D I G I TA L PAT H O LO G Y 1 5 and colon cancer. Mitoses counting in breast can- cer usually takes pathologists 10-15 minutes. An AI algorithm could be used as a pre-screener that finds candidate mitoses to expedite the process. ‘We could click on the ones we like and don’t like and end up with a number of mitoses for a dif- ferent area, of typically two square millimetre.’ In nuclear segmentation in breast cancer, pathologists could segment the nuclei with image analysis algo- rithms without having to estimate in a subjective way. AI algorithms are also particu- larly helpful in immune infiltrate quantification, an important prog- nostic feature that proves more dif- ficult to assess in colon cancer than in breast. ‘You have more tissues, muscle, fat, etc.,’ he said, ‘and you only need to count a number of infiltrating inflammatory cells in a tumour area, so all areas must be recognised,’ Finding metastases in breast can- cer is a tedious task in which pathol- ogists typically have to put under and look at 50 slides, a process that can take up to 20 minutes. The missed rate is high. Van Diest and colleagues, who recently revised the original material from a big breast study, found that pathologists missed 23% of central node patholo- gies. ‘We need something to do a better job. AI did get better results than very experienced pathologists at detecting lymph node metastasis in breast cancer,’ he pointed out. The problem remains that, so far, very few AI algorithms have been implemented in clinical cases. ‘We are not there yet.’ Van Diest con- cluded. ‘Workflow needs to be com- pletely integrated and we need fully integrated image analysis, but we need to do this step by step.’ of staining, quality assurance of per- formance, implementation controls, and accreditation. A key advance would be for the elements to come together to pro- vide algorithms that are truly inter- operable and perform on different platforms, he said. ‘Digital pathol- ogy is progressing without these algorithms at the moment, but adop- tion will be speeded up when really useful algorithms come along.’ (MN) TMAs could be a vital screening tool for radiology Infusing tissue micro- arrays with new life Report: Mark Nicholls The advent of digital pathology is help- ing to address some of the chal- lenges surrounding tissue microarrays as they are integrated into the digital workflow, in some ways giving them ‘a new lease of life’, according to Professor Inti Zlobec, who spoke at the Digital Pathology and AI Congress in London last December. As Head of the Translational Research Unit at the Institute of Pathology, University of Bern, she pointed out the challenges surrounding next-generation Tissue Microarray (ngTMA) and explored the benefits of coupling with digital pathology in translational research. A long work history with tissue microarrays indicates that digital pathology is now ‘vital to the whole process’ of how TMAs are constructed. Reflecting on the early days of TMAs from 1998 and the combination of different spots on one block, Zlobec said the idea uses less resources and material and, by adding associated pathological data, a research tool can be created that can last 5-10 years. It was clear for some time that TMAs might be a valuable screening tool, and her team, which has a focus on colorectal cancer, was interested in tumour budding but the challenge was how to get budding on a TMA. ‘One solution was to incorporate digital pathology into the microarray workflow,’ she said. ‘The digital slide could be used to annotate exactly the region or multiple regions you wanted. By increasing the resolution, you could really see what you were putting on the TMAs.’ The next-generation Tissue Microarray approach, which utilises slide scanning and digital annotation Snapshot of an H&E stained next- generation Tissue Microarray as a basis for TMA construction, has been used in her lab since 2012. ‘The beauty of that is that you have fully annotated – and documented – spots and images. There was also the ability for every single spot to have clini- cal and pathological information and the clinical outcome for every single patient.’ Her Bern group created 720 TMA blocks with about 150,000 punches thumped physically into the blocks. While colorectal cancer was the main focus – about 30% of the blocks – the team also created tissue blocks for lung, pancreas, oesophagus, pros- tate and other areas. Ultimately the process led to more than one million images in the TMA archive from different tumour entities (after TMA cutting and staining) and all with the associated pathological annotation, histological information and clinical outcome… and poten- tially available for image analysis. ‘Now the problem is how to manage that number of TMA slides,’ she said. ‘A solution is with machine learning tools and having the TMA spots linked back to patients, whole slide images, annotative infor- mation and image analysis results, though this is still work in progress.’ Another question Zlobec posed was whether, with the ever-advancing technology, TMAs are still relevant and, she noted, the num- ber of annual publications on the sub- ject had plateaued in recent years at about 600. The original aim of TMAs of screening and resource efficiency, she said, was still relevant and, they can be used to study heterogene- ity, biomarkers, and to capture the tumour environment. More recently, the Bern group has developed a tool (Scorenado) that provides an inter- mediate solution to image analysis, Professor Inti Zlobec heads the Translational Research Unit at the Institute of Pathology, University of Bern. Her interests lie in inter- disciplinary translational research to improve the clinical management of patients with colorectal cancer and in histomorphological biomarkers as prognostic and predictive features of tumour response to therapies. Single spot from a next-generation tissue microarray of colorectal cancer stained with pan-cytokeratin facilitating user-friendly visual assess- ment on digital slides. Today, TMAs are a resource in the machine learning era for classification or prognosis and prediction for patient outcomes and also for translation into clinics. ‘They also combine technology with molecular applications – we can take the punch we put on a TMA and then put it into a tube to do the sequencing, giving us the kind of documentation we really need,’ Zlobec explained, also pointing out that digi- tal pathology also can complement mass spectrometry, identifying and validating new biomarkers. As digital pathology evolves, its added value to TMA use becomes clear. It improves, she pointed out, the quality of bio- marker studies with precise ngTMAs that quickly capture ROIs; uses digital image analysis; improves documenta- tion; aids biobanking and the accredi- tation process; it also offers high quality pathology standard annotation of images, and can be combined with molecular applications. Additionally, using the image as a biomarker adds a huge wealth of morphology infor- mation. Pathologists will hold a pivotal role Continued from page 13 In 1997, NHS consultant pathologist Professor David Snead works at the University Hospital of Coventry and Warwickshire, where he is clinical lead for Cellular Pathology at Coventry and Warwickshire Pathology Services. Having led the implementation of digital pathology at the centre, his main research interests lie in extending the capability of digital pathology for routine diagnostic use and the advancement and deployment of computer assisted diagnostic algorithms. the haematology survey in progress. Data from the clinical biochemistry survey is currently being analysed with results imminent, but findings from the histopathology report have been described by Martin as ‘grue- some’. ‘The survey showed that 97% of laboratories do not have enough histopathologists, and it’s going to get worse. 2021, we are expecting somewhere around 25-28% gap in the histopathology workforce. ‘RCPath is making the case for more pathologists, with support from a range of charities and other organisa- tions, in its efforts to get investment in the laboratory and diagnostic work- force. This data means we can show exactly what the issue is.’ Pressure has grown in recent years because of increased workload, not enough trainees, with a lower pro- file in medical schools and limited opportunities to meet pathologists. Additionally, an ageing workforce exists with a quarter of histopatholo- gists aged 55 years or over, as well as the risks of stressed practitioners opt- ing for early retirement. ‘We need to de-stress the system to take the heat off current people so that they are prepared to stay,’ Martin emphasized. Digital pathology and AI support will help, she believes. With high levels of pathology expertise in the NHS, she remains optimistic about the future for pathology with the various disciplines poised to play a pivotal within future diagnosis and care delivery. (MN) Lluís Donoso-Bach gained his medical degree from the School of Medicine of the Autonomous University of Barcelona in 1981. He completed his residency in radiology at the Hospital de Sant Pau in Barcelona, in 1992, and in that year was appointed Chairman of the Radiology Department at the UDIAT Diagnostic Centre. In 1998 he became its Executive Director. Since 2006 he has led the Diagnostic Imaging Department at the Hospital Clínic of Barcelona and been Professor of Radiology at the University of Barcelona. gration of image analysis and deep learning. During diagnostic work, we often use completely subjective features to make a diagnosis. For example, we diagnose cancer when we see, within the tissue, big nuclei, variation of shape or size between the nuclei, or when the chromatic pattern of nuclei are irregular, etc. But these are all subjective, and we’re easily tricked in our visual system in assessing size, so we’re going to measure it,’ he explained. Ideally pathologists could use AI and image analysis in several appli- cations, e.g. mitoses, nuclear area, degree of tubule formation and metastases finder, ‘a tedious task, where we might miss things.’ Integration would also make sense in immune infiltrate quantifi- cation; IHC membrane scoring and positive nuclei scoring; and tumour vs. stroma quantification, which is becoming quite popular in breast el analysis efits in tumour cell morphometrics to predict survival in lung adeno- carcinoma. Similar work is being conducted for prostate cancer and breast carcinoma, he said. In several tests/challenges Snead acknowledged how the computer is at least as good as the pathologist and sometimes probably better, e.g. to predict which lung cancers har- bour mutations in the EGFR gene. The UK government is funding five centres to advance AI in image diagnostics. Digital pathology is central to three of these centres including the Coventry PathLAKE centre, of which Snead is director. This initiative is designed to con- nect the UK’s excellence in AI, its strength in real world NHS clinical data and the innovative SME start biotech sector, to encourage innova- tive ideas and solutions. However, he warned, ‘Such solutions will only get into the clinic if we can validate these tools and get them approved through our regulatory framework.’ He remains confident that this can be achieved by working with regula- tors and helping to align processes with the FDA, and across the EU. Digital image analytics is mov- ing into practice, he said, but there are many challenges that remain including interoperability across platforms, variability of sections and www.healthcare-in-europe.com