4 D I G I TA L PAT H O LO G Y Experience report AI implementation: Digital pathology switch bears fruit Fully digitising pathology operations has led to greater efficiency, cost savings, and quicker diagnosis for the Laboratory of Pathology East Netherlands (LabPON). The move, made six years ago, is showing measurable benefits and now, the institution is beginning to explore the potential of deep learning computational pathology algorithms, which might push the efficiency gains even further. Report: Mark Nicholls Details were outlined to the 8th Digital Pathology and AI Congress in London by Alexi Baidoshvili, Professor of Pathology, who spe- cialises in digital and computation- al pathology at LabPON. He report- ed that LabPON had completely digitised its diagnostic routines for clinical cases in July 2015 and has continued to evolve the technology and processes ever since. Additional milestones were the development and introduction of AI software for image analysis in 2018 and a new Image Management System (IMS) in 2021, with further AI develop- ment and integration planned for 2022. In his presentation, the expert discussed the investments neces- sary to introduce digital pathology. Areas to be considered include flow analysis in the laboratory, storage, scanners, the need for a robust IMS, training of staff, consultation workflows, software and enabling remote working. On average, setting up a laboratory for digital pathology requires at least 1.5 million euros per year, he calculated, with AI sys- tems not included. However, invest- ments might be a bit lower for insti- tutions which already have a robust IT infrastructure in place. ‘It is very important to choose the right IMS,’ he added, pointing out desirable features such as open architecture, integration with commonly used scanner types, robustness, scalabil- ity, support for external consultation and – not least – a user-friendly interface. Lung tissue with special Grocott staining and more efficient internal consulta- tions and better support for multidis- ciplinary teams (MDTs). Digitisation also facilitated remote consultations, improved access to information and images, and better means to organ- ise education and research. Data from LabPON backs up these perceived improvements: for exam- ple, an experienced digital patholo- gist can diagnose faster via digital methods than with the conventional microscope. Time savings are signif- icant, reported Baidoshvili, reducing the workload of pathologists and releasing time to perform other tasks, such as participation in multi- disciplinary meetings. The expert noted that all staff segments benefit from this; lab personnel, techni- cians, and pathologists. Additionally, the switch from subjective to objec- tive diagnosis led to savings on immunohistochemistry, improved service delivery to the clinic and patient, and yielded financial advan- tages. ‘The digital transition saves on archive time and the work of secretaries and of technicians,’ he added. Current and future AI benefits After transition to digital diag- nosis, and the seamless integra- tion of interoperable AI, Baidoshvili noted an ‘impressive’ accuracy of the system, for example in cancer detection and Gleason grading of cases. Referencing the lab’s expe- rience with prostate AI solutions Concentriq from Proscia and Galen from Ibex, he praised their seam- less and intuitive workflows, with initial results showing the potential Alexi Baidoshvili in Alexi Baidoshvili is a Professor of Pathology at DT Medical University, Georgia. He specialises digital computational pathology at and the Laboratory of Pathology East Netherlands (LabPON), an institution with 120 employees, 18 pathologists and 90 lab technicians. Furthermore, Prof Baidoshvili is an active board mem- ber of several international organisa- tions and organiser of conferences. At LabPON, he and his team are working on the development of various image recognition programs. to reduce immunohistochemistry costs, subject to further validation. The expert also pointed out the potential of automatic report gen- eration with AI assistance. ‘Thanks to AI, we can save 1:18 minutes of diagnostic time in every colon diagnosis.’ Applying this to the 500+ cases LabPON processed in 2020, Baidoshvili calculated more than 11 hours of saved diagnosis time for pathologists. Overall, the switch to digitisation had widespread benefits with much improved logistics, handy tools in IMS resulting in high efficiency, flexible and remote working, easy access to computer archives, better and more efficient diagnostic qual- ity and improved diagnostic logis- tics, leading to a happier pathology workforce. Lung tissue with Pas-D staining Safer, happier, more productive Despite these initial hurdles, Prof Baidoshvili found the benefits from going digital to be well worth the effort for LabPON, which handles 55,000 histology, 20,000 cytology, 10,000 molecular and 12,000 immu- nohistochemistry cases per year: ‘What we have achieved from our investments is high efficiency and quality improvement,’ he said. Working with glass slides is a thing of the past, reducing the burden on logistics, streamlining workflows, and opening new possibilities for external consultations. Pathologists’ workplaces also saw subsequent improvement, for exam- ple in enhanced safety, better ser- vice, and more flexibility, continued Baidoshvili. ‘In the future, there will be a move from subjective diagnosis to objective with AI applications.’ He emphasised how streamlined work- flows were important in creating a ‘happier workplace’, leading to quicker replies to inquiries, faster NPIC network Pathology in North England goes digital Reaching out across 15 hospitals and a population of six million people, a network established by the National Pathology Imaging Co-operative (NPIC) is an ambi- tious move to advance digital pathology in the north of England. The project, led by the Leeds Teaching Hospitals, will see digital pathology services operate across a network of over 20 NHS trusts and 238 pathologists, scanning over 2.4 million images and generating 3 Petabytes of image data per year. In addition, 48 scanners will feed into a single, scalable, vendor neutral archive (VNA) with DICOM compat- ibility. Progress on the network was outlined by NPIC operations direc- tor, Dr Daljeet Bansal, at the 8th Digital Pathology and AI congress in London. The aspiration of NPIC is to cre- ate a globally leading infrastructure for digital pathology and AI. The groundwork for this has been laid as far back as September 2018, with the complete switch to digitisation of slides at the Leeds Teaching Hospitals, and the introduction of next generation scanners in 2021. Dr Bansal explained that NPIC’s aims are to drive clinical use of digi- tal pathology, create an environment for development and evaluation of AI and support further research and innovation. The project – which involves additional academic and industry partners working with cli- nicians – includes processing scan- ners, reporting, storage, data access and research elements and comes with £50 million of investment to fully deploy digital pathology. Focus on image analysis automation The network connects the regions of West Yorkshire, Harrogate, York, Hull, and North East England, ena- bling shared knowledge and proce- dures as well as research partner- ships. Areas of focus are the devel- opment and validation of automated image analysis and AI tools for the diagnosis of cancer with existing exemplars in breast cancer, skin, and lung cancer. ‘Research partner- ships in AI diagnostics and clinical trials will be of benefit for patients and the public,’ added Dr Bansal. The scale-up of the NPIC network will also include two national spe- cialist networks; one in paediatric tumours, the other focusing on bone and soft tissue tumours. ‘There will be a national knowledge and train- ing centre, customised data sets for AI and research, and research partnerships in AI diagnostics and clinical trials,’ the expert went on. ‘Training will be a key component with hands-on workshops and train- ing for pathologists and biomedical scientists, a centre for industry/NHS collaboration, and a multiple scan- ner platform for research and AI.’ In cooperation with Genomics England, a company of the UK Department of Health and Social Care, NPIC will also create an image library to complement the genomic and clinical data for each patient and develop patholo- gy-genomic correlation. Dr Bansal stressed the importance of the net- work as the first and largest-scale project in terms of digitisation, calling the project a ‘key element of a national digital pathology vision in the UK’. She concluded: ‘It will be a platform for many applica- tions: clinical diagnosis, digital pathology and life cycle of AI to establish ourselves as the national digital pathology system platform for use in the NHS and AI research, supporting clinical trials and diag- nostic innovation.’ (MN) Daljeet Bansal Dr Daljeet Bansal is Operations Director of NPIC (National Pathology Imaging Co-operative), overseeing the scale-up of digital pathology across hospitals across the North of England and two national networks. NPIC is a unique collaboration between NHS, Academic and Industry partners, in a network that will also develop artificial intelligence tools to help diagnose cancer and other diseases. EUROPEAN HOSPITAL Vol 31 Issue 1/22