Deep learning methods have been shown to achieve excellent performance on diagnostic tasks, but it is still an open challenge how to optimally combine them with expert knowledge and existing clinical decision pathways. This question is particularly important for the early detection of cancer, where high volume workflows might potentially benefit substantially from automated analysis. Here, we present a deep learning framework to analyse samples of the Cytosponge(R)-TFF3 test, a minimally invasive alternative to endoscopy, for detecting Barretts Esophagus, the main precursor of esophageal cancer. We trained and independently validated the framework on data from two clinical trials, analysing a combined total of 4,662 pathology slides from 2,331 patients. Our approach exploits screening patterns of expert gastrointestinal16pathologists and established decision pathways to define eight triage classes of varying priority for manual expert review. By substituting manual review with automated review in low-priority classes, we can reduce pathologist workload by up to 66% while matching the diagnostic performance of expert pathologists. These results lay the foundation for tailored, semi-automated decision support systems embedded in clinical workflows.
Tissue Level Based Deep Learning Framework for Early Detection of
Dysplasia in Oral Squamous Epithelium