in the DICOM standard and future work will be for total body photography and reflec- tance confocal microscopy. ‘There are numerous issues to resolve,’ Rotemberg pointed out. ‘No consensus exists yet on the specific fields to collect and store for every type of image. Patient level varia- bles, such as ethnicity and skin type, need to be considered. Colour accu- Metastatic melanoma cells a i c n e l a V . C o i l u J / h c r a e s e R r e c n a C r o f r e t n e C ) I C N ( e t u t i t s n I r e c n a C l a n o i t a N : e c r u o S Advancing personalised health and genomics Continued from page 1 arily digitised or primarily digitally acquired slides, employing machine learning methods to identify pat- terns predicting malignancy. While radiology and histopathol- ogy are approaching the cancer cell from a macroscopic to a micro- scopic level, genetic sequencing is working its way up from a sub- microscopic level. High-throughput sequencing methods have all but replaced the conventional Sanger sequencing approach and have made it possible to sequence a cell’s whole genome in a matter of days or even hours. Massive genetic databases enable scientists to algorithmically identify pathogenic mutations within the overwhelming genetic variability of the human species. Minimal collateral damage Ideally, detailed knowledge about a cell population’s genetic aberra- tions helps to identify pathways that are instrumental for the growth of a specific tumour and can be inhibited with minimal collateral damage to the patient’s healthy cell populations. This so-called molecular profiling was implemented, among others, in the NIH-funded I-PREDICT trial. Of 83 patients with previously ther- apy-resistant tumours, 73 received an individual combination of drugs matched to their respective molecu- lar profile, with encouraging results regarding prolonged stable disease. To make results like these accessible to the medical community, there is now even a journal dedicated to personalised and precision oncolo- gy, Nature Precision Oncology (npj). The principle of tailoring thera- pies to a patient’s genetic make-up is not limited to oncology: in phar- macogenomics, scientists investigate the influence of genetic variation on cellular pathways and thus the way that individual patients react to or metabolise drugs, finding out which drugs work best on a specific patient or subgroup of patients and which to avoid. However, person- alised medicine poses a particular Personalised medicine is key to any patient’s journey to health pathways, traditional cancer thera- pies are highly noxious to both nor- mal and malignant cells. They cause a wide range of side effects, mostly in tissues with a naturally high proportion of healthy fast-growing cells, such as the gastrointestinal tract, bone marrow and skin. Their side effects may be as fatal as the underlying disease. Thus, the holy grail of cancer medicine are therapies that are pre- cisely targeted to eliminate malig- nant cells only. To develop such therapies requires a deep under- standing of the cancer cell down to the molecular level. Insights into the cancer cell There are three main fields of medi- cal science that contribute to an intimate knowledge of the workings of a cancer cell: radiology, histopa- thology and genetics/genomics. In recent decades, radiology has pro- gressed from giving a rough orienta- tion of a lesion’s spatial dimensions and internal structure to providing detailed information almost on a cellular level. In the newly developed field of radiomics or quantitative imaging, machine learning algorithms exploit subtle differences in signal strength that are invisible to the human eye. Similarly, histopathology has pro- gressed from traditional H&E stain- ing for visual examination under the microscope to far more sophisti- cated immunohistochemistry, using specific antibodies that bind to cer- tain cell components. Information on a sub-visual level is being utilised as well, on second-

4 S U RG E RY Women in medical R&D Innovation depends on more than technical skills Cécile Geneviève is one of the few women who lead research and development (R&D) at a major compa- ny and her increasingly female team reflects women’s growing interest in the field. But while gender bal- ance is an important criterion, it takes a broad palette of skills to innovate to alleviate pain for millions of patients, she explained in an interview with European Hospital. Report: Mélisande Rouger Eagerness is what emerges from a conversation with Cécile Geneviève, a French mechanic and development engineer who leads system engi- neering at Zimmer Biomet Robotics. It’s easy to imagine how she inspires her team of 20 engineers to follow in her footsteps and deliver their best to innovate in surgery robotics. ‘I like to understand a complex subject and detail it, dissect it, and then con- trol and learn from it,’ she said. ‘This is what attracted me to physics and mathematics, and science in general. R&D presents a dynamism and ener- gy in which I can constantly evolve.’ She started working in the automo- tive industry in the late 1990s, drawn by the many opportunities to learn and grow in a complex and technical world. In 2010, her focus switched to the medical device industry and robotic-assisted surgery and she never let go. ‘The pace of innovation Below: A surgeon interacts onscreen and with the ROSA robot and opportunity to deliver a service to patients and users is very exciting; it’s an area where so many things are happening right now.’ What makes a good team? Leadership positions continue to be dominated by men in many industries, often more so in techni- cal departments. But the world is changing. An increasing number of women are taking an interest in R&D leadership, such as Liane Teplitsky, Vice President and General Manager of Technology & Data Solutions at Zimmer Biomet’s. ‘Gender divide is fading out,’ Geneviève said. ‘I don’t notice a difference in the way we work between France, Canada and the United States. An HR publi- cation was sent out recently that showed almost no difference in sala- ries between men and women in our company. I think this helps to inspire more women to work in this field.’ Geneviève witnessed the change early on. When she studied mechan- ical engineering at the National Institute of Applied Sciences (INSA, in French) in Lyon back in the 1990s, 30% of her classmates were women. ‘This was the first time so many women joined that course,’ she said. ‘Some areas, for example software engineering, remain male-dominat- ed. There are just not enough female students in the pipeline.’ Geneviève’s team now comprises 40% of women system engineers. The proportion wasn’t an objective but reflects a sense of fairness and desire to attract the right skills in her department, she explained. ‘I am very sensitive to the question of equality and, for me, it’s crucial to offer equal opportunities to all quali- fied applicants, regardless of gender or background. Technical and soft skills, motivation and respect for oth- ers is what really matters.’ Geneviève and her team work closely with healthcare professionals to speed the pace of innovation and expand the possibilities for successful treat- ment outcomes. ‘Collaborating with physicians, we’ve created a compre- hensive inventory for joint recon- t e m o B i r e m m i Z © struction, bone and skeletal repair, sports and spinal injuries, and dental procedures available.’ Launched: The Woman in Orthopedics program There are more women in orthope- dics than first thought and the num- ber is growing. The Zimmer Biomet Institute has recently launched the Woman in Orthopedics program to support the trend. Surgeons are very demanding partners regardless of gender. ‘They have very high expec- tations and that’s why the develop- ment of a product such as the ROSA robot is so interesting for engineers,’ she said. ‘The key challenges are related to the variety of surgical techniques used by surgeons. We need to develop features that are flexible enough for good adaptabil- ity to these techniques, while achiev- ing the highest level of quality for performance and safety.’ ‘Patient and user safety is at the center of our activities, and risk management and quality are a daily focus for R&D teams,’ she said. ‘ROSA ONE Brain is a good exam- ple of this patient-focused work frame.’ The robot assists surgeons in planning and performing complex neurosurgical procedures through a small drill hole in the skull. The technology enables surgeons to per- form less invasive procedures than traditional craniotomies – enabling smaller incisions and potentially enhancing patient comfort. ‘The device may give access to a therapy for patients who could not have sup- ported a conventional procedure, because of the operating time, for example the SEEG, which aims to identify epilepsy foci for drug resist- ant patients,’ Geneviève said. Another innovation, ROSA ONE Spine, enables to perform mini- mally invasive surgeries and has been proven to significantly reduce blood loss, lower infection rates and reduce complications. Studies show that use of robotic guidance during spinal fusion significantly reduces radiation exposure and length of Cécile Geneviève leads a team of R&D engineers at Zimmer Biomet Robotics, a global musculoskeletal healthcare leader. Based in Montpellier, France, Cécile and her team are developing robotic solutions that assist surgeons to treat patients suf- fering from disorders of, and injuries to bones, joints its supporting soft tissues and performing complex neurosurgical procedures. stay. Technology is increasingly inte- grated in the operating room and robots are only one part of the equation. In the future, there will be a stronger integration between implants and technology, artificial intelligence and data insights, she believes. ‘Many devices and products are now robotic, connected and/or equipped with sensors: operating tables, lighting, imaging systems, microscopes, drills and reamers. All these devices will work as integrated units, delivering new sources of data insights to further improve patient care.’ Robotics is a part of ZBEdge, Zimmer Biomet’s suite of integrat- ed digital and robotic technologies, which has been engineered to deliv- er data-powered clinical insights, shared seamlessly across the patient journey. ‘Data processing and artificial intelligence make it possible to re- run surgeries for training purpos- es, share a specific technique with other surgeons or train young sur- geons. The possibilities are endless,’ Geneviève concluded, ‘and that’s what makes it so exciting.’ Changing processes increased output by 30 percent Streamlining and higher quality Rein Medical, a leading supplier of customised all-in-one computer systems, high-resolution displays, customised OR wall modules and video and image management software solutions across Germany and Europe, produces high-quality, robust products. ‘To maintain these high standards going forward, the JVC subsidiary has enlarged and redesigned the warehouse and production area and optimised the processes,’ the company reports. The company’s enlarged and redesigned warehouse and production area ‘We are focusing on two separate production areas. The mechanical components are manufactured in the separate pre-production area, which keeps most of the dust and impurities outside the electronic assembly. ‘In the production area, which is protected from electrostatic dis- charge (ESD) and air-conditioned, the assemblies are created on mobile workstations and then inserted into the pre-assembled housings. This protects the devices from uncon- trolled electrostatic discharges, which can cause invisible damage inside the components. ‘Thanks to this more efficient production process and the enlarge- ment of the production areas, we could increase production output by 30 percent, with less work for the production technicians and with the same number of production days. ‘We have also improved quality assurance with the new ESD area,’ said Stephan Rein, founder and managing director of Rein Medical, when describing the new processes and their advantages. During the redesign, special attention was paid to ergonom- ics at the workplace, he added. ‘After all, we want to create the best working conditions for our employees, and improve the qual- ity of their work and products.’ The Mönchengladbach-based com- pany benefited from the experience of its colleagues at JVCKenwood Corporation in Nagaoka ( Japan) in regard to more efficient production and logistics processes. Details: www.reinmedical.com EUROPEAN HOSPITAL Vol 30 Issue 3/21

6 S U RG E RY 3-D printed titanium-alloy knee plates High tibial osteo- tomy simplified 3-D metal printing technology is producing personal- ised medical-grade titanium-alloy plates that perfectly fit individuals suffering arthritis of the knee. Engineers at the University of Bath’s Centre for Therapeutic Innovation (CTI) working with 3-D Metal Printing Ltd, are using the TOKA (Tailored Osteotomy for Knee Alignment) treatment to improve the surgical procedure and fit of high-tibial osteotomy (HTO) plates to realign a patient’s knee. The sys- tem improves stability, comfort and weight bearing over existing generic plates. The technique also simplifies high tibial osteotomy (HTO), mak- ing operations quicker and safer. The HTO plates have been safe- ty tested in a computer-based vir- tual ‘in-silico’ trial using CT scan data from 28 patients. The in-silico clinical trial, globally the first to demonstrate the safety of an ortho- paedic device, modelled the stress- es that would be exerted on the bespoke plates and showed these would be comparable in safety to the standard treatment. ‘Knee osteoarthritis is a major health, social and economic issue and does not receive as much atten- tion as it should,‘ Professor Richie Gill, from the Centre for Therapeutic Innovation, pointed out. A quarter of women over 45 have it, and about 15 percent of men, so it’s a significant burden that many live with. Knee replacement is only use- ful for end-stage osteoarthritis, so you can be in pain and have to live with a disability for a long time, potentially decades, before it’s possible. We hope that the new TOKA process we’ve developed will change that.‘ Knee osteoarthritis patients receiving TOKA will undergo a 3-D CT scan of their knee, before a per- sonalised 3-D printed surgical guide and plate, both shaped to their tibia (shin bone) is created. The surgical guide simplifies the surgery and is designed to improve accuracy. The process also sees the first implementation of 3-D printed screw threads into the HTO plates, meaning they can be optimally posi- tioned to help secure them against the bone. Upcoming trials When clinical centres return to car- rying out elective surgery, expected later this year, trials will begin. UK hospitals in Bath, Bristol, Exeter and Cardiff will take part in a ran- domised control trial to compare patient outcomes with an existing generic HTO procedure. Tests of the TOKA tech- Italy nique have already begun where, so far, 25 patients received new person- alised HTO plates as part of a trial at the Rizzoli Institute in Bologna. HTO in sur- gery realigns the knee joint by making a cut to the tibia and opening a small gap, which needs to be stabilised by a metal plate. This moves the loading to a less worn part of the knee. Patient outcomes depend on how accurately the cut is made and the gap opened. realignment ‘HTO surgery has a long clinical history and has very good results if done accurately. The difficulty sur- geons have is achieving high accu- racy, which is why we have created the TOKA method,’ Gill explained. ‘This starts with a CT scan and digital plan. 3-D printing the custom knee implant and doing the scan- h t a B f o y t i s r e v i n U © ning before operating means sur- geons will know exactly what they’ll see before operating and where the implant will go. In addition to a surgeon being able to precisely plan an operation, a surgical guide (or jig) and a plate implant, each per- sonalised to the patient, can be 3-D printed automatically, based on the scanning data. Importantly this type of treatment relieves the symptoms The implant preserves the existing joint and can be used at an earlier stage of arthritis, that means before a knee replacement becomes necessary. of knee osteoarthritis while preserv- ing the natural joint.’ Pre-planning greatly simplifies surgery and could cut time on the operating table from two hours to around 30 minutes. (SKE) Modern cementing technology Cemented hips help preventing inf For decades, hip arthroplasty has been a routine procedure. In Europe, both total and partial hip replacements are the most frequent surgical interventions for patients with hip fracture or osteoarthritis. The treatment relieves pain and has good long-term outcomes. In cemented hip replacement, the artificial acetabulum and/or the femoral stem are fixed with bone cement and implant and bone bond immediately. To prevent infections, the cement contains one or two antibiotics. Comprehensive studies have con- firmed many benefits of this meth- od. to The trend towards uncemented implants Nevertheless, in Denmark, currently around 70 % of the patients receive uncemented implants, whereas Norway and Sweden are more con- servative, but the use of uncement- ed implants has increased even in these countries, in Sweden from 3% in 2000 to 28% in 2019. the German According Arthroplasty Registry’s (EPRD) annual report 2019, in Germany, the number of cemented total hip replacements decreased from 8 % to 5 % between 2014 and 2018 across all patient age groups; at the same time, the percentage of entirely cement-free replacements increased from 75 % to 79 %. ‘Comprehensive studies have confirmed the many advantages of cemented hip replace- ments, such as a reduced frac- ture risk. Particularly certain patient culture in healthcare systems. A sustainable bonding solution The goals of any surgical interven- tion are pain reduction, maintain- ing long-term function and avoiding complications and revision surgery. The outcome is shaped by sev- eral factors – which can be partially controlled by the surgeon and the patients. ‘Advanced patient age is a given,’ says Overgaard. But ‘con- sumption of tobacco, alcohol or medication and the willingness to undergo surgery are issues that patient and physician can discuss.’ In addition to patient selection, perioperative care and rehab, the choice of bonding components plays a crucial role. This decision should always be based on reliable data. ‘There is plenty of evidence from high-quality studies,’ says Overgaard and adds that his team ‘chooses the individual suitable fixing methods based on this evidence.’ The high percentage of uncemented implants shows that this is not best practice in all hospitals. Overgaard under- lines a lack of experience par- ticularly among young physicians. ‘We surgeons only teach what we can do,’ he says, but ‘the objective should be to decide on the basis of reliable evidence in order to achieve best outcomes for the patient and reduce complications.’ Periprosthetic fracture is more common in uncemented stem fixation The most common complications after hip replacement are loosen- groups, above all older patients or patients with hip fractures, profit from this type of implant. ‘However, in many countries, clini- cal practice does not follow scientif- ic evidence,’ says Professor of ortho- paedic surgery Søren Overgaard of Copenhagen University Hospital, Bispebjerg, Denmark. The reasons for this trend, he assumes, might be the fact that surgeons are not familiar with the current evidence and the differences in training and Dr Georgios Chatziagorou, University of Gothenburg, Sweden Professor Nils Hailer, Uppsala University Hospital, Sweden Professor Søren Overgaard, Copenhagen University Hospital, Bispebjerg, Denmark ing of the components, infection, periprosthetic fracture, and disloca- tion. Among these complications, periprosthetic fracture (PPF) had the highest mortality, according to the Swedish Arthroplasty Register. ‘Thus, prevention of a PPF not only can increase the quality of life, but also might save lives,’ con- cludes Dr Georgios Chatziagorou from the University of Gothenburg in Sweden. ‘It’s therefore, important to iden- tify the patients most at risk of PPF,’ he emphasises. Increasing age is a known risk factor, and uncemented femoral components resulted in higher risk of revision surgery, according to a large reg- ister study. Among male patients aged 60-69 years who received an uncemented stem, the risk of PPF was four times as high as among the same age group with cemented stem. However, among women of the same age-group, the risk was thirteen times higher for those with uncemented stem fixation compared to women with a cemented femoral component. Women older than 55 years should receive cemented hip replace- ments, according to data from the Norwegian Arthroplasty Register. A Danish study showed positive effects of changing from uncement- ed to cemented fixation. The number of periprosthetic fracturea decreased from 4.1 % to EUROPEAN HOSPITAL Vol 30 Issue 3/21

8 R A D I O LO G Y Triple celebrations in Sw RTI: Ensuring X-ray quality for 40 years The Swedish RTI Group reports three notable milestones in 2021, including its 40th year in specialised X-ray safety manufacturing. In the early ’90s, during Sturesson’s studies in Medical Physics at Sahlgrenska University Hospital, Gothenburg, he was working on dosimetry in diagnostic radiology departments. ‘After reading about the X-ray developments at RTI, I applied for work there as part of my master’s thesis evaluating solid-state detectors and scintillating materials for radiation protection in medical X-ray. I graduated in ’95 and worked part-time at the University Hospital and at RTI, before joining the latter full time,’ bringing him understand- ing and schooling in clinical and manufacturing in dosimetry. Scatter probe from RTI In 1981, the DIGI-X model – the first commercially available multimeter to measure non-invasive kV – was launched. ‘Subsequent develop- ments have included world firsts, in both quality assurance (QA) hard- and software, of the Mini-X, PMX-II, Barracuda and oRTIgo,’ the firm reports. ‘Today, RTI’s trademark Piranha and Cobia meters are used by a plethora of hospitals, major manufacturers of X-ray equipment, service providers and government authorities worldwide.’ Early in 2021, RTI launched its own Scatter Probe, a combined leakage and scatter detector, an upgraded version of its industry- leading QA software Ocean. At the forefront of these key devel- opments in X-ray QA have been Olle Mattsson (Area Sales Manager Nordic & Rest of the World) and Sören Sturesson (Product Management & Technical Manager Calibrations), both RTI’s longest-serving “veter- ans”, now celebrating 25 years of service in the Gothenburg HQ Sales and R&D departments. What drew them to RTI? After graduating as an electrical engineer, Mattsson saw a RTI job advertisement. ‘Like many others, I’d never heard of the concept of meas- uring on X-ray machines non-inva- sively. RTI, therefore, sounded like a very open-minded, creative and forward-thinking company.’ The job interview was a resounding success. How has RTI’s ‘firsts’ impacted on the X-ray market and influ- enced your own work? ‘If RTI has been the first to do some- thing, most of my time in Sales has been spent explaining our unique solution,’ Mattsson said. ‘Customers are very keen to get the latest and best on offer. Thanks to RTI, the X-ray QA market has advanced sig- nificantly in user-friendliness and meeting new demands. Naturally, this will continue to improve in future.’ Research and development has been challenging, but working with generations of X-ray multimeters has been exciting, Sturesson point- ed out. ‘There have been increased demands for user-friendliness to meet the new technical inno- vations within diag- nostic imaging. When the company released DIGI-X in 1981, every X-ray tube/filter combina- tion was almost the same, whereas today there’s a large variety in X-ray spectra, nar- row-beam geometry.’ How did the X-ray industry respond to RTI’s break- through devel- opments? ‘Due to all its “firsts”, many times our competitors simply had to fol- low us,’ enthused Mattsson. ‘The Barracuda meter release in – since I joined. Since the early 2000s, beginning with the formation of RTI Group Inc. in the USA and, more recently, with Group APAC in Singapore, our presence and expo- sure in North America and Asia- Pacific also increased significantly.’ Sturesson: ‘The basics are still the same – safety and proper func- tionality of the X-ray equipment. The major change was when X-ray film was phased out and replaced with digital detectors, opening the door for major development of X-ray diagnostics, which has led to a wide range of applications with which the X-ray multimeters of today must comply.’ How will the increased implemen- tation of software shape future X-ray QA? ‘I don’t think anyone could fore- see how big the focus on software would be, nor how many differ- ent software programmes would be available today to assist users in so many ways,’ Mattsson noted. ‘Software has become essential, and will continue to grow to ensure compliance, easy management of measurements and the integration of other important data. Therefore, RTI has invested in its Academy training and online presence to increase accessibility of the Ocean software portfolio.’ ‘In the late ’80s, RTI launched oRTIgo, the first software tool for X-ray QA,’ Sturesson added. ‘This continued to develop and has now evolved to the Ocean software we offer today. Sören Sturesson, Product Management & Technical Manager Calibrations 2001 was a major paradigm shift for the X-ray multimeter market,’ Sturesson observed. ‘Our in-depth research and testing resulted in a meter without the shortcomings pre- sent in other non-invasive X-ray multimeters then available. Our com- petitors followed ever since, while we’ve continued to lead in the sim- plification of using our meters, not least with software tools like Ocean.’ How has X-ray testing technology changed in 25 years? ‘The X-ray machines and QA solu- tions available have moved the focus away from hardware towards dif- ferent software solutions to help the users,’ Mattsson explained. ‘Consequently, QA testing standards had to change due to new regula- tions in the X-ray market, often after the industry has already developed the new hardware to meet demands. ‘In RTI Sales, we have a far bigger and more active global distributor network – more than 100 contacts The Synthetic Brain Project Supercomputer creates 3-D brain models Scientists are using artificial intelligence and the Cambridge-1 supercomputer to synthesise artificial 3-D MRI images of human brains and create models that show disease states across various ages and genders, Mark Nicholls reports The Synthetic Brain Project is focused on building deep learning mod- els which have been developed by King’s College London (KCL) and NVIDIA data scientists and engineers as part of The London Medical Imaging & AI Centre for Value Based Healthcare. Computing power Dr Jorge Cardoso, Senior Lecturer in Artificial Medical Intelligence at KCL and CTO of The London AI centre for Value Based Healthcare, said that the Cambridge-1 supercomputer has the necessary computing power to train models of the size needed for this process, along with the speed to make development cycles fast and feasible. The AI model has learned what human brains look like by examining many images, Cardoso Synthetic brain model created by KCL and NVIDIA using Cambridge-1 supercomputer explained, with Cambridge-1 being used to teach AI models to gener- ate synthetic brain images from previous MRI brain scans from the UK Biobank and NHS. ‘We can then sample from this model by provid- ing the subject characteristics, such as age and sex, and retrieve the newly-generated images. ‘This solves the problem of scarce data, as we can generate as much data as we want, and also the prob- lem of privacy, because we can learn from synthetic data rather than real data. ‘As the models learn to gener- ate brains with/without pathologies, and in young and old age, we can then study these images to better understand the effects of pathology and ageing to give us a better under- standing what actually changes in these brains.’ Disease diagnosis This, he added, will help to diag- nose neurological diseases, such as stroke, dementia, brain cancer and multiple sclerosis, based on brain MRI scans – or predict diseases that a brain may develop over time and enable preventive treatment. Researchers hope that by generat- ing an infinite number of images with/without pathology from many ages, genders and ethnicities, they can then train other AI algorithms to learn to identify which images are pathological and which are not. Cambridge-1 enables accelerated generation of synthetic data that offers an understanding of how different factors affect the brain, anatomy, and pathology. Scientists can ‘ask’ the models to generate an almost infinite amount of data, with prescribed ages and diseases and, from there, help tackle problems such as how diseases affect the brain and when abnor- malities might exist. ‘This could mean better and earlier diagno- sis, which would be naturally ben- eficial to most patients, because early treatment commonly improves outcomes,’ Cardoso pointed out. ‘A better understanding of how dis- eases affect the brain can also help develop new therapeutics.’ Data availability The Cambridge-1 supercomputer has been pivotal in this aim, because the previous biggest computers – albeit very substantial in computing power – could not scale to the AI model size needed to produce good quality data. Data availability can be a major issue, but Cambridge-1 is helping to solve that via its massive com- putational power to generate the large synthetic datasets for research development. and algorithm EUROPEAN HOSPITAL Vol 30 Issue 3/21

1 0 R A D I O LO G Y Reduce MRI scanner anxiety Float off into virtual reality Claustrophobia or anxiety can overwhelm small children and people with cognitive difficulties, especially in a confining and noisy MRI scanner tube. Their restless reactions can then ren- der scan images useless. Report: Mark Nicholls To help such patients to relax dur- ing scanning, a team from King’s College London (KCL) has designed an immersive environment with a special virtual reality (VR) headset for use with MRI scanners. Christmas gift inspiration The idea for this distracting system originated when Dr Tomoki Arichi, a Clinician Scientist and Clinical Senior Lecturer in the KCL Centre for the Developing Brain, presented Jo Hajnal, Professor of Imaging Science in the Centre, a Christmas gift of VR goggles. ‘Professor Hajnal real- ised that whilst using the goggles, the strong immersive experience meant he was completely unaware of what was going on around him and this inspired us to create similar technology for use inside an MRI scanner,’ lead researcher Dr Kun Qian explained. ‘We then went on to develop the technology through a multi-disciplinary collaboration of colleagues with a background in computer science, engineering, opti- cal science, clinical and physics.’ Qian, a research associate in the Centre for the Developing Brain at KCL’s School of Biomedical Engineering & Imaging Sciences, acknowledged that having an MRI scan can be an ‘alien experience’, with scanning failure rates in under five-year-olds as high as 50 percent. Time and resource wastage can delay clinical decision making, he pointed out. ‘Even in those children who manage to complete the scan, movement artefacts can significantly degrade image quality and affect interpretation. Whilst this can be alleviated with sedation or anaes- thesia, these are resource heavy and carry their own associated clinical risks.’ The researchers had observed that, when fully immersed in a virtual reality environment, people become unaware of what occurs in real- ity around them. Thus, they devel- oped a system to replace a patient’s actual visual scene when having a scan with tailored VR content that n o d n o L e g e l l o C s ‘ g n i K © The VR headset can be used safely inside an MRI scanner. The device is designed to be light tight, so that the user cannot see their surrounding environment at all. is immersive and interactive, Qian said. ‘This removes the sensation of being inside an MRI scanner. We believe that it can potentially benefit all groups having a MRI scan, but in particular will help vulnerable populations with anxiety problems, and those with cognitive difficulties.’ The key component This is a custom-built 3-D printed MR-compatible VR headset; a pair of MR-compatible cameras (for eye tracking) and a MR-compatible pro- jector for VR content display, which can be set up and removed between patients. The research team also designed the system to precisely match exist- ing scanner hardware – in the case of King’s College London, the head coil in a Philips Achieva 3T system. With everything connected to a standard PC in the control room, patients control the system by using their eyes. They also can communi- cate with a carer outside the scan- ner. The aim is for the system to transform the MR examination from a passive and isolating experience 30-second MRI scan to estimate placental health Detecting early signs of pre-eclampsia From routine MRI scans, a machine learning algorithm is being used to model the normal distribution of placental tissue properties and show deviations from normal placental ageing and signs of pregnan- cy complications. To identify adverse conditions in preg- nancy, researchers at the School of Biomedical Engineering & Imaging Sciences at King’s College London (KCL), united with the women’s health department at the city’s St Thomas’ Hospital, where the emerg- ing technique was devised in col- laboration with obstetricians, physi- cists, radiologists, midwives, engi- neers and pathologists. Patients are scanned with a T2* sequence MRI, with imaging data fed into the neural network pipeline, which rapidly performs automatic segmentation before a regression model delivers a placental health score. The scan, combined with the automatic pipeline, can be included as part of any foetal MRI examina- tion but has the benefit of acquiring additional information not available under routine obstetric screening using ultrasound. Lead researcher Dr Maximilian Pietsch explained that the method is most sensitive early in the second half of pregnancy – also typically when the onset of pre-eclampsia begins – thus allowing early detec- tion and monitoring. The method Overlay of a placental T2* map (functional MRI) on an anatomical MRI image. Bright yellow indicates high oxygen content, dark red low oxygen content. develops a clinical marker and auto- mates its extraction, which would otherwise be prohibitively time-con- suming in clinical practice. Oxygenation levels Research Fellow Dr Jana Hutter said the approach may provide addi- tional information to ultrasound, Anatomical MRI image of the placenta (left side) and the foetus (right side) in the 2nd trimester which cannot acquire information from inside the placenta. ‘We use MRI because it has no ionising radiation, so is completely safe to use in pregnancy, and the T2* imag- into an active, engaging and interac- tive one. ‘Our VR headset is designed to be entirely light tight, so that the user cannot see their surrounding envi- ronment and is unaware of visual reminders of their position inside the MRI scanner,’ Qian added. Immersive content With the patient positioned in the scanner, the projector’s sys- tem immediately shows immersive 3-D images which appear to sur- round the user. This continues till the examination ends. ‘Crucially, it means that a patient may be com- pletely unaware of their surround- ings during their whole time inside the scanner,’ said Qian. The research team has incorpo- rated real-life sensory experiences into the VR environment, such as the visual scene moving when the examination table moves, or animat- ed characters responding to scanner noise. Using their eyes, the user can navigate through the virtual world to select content, such as films or games, and initiate, or terminate, a video link to their carer, meaning they can interact at any time dur- ing the scan with a companion via a webcam with microphone and a display monitor. Value in studies of cognitive processes While the proposed system has clinical applications for the popula- tions who find MRI stressful, the team believe it also has potential for neuroscience as a platform for MRI experiments studying fundamental Dr Jana Hutter is a Senior Research Fellow at KCL and UKRI Future Leaders Fellow based at St Thomas’ Hospital. After gaining her PhD in Erlangen, Germany, she moved to KCL, where she now leads her own research to focus on functional foetal and neonatal MRI. ing technique allows us to look at oxygenation in the placenta.’ Chronic hypertension in pregnan- cy is increasingly common, raising the chance of pre-eclampsia. ‘This technique allows us to look for this before any clinical symp- toms start in this high-risk popula- tion,’ Hutter said. ‘This gives obste- tricians additional valuable informa- tion to closely monitor the mother and baby to optimise the time point of the birth – late enough for the baby but early enough to avoid any effects to the mother.’ The technique may assess the health of an unborn child and the mother, so as to intro- duce interventions to prevent pos- sible complications from growing more serious. Other applications could be to identify evidence of foetal birth restriction, in which a baby does not grow as big as it could, plus late stillbirth, and other complications, such as gestational diabetes. EUROPEAN HOSPITAL Vol 30 Issue 3/21

1 2 R A D I O LO G Y Advertisment Continued from page 11 the of Radiology at Thomas Jefferson University Hospital, Philadelphia, delivered ‘Serendipity and Unintended Consequences’ lecture, it reflected the visionary perspective of the late PACS pioneer Professor Samuel J Dwyer III, whose seminal research on the retrieval rates of radiology exams formed the basis for PACS storage systems. Flanders likened the surge in AI tool development, and image data set expansion, to the 1960s space race, which launched the digital age, and to the global campaign to end the use of DDT. The latter produced a negative unin- tended consequence: injury to wildlife was reduced, including mosquitoes, which then caused a large increase in cases of malaria. The promise of AI technology to radiology This is huge – potentially it could improve diagnostic quality, to effi- ciently extend access and be a pana- cea to major cost reductions in medi- cine. Flanders referenced experts who predict that healthcare AI overall will be a US$36 billion industry by 2025, generating a US$150 billion annual savings in AI-enabled healthcare by or Europe. Is there enough rigor- ous evaluation? Flanders thinks not, nor do many other radiologists now speaking out publicly. from Researchers Radboud University Medical Centre in Nijmegen, the Netherlands, evaluated 100 CE-marked AI software products for clinical radiology from 54 vendors listed in www.aiforradiology.com. They reported in European Radiology that sixty-four products had no peer-reviewed evidence of efficacy, and that only eighteen of the hundred had evidence of efficacy rated as 3/6 or higher, validating impact on diag- nostic decision-making, patient out- come or cost. International standards for performance are needed Between January and October 2020, 320 published papers described new machine learning-based models to detect Covid-19 in chest CT images and chest radiographs. Writing in Nature Machine Intelligence, research- ers from Cambridge University report- ed that none of these models were of potential clinical use due to methodo- logical flaws and/or underlying biases. Most failed to describe a reproduc- ible methodology or failed to follow l a t i p s o H y t i s r e v i n U n o s r e f f e J s a m o h T © Adam E Flanders MD, Professor of Radiology and Rehabilitation Medicine at Thomas Jefferson University Hospital in Philadelphia, is co-director of neuroradiol- ogy/ENT radiology and vice chair of imag- ing informatics, also chairing the Imaging Informatics Council for the Enterprise Radiology and Imaging Service Line for Jefferson. image data sharing and collaboration. One example is the Medical Imaging and Data Resource Center (MIDRC), a consortium formed by the American College of Radiology (ACR), the Radiological Society of North America (RSNA), and the American Association of Physics in Medicine (AAPM) for image collection, curation and distri- bution of Covid-19-related imaging data for AI research. Left: Peer-reviewed articles were present for 36 of the 100 commercially available AI products; Right: Levels of efficacy of the included papers. The search strategy yielded 239 peer-reviewed publications on the efficacy of 36 out of 100 commercially available AI products. A single paper could address multiple levels. (Image source: van Leeuwen et al., European Radiology 2021) 2026. HealthITAnalytics reported, in 2019, that the top three areas of healthcare AI investment were robot- assisted surgery ($40 billion), virtual nursing assistants ($20 billion), and administrative workflow ($18 billion). By comparison, AI for preliminary diagnosis and automated image diag- nosis ranked at the bottom of its list, at $5 billion and $3 billion respectively.’ Artificial intelligence is fun- damentally limited Nonetheless, there has been a tre- mendous surge in interest in AI in radiology. ‘But is diagnostic AI on the right track? Is it happening too fast? Are image databases used to train AI software diverse enough, inclusive enough? Is there enough information?’ Flanders asked. ‘AI needs guardrails. It doesn’t know when it doesn’t have enough information when making a decision.’ Flanders cited an example of a brain CT angiogram, whereby the AI interpretation was a technical error of insufficient contrast mate- rial when, in fact, the patient was brain dead. While a radiologist can infer the correct diagnosis in this instance, it underscores the fact that AI is fundamentally limited to enti- ties for which it is trained. As of 2021, 222 AI medical devices were cleared by the FDA. Of these, 129 were for radiology applications, with 77% manufactured by small companies. But there is no agreed definition of an AI device or specific regulatory pathway, either in the USA best practice for ML model develop- ment, and/or failed to show sufficient external validation to justify wider applicability of the method. Flanders said a recent editorial in Radiology ‘Artificial Intelligence of Covid-19 Imaging: A Hammer in Search of a Nail’ was even blunter. ‘To a large extent, the large quantity and rapid publication of articles on AI for Covid-19 are emblematic of current trends in other areas of radiology AI. ‘It’s now so much easier to design and conduct a radiology AI experi- ment. The only prerequisite seems to be possession of a large data set,’ observed Ronald M Summers, MD PhD, of the National Institutes of Health Clinical Center’s Imaging Biomarkers and Computer-Aided Diagnosis Laboratory of Bethesda, Maryland. ‘International standards to enable meaningful comparisons of the per- formance are needed for AI software, emphasised Flanders. ‘Prospective outcomes studies are necessary to determine whether use of AI leads to changes in patient care, shortened hospitalisations, and reduced morbid- ity and mortality.’ Governments support image data collection Access to data is critical, but has historically been monolithic, isolated, and difficult. Governments are now starting to support image data col- laboration. Covid-19 brought a huge opportunity, because significant R&D funds became available to facilitate The world’s largest open Covid-19 images database MIDRC supports 12 internal Covid- related research projects. Its reposi- tory is expected to be the largest open database of anonymised Covid-19 me- dical images and associated clinical data globally. ‘There’s no organ system in the body not affected by Covid-19. I predict this image repository, which welcomes image data contributions from throughout the world, will be of tremendous benefit to AI research and innovations for radiol- ogy,’ said Flanders. Continuing AI radiology research He advises contribution of data to public data repositories to mitigate bias and strength diversity in AI research and development. He rec- ommends promotion of new interna- tional evaluation checklists for test- ing, and radiologists should retain an appropriate level of scepticism when assessing products for clinical use, and when purchasing radiology artifical intelligence tools, mandate vendors to provide a complete account of performance. ‘Most importantly, remain engaged in this very exciting, accelerating, and ever-changing process. ‘Just as radiologists’ input helped to perfect PACS to become the utilitarian technology it now is, most likely they will do the same for radi- ology artificial intelligence tools as they become ubiquitous.’ EUROPEAN HOSPITAL Vol 30 Issue 3/21 Comfortable and secure: The swivelling handle system helps patients to position themselves on the examination table with as little assistance as possible – even when transferring from a wheelchair. Febromed provides support for everyday radiology work Safe and hygienic: „get up“ Whether in magnetic reso- “get up” by FEBROMED helps nance imaging (MRI), com- reduce this strain in day-to-day puter tomography (CT), diag- work to a minimum. nostic radiology, or radiation therapy: in radiology, it’s all Safety isn’t just a handling is- about the details. Highly spe- sue; everything has also been cialised equipment in a pro- carefully considered in terms fessional environment helps of hygiene. The “get up” handle to make exact diagnoses and system by FEBROMED is easy to implement precise therapies. disinfect and meets the highest Everything else needs to be hygienic requirements of a me- just as professional. dical environment. The material is extremely durable, an invest- We at FEBROMED can help ment in safety but also in cost with daily operations with the efficiency. “get up” handle system. With our handle system, patients can perfectly position themsel- ves on the examination table at their own pace. This takes the pressure off of medical person- nel, letting them concentrate on what’s essential: the examinati- on. Reducing workload In radiology, a large section of patients have restricted mobi- lity, something that can often present a considerable strain for medical staff. They have to use their full physical strength to move patients – and are in danger of becoming a medical emergency themselves. But be- sides the physical complaints, there are also costs for the em- ployer and the social system. The “get up” handle system can not only be mounted on the ceiling, but also on the wall. Photos: Febromed www.febromed.de Febromed GmbH & Co. KG  Am Landhagen 52 | D-59302 Oelde Telephone Telefax E-Mail    0049 2522 92 01 920 0049 2522 92 01 919 info@febromed.de

1 4 D I G I TA L PAT H O LO G Y Genomics, transcriptomics, proteomics, metabolomics Molecular profiling boosts personalisation ing. When discussing personalised laboratory medicine in genomics, he emphasises that it is important to clearly define the genome variation. Germline DNA from individuals will ease this pre-emptive approach while, in cancer patients, personal- ised laboratory medicine must focus on DNA from the tumour. For patients with a tumour, where Next Generation Sequencing is cur- rently mainly conducted only in cancer relapses, he suggests that, within the next decade, this will be a routine pre-emptive pharmacoge- netic test. Better treatment and drugs that work To Schwab, the benefits of per- sonalised laboratory medicine are clear. ‘It’s about better treatment and selective drugs that will work if you carry a specific mutation, meaning patients will receive individualised To avoid adverse reactions, personalised laboratory medicine can help to predict a patient’s drug response. Investigations based on DNA and other omics technologies – e.g. genomics, transcriptomics, pro- teomics, metabolomics – along with microarray technologies, is making a particularly valuable contribution to cancer care, in which personalised approaches are becoming possible through molecular profiling. Report: Mark Nicholls Matthias Schwab, Professor and Chair of Clinical Pharmacology at the University of Tübingen, who also heads the Dr Margarete Fischer-Bosch Institute of Clinical Pharmacology in Stuttgart, Germany, believes that personalised labora- tory medicine will have an increas- ingly important role in patient care over the next decade. However, he emphasises that the technique must be used in line with guidelines and recommenda- tions – and where there is a clear consequence to the action, such as in drug selection or adjustment of the dose. Cancer drugs and pharma- cogenomics This is particularly important in drugs that have a narrow therapeu- tic window, where the dose used is close to that which may result in drug-related toxicity. ‘Personalised laboratory medicine in healthcare Figure adapted acc. to Emami-Riedmaier A, Schaeffeler E, Nies AT, Mörike K, Schwab M. Stratified medicine for the use of antidiabetic medication in treatment of type II diabetes and cancer: where do we go from here? J Intern Med. 2015 Feb;277(2):235-247 laboratory medicine also has value in research, by using consented information from a patient’s DNA to answer specific research questions.’ Latest trends and applica- tions Trends and applications in personal- ised laboratory medicine involve the application of newer drugs in cancer therapy, such as epidermal growth factor receptor (EGFR) inhibitors in lung cancer and can support genetic testing in therapy selection. With around 15 different drug targets (e.g. genes) and genetic vari- ation, the technology can help cli- nicians select the right drug and dose in a process mostly conducted together with a pathologist using lung tissue. Schwab feels this is the most promising application, but he fore- sees a move towards pre-emptive testing, not only in cases of drug failure after conventional cancer therapy, but also at the onset of therapy, where genetic information will be available when prescrib- Matthias Schwab & Elke Schaeffeler. Pharmacogenomics: a key component of personalized therapy Genome Medicine volume 4, Article number: 93 (2012) is currently mostly implemented for treatment of oncology patients to avoid adverse drug reactions,’ he explained, adding that, in some areas, such as colorectal cancer, the approach is clearly defined – and where drugs (e.g. 5-Fluorouracil) to be used in clinical practice should be laboratory tested with a person- alised medicine approach, for exam- ple to prevent haematoxicity and/or gastrointestinal side effects. ‘ Personalised laboratory medicine – a crucial step towards correct doseage ‘Personalised laboratory medicine is a crucial step in ensuring the dose is correct, that the treatment is correct, and matches the patient’s DNA, genetic profile and condition,’ Schwab emphasised. ‘Guidelines for healthcare imple- mentation, that are regularly updat- ed and publicly available, are there for doctors to select the right drug and adjust dose, or in some cases choose another drug. Personalised Matthias Schwab is Professor and Chair of Clinical Pharmacology and heads the Dr Margarete Fischer-Bosch Institute of Clinical Pharmacology in Stuttgart and the Department of Clinical Pharmacology, University Hospital, Tübingen, Germany. His major clinical and research activities are personalised medicine and pharma- cogenomics with focus on cancer. treatment and not suffer toxicity.’ Such a medical approach prom- ises more effective diagnostics with more effective and safer treatment, as well as faster recovery and resto- ration of health and improved cost effectiveness. ‘Personalised medicine in the genomics area is the future,’ Professor Schwab concluded, but, he added, ‘in the next few years we will also talk about liquid biopsies because we can also use this infor- mation for epigenetic or metabo- lomic markers.’ ERBB2 gene amplification status Continued from page 13 instead of annotation by an expert. ‘This also allowed us to create high amounts of training data in a very short time and we showed, in a validation set of 250 images, including immune cells, that the accuracy on a pixel level was very high, which shows that this method should work,’ he added. However, the big question remains as to whether AI is going to be just another subjective expert. ‘How you select and annotate training data is crucial, especially for expert-super- vised algorithms,’ Lundin continued. ‘Ground truth is provided by human observations and will always be subjective. To get more reliable ground truth, should we,’ he ques- tioned, ‘use alternative and more objective endpoints than the expert annotations? ‘Maybe it’s time for a paradigm shift, as most of the algorithms so far created for pathology are based on expert interpretation. The whole idea of performing those are often that we want to make a prediction, a diagnosis, but that could also be a prediction of survival, risk of cancer recurrence and evaluation of the treatment efficacy.’ Learning with 1,300 breast cancer patients Lundin’s question was whether a middle step of human annotation is needed, or could that be replaced with machine learning, going direct- ly from the sample to the predic- tion? He detailed how this has been examined by his team for direct out- come supervised learning with 420 colorectal cancer (CRC) and 1,300 breast cancer patients using micro- arrays where the long-term outcome was known. Pathologists performed a risk prediction visually before the images were put through a neural network trained to directly pre- dict outcome. High discrimination was achieved using this outcome- supervised learning and in CRC, it was better than the consensus score of three experienced pathologists, demonstrating that valuable infor- mation can be extracted from the tissue. Further, the centre’s ERBB2 study of breast cancer patients was based on nationwide data from Finland, and saw the team train a set of 700 samples where they knew the ERBB2 status. Having shown it is possible to predict ERBB2 status from the morphology with super- vised learning, this was validated in samples from a randomised trial study on the effect of Trastumuzab treatment. Here, the algorithm predicted ERBB2 status from the morphology and also discriminate the patients who received Trastumuzab into two different prognostically different groups. ‘This is among the first studies to show that an improved efficacy of a molecularly targeted therapy can also be predicted from the morphology based on features learned by AI,’ the professor noted. ‘From these studies, we see that, using morphology and patient status as endpoints, you can clearly predict the survival and do it at the level of a highly-trained expert, and in some circumstances even better.’ n r o b r i S f l U © Johan Lundin is Research Director at the Institute for Molecular Medicine Finland (FIMM), University of Helsinki, Finland and Professor of Medical Technology at the Department of Global Public Health at the Karolinska Institute, Stockholm, Sweden. He is also co-found- er of Aiforia Technologies, a FIMM spin- off company developing a cloud-based platform for AI-supported digital diag- nostics within pathology. EUROPEAN HOSPITAL Vol 30 Issue 3/21

1 6 L A B O R AT O RY Lab interoperability is essential Demand for molecular LIMS increases Fast, flexible laboratory information management systems (LIMS) that cope with data and workflow com- plexities of molecular and genetic testing now work in laboratories internationally. Here, in the first in a new Lab Pinnacle Series, experts from the CliniSys Group, Sunquest Information Systems and Data Innovations (all owned by Roper Technologies), discuss the value of a LIMS in molecular and genetic testing. flows,’ said Filip Migom (CliniSys/ MIPS), adding, ‘within the human molecular testing domain, if your focus is testing with yes/no results, a traditional LIS can handle that. However, if you want to do more complex procedures, it’s better to manage this outside the standard LIS.’ ‘Techniques change rapidly, but generic approaches are possible, as done in Europe with our GLIMS. We’ve built a special purpose mod- ule, GLIMS Genetics, to support the flexible workflows needed in those environments. Instead of dedi- cated haematology or biochemistry modules, we have a more general approach, so we can support some complex workflows.’ the instrument technique and report- ing code. Middleware plays a key role in facilitating this level of integration and interoperability.’ When should molecular testing be brought in house and LIMS be implemented? ‘For private labs, when it can be profit- able,’ answered John Lebon (CliniSys/ MIPS). ‘Plan early,’ he advised; vol- umes grow beyond the capabilities of the current LIS. Interfacing with vari- ant analysing systems within health- care systems is difficult and often performed manually. Shirley Li agreed, profitability is the primary driver in the USA – i.e. scalability and reimbursement above you already have LIMS or other soft- ware that provides a process to man- age testing.’ An existing good volume of testing also underlines a need to introduce a LIMS. Steve Abbs: ‘Since the first big genetics push in early 2000s and still today, many UK labs use NHS funding to purchase LIMS. Under the current national genomic medicine initiative, labs are expected to deliver change by standardising and indus- trialising testing. To that end, NHS funding is available for purpose-built LIMS technology.’ Could LIMS take molecular/genet- ic testing beyond workflow into analysis, particularly for NGS? Shirley Li: ‘Molecular and genetic testing involves large volumes of complex data – especially true with NGS, where a typical whole exome sequencing identifies around 10,000 Middleware ‘Across the board, interfacing gives you efficiency, quality, and scalabil- ity,’ said Carol Beth Casto (Data Innovations). ‘For molecular and genetic testing, instrument interfac- ing is in its infancy. Different molecu- lar and genomics labs have differ- ent workflows. At Data Innovations, we are making inroads with work- flow variability by establishing that standard, lower-level protocol for the output, and able to tweak here and there. We work with instrument vendors and LIMS systems to pro- vide standardised translation from the system to the instrument and back with our Instrument Manager middleware.’ ‘In Europe, test orders come in and results go out through different systems with different pro- tocols, formats and languages,’ Filip Migom explained. ‘Middleware ena- bles our GLIMS to manage all those variables. We must translate all types of requests to standardise within our database, and then we must translate again for the reporting and business side. In the middle, we must also translate all the interoperability with the instruments because performing a download of a plate depends on Automation of processes and cross- discipline data availability can produce maximum efficiency cost. Looking at NGS, about two- thirds of lab costs are for reagents. Additional efficiencies in staff time and resource allocation are worth- while. ‘Moving from paper-based, offline systems into something like Sunquest Mitogen LIMS helps under- stand where lab costs are and how to improve cost analysis measures.’ Holding power within an institution ‘Testing within the system, sometimes even before it’s profitable, can be bet- ter for the overall business strategy than sending samples out to a pri- vate lab,’ Stéphane Decap reasoned. Various diseases are tested in differ- ent lab specialties, such as pathology and genetics. In a one lab system, using a sophisticated LIMS to auto- mate the workflow, processes can be streamlined, with higher volumes and fewer people. Laura Voeghtly: ‘If a testing method emerges that drastically reduces lab costs, labs will often make the case to bring it in – that’s easier to do when variants per patient, per sequencing run. Doing this manually carries risks error and data loss, and challenges information filtering. Today’s LIMS help data files and assets move across the lab, corresponding to the testing and ultimately being translated for the different analysis pipelines labo- ratories use.’ Steve Abbs: ‘The UK has independ- ent bioinformatics and variant inter- pretation software providers; a cur- rent priority is interfacing those solu- tions with the LIMS. With the bioin- formatics process being so complex, interfacing is necessary to get that output into the LIMS and ultimately into the diagnostic report, which then goes into the electronic patient record. For more basic genomic tests that require a simple genotype and produce a positive/negative result, LIMS users can apply rules-based logic to automatically generate a report based on the result.’ ‘Inside the LIMS, we collect a lot of data about the phenotype, clini- cal information and family history, which can be very useful for NGS techniques,’ added Stéphane Decap. To manage all that data, connec- tivity must be beyond the instru- ‘When precision medicine initiatives gained momentum in the USA,’ observed Laura Voeghtly (Sunquest), clinicians focused on molecular test- ing, which was new to many of them. Thus, higher volume and greater test complexity needed support, and clinicians needed to know the diag- nostic, prognostic, and therapeutic implications of test results. Thus, LIMS automation, instrument integra- tion and general laboratory interop- erability became essential.’ To deliver a genomic medicine service, NHS England commissioned seven Genomic Laboratory Hubs (GLHs) and developed a National Genomic Test Directory for GLHs. ‘The initiative includes goals for standardisation, equity of access to testing, economies of scale, and volume growth that are collectively increasing demand for well-integrat- ed LIMS technology,’ said Steve Abbs (CliniSys). ‘NHS labs also do a lot of genetic testing for infectious disease,’ Tony Oliver (CliniSys) pointed out. ‘Just think of Covid-19 to grasp the volume of molecular PCR amplifi- cation techniques and sequencing events. Those are generally done in metropolitan teaching hospitals, but with the commoditisation of molecu- lar testing, we see a greater need for specialised LIMS technology.’ Stéphane Decap (CliniSys/MIPS) pointed to a sharper distinction between molecular and genetics tests in the EU. Molecular techniques are routine for human genetics and by other labs, for example for virol- ogy. University hospitals mostly cover human genetics testing but genetics markets vary greatly between coun- tries, e.g. in Germany many private labs do human genetics tests, but France has only a few, and only two genomics hubs do full genomics testing. ‘The need for GLIMS – our lab information system that supports genetics – has not been as high in France or Belgium; but, as those countries follow the UK’s increased testing, demand will grow. Genetics and genomics tests are very complex and specialised, which calls for spe- cialised technology that integrates with all necessary databases and sys- tems from order to invoicing.’ LIS or LIMS? ‘LIS are built to accommodate clini- cal pathology, so they can be fairly rigid in their database structures,’ Shirley Li (Sunquest) underlined. ‘With molecular and genetic testing, applying different ways to sequence new modules requires a flexible, agile workflow solution. The applica- tion of different assay technologies varies among laboratories and even for a given instrument there can be differences in analysis and data inter- pretation needs.’ Next generation sequencing (NGS) This needs more sophisticated analy- sis for bioinformatics, which can also differ from one lab to another. LIMS can be key to supporting the right level of configurability and modular- ity for labs to define their own plans and strategies in this field.’ ‘We can work with middleware partners to support molecular work- ment, beyond the wet lab. ‘It needs to connect to a Human Phenotype Ontology (HPO) database to manage the NGS information, and needs to connect to the variant analysis infor- matic pipeline.’ Covid-19 challenges ‘Many existing interfaces were avail- able for PCR instruments that were connected directly to our LIMS,’ said John Lebon. ‘But suddenly every lab bought new instruments and wanted to increase PCR capacity significantly. Our team made it work through remote installation and a lot of logis- tical problem-solving.’ Reagents were not available for the instruments they had, noted Tony Oliver, thus labs bought many additional instruments, resulting in an exploding need for interfacing in the UK. ‘We deployed to data centres, but cybersecurity regulations make processing text files problematic when data-centre- deployed interfaces are involved. The sudden surge in new labs and instru- ments caused a surging need for text file interfaces. That was really the only notable technical challenge we faced in the UK.’ Interfacing demand also spiked in the USA, Laura Voeghtly said – not just for new instruments but also for existing platforms as new reagents received emergency use authorisa- tion (EUA). ‘Standard regulations were not yet in place for these EUAs, creating a “wild west” situation with many moving parts – particularly as we worked to support the many Covid-19 labs that were entirely new to molecular testing.’ Pop-up labs ‘The Belgian government created a pop-up lab for faster testing of the entire population,’ added Filip Migom. ‘Everything was centralised and pre- labelled with barcodes through the LIS, but the lab had many different instruments using a range of tech- niques for Covid-19 testing. In a normal lab workflow, you know in advance which instrument you will use, but the LIS is not designed for molecular testing workflow and does not know upfront which instrument to use for any given plate. Every instrument has its own plate structure format, so an urgent need arose to select and correctly upload a file for the plate configuration. We were all very motivated; we achieved it, but it was a big challenge.’ ‘We had to be nimble to sup- port USA pop-up labs,’ said Laura Voeghtly. ‘We encountered two main types: (1) pop-up labs in health sys- tems and physician practices focused on acquiring specimens and clients from across the region and country and (2) brand new laboratories at academic institutions to test students, faculty and staff. For healthcare organisations we provided quick integration with our ordering process through Sunquest Atlas, providing a physician portal to acquire countrywide specimens. For the academic institutions, we managed the integration of necessary components to support Covid-19 test- ing for new and potentially inexperi- enced labs.’ Prediction: Next Generation Sequencing may become the highest diagnostic yield assay adopted by labs. EUROPEAN HOSPITAL Vol 30 Issue 3/21

1 8 CA R D I O LO G Y AI elevates imaging of coronary arteries Coronary plaque scanned in six seconds Coronary computed tomography angiography (CCTA) is a non- invasive imaging test which can be used to evaluate coronary artery stenosis and measure plaques. Current plaque analysis is time-consuming and needs expert readers in order to help assess a patient’s heart attack risk. That’s about to change. Report: Mark Nicholls During a session focused on artificial intelligence in imaging, at the forth- coming virtual European Society of Cardiology Annual Congress 2021, Dr Andrew Lin will explain how using artificial neural networks, AI can generate automated predictions directly from medical image data. Healthcare in Europe spoke with the post-doctoral researcher about his work at the Biomedical Imaging Research Institute at Cedars-Sinai Medical Center, in L.A. California. Research aims? To develop a deep learning system for plaque quanti- fication from CCTA and compare it with expert readers and intravascu- lar ultrasound. A dataset of 921 patients was used to train an automated deep learning neural network to measure plaque volume and coronary artery steno- sis from CCTA images. The deep learning network was then applied to an independent test set of 275 patients, where its diagnostic per- formance was evaluated. ‘There was excellent agreement between deep learning and expert readers for measurements of total plaque volume and coronary artery steno- sis,’ Lin confirmed. ‘Deep learning also performed in close agreement with intravascular ultrasound, which is the invasive gold standard for plaque measurements. Speed and high accuracy The time taken for deep learning analysis per scan was around six seconds, compared with 25-30 min- utes taken by experts.’ Deep learning offers significant promise for plaque quantification from routine CCTA images, by auto- matically detecting and contouring the heart arteries and underlying atherosclerotic plaques. This, Lin pointed out, enables measurements of plaque volume and coronary artery stenosis directly from stand- ard CCTA images, with high accu- racy compared to expert readers and at a fraction of the usual analy- sis time. Lin also noted that deep learning from CCTA agreed with the invasive reference standard of intravascu- lar ultrasound for measurements of Example of deep learning plaque segmentation. Left: Plaque in the left anterior descending artery on CCTA. Middle: deep learning segmentation of calcified plaque (yellow) and noncalcified plaque (red). Right: 3-D rendered view of the coronary tree showing the individual segment analysed by deep learning. Tackling continuous remote monitoring Continued from page 17 will transmit the large amounts of data quickly and in real time. Diverse expertise Diverse expertise is vital in the con- sortium, with HHI as a technical part- ner alongside technology companies, and the LVAD experience of DHZB. Dr Slawomir Stanczak, Head of the HHI Wireless Communication and Networks Department, stressed the obvious need to integrate data protection and security into the data transmission. This research is in the earliest stages. It will assess the tech- nology and compare conventional datasets with those offered by the 5G campus map, before leading to a patient study and a clinical trial. LVAD patients still need support from specialised centres for what remains a complex therapy and while remote monitoring of class 2 or class 3 heart failure patients is being conducted, the hope is that this technology will enable more severe heart failure patients to be moni- tored remotely, beyond the specialist Katrin Schreiter (KS) Dr Christina Czeschik (CC) Spain: Eduardo de la Sota (EdS) The Netherlands: Madeleine van de Wouw (MvW) USA: Cynthia E. Keen (CEK), Lisa Chamoff (LC) Subscriptions Dorothea Fleischer, Theodor-Althoff-Str. 45, 45133 Essen, Germany Subscription rate 4 issues: 32 Euro, Single copy: 8 Euro. Send order and cheque to: European Hospital Subscription Dept Printed by: WVD, Möhrfelden, Germany Publication frequency: quarterly Representatives China & Hongkong: Gavin Hua, Sun China Media Co, Ltd. Phone: +86-0755-81 324 036 E-Mail: 627416876@qq.com Germany, Austria, Switzerland: Ralf Mateblowski Phone: +49 6735 912 993 E-Mail: rm@european-hospital.com France, Italy, Spain: Eric Jund Phone: +33 493 58 77 43 E-Mail: jund@european-hospital.com Editor-in-Chief: Brenda Marsh (BM) Editorial team: Wolfgang Behrends (WB), Sonja Buske (SB) Senior Writer: Mark Nicholls (MN), Great Britain Mélisande Rouger (MR), Spain Publishing Director: Mareike Scholze (MaS) Executive Director: Daniela Zimmermann (DZ) Founded by Heinz-Jürgen Witzke ISSN 0942-9085 Correspondents Austria: Michael Krassnitzer (MK), Christian Pruszinsky (CP) France: Jane MacDougall (JMD) Germany: Cornelia Wels-Maug (CWM), Karoline Laarmann (KL), ECG INNOVATION MADE IN GERMANY THE NEW CARDIOPART 12 WLAN CONVINCING WORKFLOW WIRELESS, ROBUST, ULTRAMOBILE AND HYGIENIC The CardioPart 12 WLAN is a 12-lead resting ECG which fits into your pocket. Its compact design is equipped with a brilliant touch screen display and available on a 24/7 basis. Chargeable using the associated docking station. Learn more at www.amedtec.de. Dr Andrew Lin is a postdoctoral researcher in the Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, California, (Mentorship: Dr Damini Dey). His research focuses on the application of artificial intelligence and advanced analytics tech- niques to plaque imaging with CCTA. total plaque volume and minimum luminal area. Workflow integration He now believes a deep learning system that rapidly and accurately estimates coronary artery stenosis could potentially be implemented into clinical practice as a ‘second reader’ and clinical decision support tool. ‘By providing automated and objective results, deep learning may reduce inter-observer variability and interpretative error among physi- cians. The system also could be used to pre-screen CCTA scans, flagging patients with obstructive disease to be prioritised for clinical reporting.’ When integrated into clinical deci- sion-making, CCTA guides the use of preventive therapies, improves event-free survival, and reduces unnecessary downstream testing. AI-enabled image analysis has the potential to improve the efficiency of CCTA workflow and enhance patient risk stratification. centres, and improve total mortality outcomes. GB, Scandinavia, BeNeLux: Simon Kramer Phone: +31 180 6200 20 E-Mail: kramer@european-hospital.com Israel: Hannah Wizer, International Media Dep. of El-Ron Adv. & PR Co., Ltd., Phone: +972-3-6 955 367 E-Mail: ronin@netvision.net.il Taiwan: Charles Yang Phone: +886 4 232 236 33 E-Mail: medianet@ms13.hinet.net USA & Canada: Hanna Politis, Media International Phone: +1 301 869 66 10 E-Mail: hanna@media-intl.com Maria Kaiser Phone: +1 250 726 4007 E-Mail: mkads@mac.com All company, brand and product names in this publication are the property of their respective holders. Users must obtain permission from those holders before copying or using the owner’s trademarks, product and company names or logos. EUROPEAN HOSPITAL Vol 30 Issue 3/21

2 0 CA R D I O LO G Y Digital innovation in cardiology What’s new? Triage HF Plus, highlighted in the BCS conference session ‘Digital Innovation in Cardiology - What’s new?’ is a digital heart failure care project that uses a customised algorithm to detect early signs of dete- rioration in patients with implanted devices. ac implantable electronic devices, which use data from up to nine monitored physiological parameters to classify patients as low, medium, or high risk of HF hospitalisation in the next 30 days. Once enrolled, all high-risk episodes are managed via the Triage HF Plus pathway and patients received a structured tel- ephone assessment to assess both During her presentation ‘Digital solu- tions to identify worsening heart failure’, consultant cardiologist Dr Fozia Ahmed discussed the UK’s NHS heart failure burden (c. one million patients). In recent years, HF hospitalisation has risen by 33 per cent, three time faster than all other hospital admissions, she said, adding that predictions from 2010 indicated the UK would see a 50 per cent increase in HF admissions by 2035 – but this milestone was reached by 2018 in the Greater Manchester area. ‘That highlighted the need to do things differently,’ Ahmed noted, ‘to identify unstable or decompensating patients earlier, in order to improve outcomes, reduce admissions and decrease overall costs.’ The traditional approach was reac- tive, but using risk-based data from implanted devices to help identify decompensating patients sooner and intervene earlier to prevent a ‘prob- lem turning into a crisis’. Clinical alert The Manchester team developed Triage HF Plus: a remote manage- ment HF care pathway which com- bines the TriageHF risk score with a structured telephone-based remote clinical patient assessment. The validated Triage HF score is a feature of Medtronic cardi- HF and non-HF related acute issues. The data in the accompanying report summarises a patient’s gen- eral health, highlighting any sig- nificant reduction in activity levels, a change in heart rate or rhythm and increasing indicators of conges- tion. ‘Without seeing this patient, the remotely monitored data can provide an indication of whether the patient is less active, more sympto- matic, and in need of clinical review, just from the remotely-monitored data,’ Ahmed explained. ‘Combined with the telephone assessment, it’s then possible to obtain a more com- prehensive picture of how stable or unstable the patient is.’ The nurse- Health report showing abnormal data led system is patient-centric and has transformed care for more than 1,000 patients in the Manchester region. Other issues, such as acute respiratory infections and worsen- ing COPD (chronic obstructive pul- monary disease), have also been identified. The team has redesigned ser- vices and delivered real-time rapid triaging and automated specialist management in the home environ- ment. ‘By implementing this, we could reconfigure and redesign our clinical services in such a way that patients get assistance on the basis of clinical need between scheduled appointments,’ Ahmed explained. ‘The digital solution is hi-tech, low-labour and low-cost, designed so that it complements existing heart failure care pathways and can augment guideline-directed inter- ventions. It’s also Covid-proof and delivers end-to-end remote monitor- ing management and intervention.’ Risk prediction in cardiomyopathy Delegates also heard from Professor Declan O’Regan (Imperial College London), who spoke of future preci- sion phenotyping and risk predic- tion in cardiomyopathy, and the added value that machine learn- ing can bring to every stage of cardiovascular imaging. ‘That starts from the point of image acquisi- tion and reconstruction, to image enhancement like super resolution, to labelling images, and also iden- tifying pathology and new imaging biomarkers of disease, such as the concept of radiomics, and also mov- ing beyond conventional diagno- sis towards prediction tasks, which really drives management decisions Dr Fozia Ahmed is a consultant cardiolo- gist at the Manchester Heart Centre, part of Manchester Royal Infirmary, where she specialises in heart failure and cardiac devices. In December 2020 she received a major UK award for her outstanding contribution to heart failure services and excellence in HF care. Her research inter- ests include risk prediction models, remote monitoring in heart failure, and preven- tion of cardiovascular infection; with a focus on re-designing clinical pathways to improve patient outcomes. in the clinic.’ O’Regan covered three main topics – how computational imaging can be useful in cardiovas- cular medicine, ranging from inte- grating data, and bringing imag- ing and non-imaging data together; using AI and imaging for discov- ery science and understanding new mechanisms of heart disease; and for early diagnosis and risk predic- tion. Finally, on the vexing issue of overload, Professor Tim Chico (University of Sheffield), questioned and suggested how to cope with the amount of data being generated a in his presentation ‘A cardiovascular digital twin; the brain assist device we need to make sense of data overload?’ C-arm adds accuracy to cardiovascular interventions Flexibility and freedom of movement Alongside the worldwide increase in longevity, inevitably with chronic health conditions rising, cardiovascular procedures and OR utilisation have surged. Thus the efficient and safe performance of surgeries is even more important. tem temperature and avoid sys- tem failures due to overheating, it is equipped with Advanced Active Cooling. Additionally, it is the only system on the market to offer motor- isation of all four axes for easy control. ‘The system meets all the The intraoperative use of mobile C-arms increase, accuracy, improv- ing clinical outcomes, which can sig- nificantly reduce revision rates and thus overall healthcare spending, the manufacturer Ziehm reports, also noting: ‘Mobile X-ray imaging devices also have lower acquisition and installation costs, which results in a faster return on investment in comparison to fixed installed systems.’ A comprehensive mobile hybrid solution ‘The Ziehm Vision RFD Hybrid Edition* mobile C-arm designed for demanding interventional procedures. To ensure constant sys- is requirements to transform conven- tional ORs into hybrid rooms in no time,’ Ziehm continues. ‘With versa- tile display options, ceiling-mounted monitors, wireless solutions and a unique Usability Concept, it does not require any changes to the OR and is ready for use immediately – without extensive construction work. ‘Connectivity to 3-D vascular navi- gation systems and contrast injec- tors make it ideal for demanding hybrid procedures such as TAVI, angioplasties and EVAR. ‘Together with its French daughter company Therenva, Ziehm Imaging is investing in the future of intraoperative 3-D vascular navigation. Therenva’s mobile image fusion system EndoNaut ena- bles physicians to achieve better accuracy during challenging hybrid surgeries. Combining preoperative CT data with intraoperative imag- es from the mobile C-arm on the EndoNaut system reduces X-ray dos- age and contrast media usage for even more precise results. features such ‘Software as Enhanced Vessel Visualization with automatic colour display of vessels help to precisely define contours and side branches, facilitating com- munication in the OR.’ Powerful cardiovascu- lar imaging in a mobile CathLab ‘With the introduction of the most powerful generator on the market for mobile C-arms, the Ziehm Vision RFD Hybrid Edition with 30 kW** enables more clarity in cardiovas- cular imaging,’ the company points out. ‘Faster and sharper imaging, reduced motion artifacts and dedi- cated parameters allow more details to be displayed. Dedicated func- tions for coronary interventions and electrophysiology provide the best possible support during demanding procedures. ‘Furthermore, special display and transmission options suitable for cath labs are available. These have become well-known and established through many years of practice in hybrid rooms. ‘Together with its Dutch partner company, Fysicon, Ziehm Imaging is going one step further. To meet the needs of interventional cardiologists worldwide, they offer a dedicated mobile haemodynamic measure- ment station. The mobile CathLab solution provides more flexibility and freedom of movement and rep- resents an alternative to conven- tional setups.’ Reflecting on the use of Ziehm Vision RFD Hybrid Edition CMOSline in his mobile CathLab, Dr Rajaram Prasad concluded: ‘Our mobile con- cept has only advantages for me, my staff and my patients. I have not yet had a case that I could have solved better with a fixed system.’ * Ziehm Vision RFD Hybrid Edition represents a group of optional hard- ware and software that creates an option package on the device named Ziehm Vision RFD. ** 30 kW generator is available in combination with dedicated cardio packages. EUROPEAN HOSPITAL Vol 30 Issue 3/21