Tagging and Auto-Captioning of Histopathology Scans for Diagnosis Assistance

Digital pathology uses modern scanners to capture high-resolution images from biopsy samples. Computer algorithms, especially artificial intelligence, can help in automatic searching for similar cases in the archive of hospitals and laboratories. Displaying similar images form the past patients, that have already been diagnosed and treated, can provide useful information to the pathologist to solidify the final diagnosis. These images, however, are very large such that their processing requires smart algorithm to distinguish between relevant and irrelevant information. In this project, we develop novel methods for identification and auto-captioning of the pathology scans to assist clinicians in their daily work. Tagging technologies using AI will be developed to index and smartly archive the images. This enables fast and accurate search for similar cases. Besides, the identification of already diagnosed cases (which are accompanied by pathology report) enables the algorithm to generate a caption for the new scans. This should help shorten the processing times by drawing attention to the critical cases.

Faculty Supervisor:

Hamid Tizhoosh

Student:

Shivam Kalra

Partner:

Huon Digital Pathology

Discipline:

Engineering

Sector:

Medical devices

University:

Program: