ARTIFICIAL INTELLIGENCE AND PATHOLOGY

The microscopic diagnosis of cancer remains challenging. Recent data from our group within the French (nationwide) LymphoPath network shows that 20% diagnoses are inaccurate, with direct impact on patient care. Indeed, molecular techniques, not affordable for all pathology departments, have become critical final diagnosis. Currently, automated solutions could help pathologists diagnostic decisions or histological grading lacking, and it is reasonable to expect be experts rare tumours if they only see a few cases per year. Digital microscopy offers unique features which available in conventional optical microscopy. Assisted by dedicated software tools, digital enables dynamic prompt access any detail stained slides at several magnifications. Furthermore, calibrated qualities number discrete pixels slide allow image analysis quantification using computer vision and, particular, deep learning approaches. Image feature extraction methods based pixel detection less often object segmentation has brought much hope improving accuracy human eye. Automatic whole-slide images recently been performed predicting tumour classification, gene mutations survival outcomes. However, positioning such technologies routine practice limited trained networks frequently do meet industrial constraints required general application as certification, qualification explainability (black box effect) algorithms. This presentation will describe different approaches machine aiming classifying subtypes through histopathologic features. Advantages drawbacks these techniques discussed special emphasis risk biased performance assessment systems. role sets quality strategies envisaged broaden use neural networks. Providing limitations taken into account circumvented, seems artificial intelligence histopathology should pave road precision medicine integration holistic dashboard oncology care teams. Keywords: Bioinformatics; Computational Systems Biology, Genomics, Epigenomics, Other -Omics, Tumor Biology Heterogeneity Conflicts interests pertinent abstract P. Brousset Consultant advisory role: Roche, MSB, Janssen Cilag laboratory Research funding: Pierre Fabre