Abstract:
Lymphoma is a type of cancer that is particularly difficult to diagnose as it exists in a large variety of subtypes and has numerous mimics. In France, an estimated 20% of lymphoma cases are misdiagnosed by non-expert pathologists, and getting a second opinion both increases the workload of expert hematopathologists and delays patient treatment by several weeks.
By reproducing the expert hematopathologists’ approach of sequentially detecting specific patterns in H&E-stained lymphoma slides, our goal is to create a tool to assist non-expert pathologists for lymphoma diagnosis.
To this end, we compare both Deep Learning and conventional Machine Learning approaches, both in terms of performance and of interpretability of their predictions. We also focus on feature engineering approaches to develop features reflecting the presence of lymphoma subtypes characteristics to narrow down possible diagnoses.
En appuyant sur le bouton "j'accepte" vous nous autorisez à déposer des cookies afin de mesurer l'audience de notre site. Ces données sont à notre seul usage et ne sont pas communiquées. Consultez notre politique relative aux cookies