AANS Online Scientific Sessions: Tumor-Automated Intraoperative Frozen Diagnosis of Brain Tumors Using Machine Learning

Automated Intraoperative Frozen Diagnosis of Brain Tumors Using Machine Learning

Video Transcription

Video Summary

In this video presentation, Siri Khalsa, a fifth year neurosurgery resident at the University of Michigan, discusses the use of machine learning to automatically diagnose brain tumor tissue during surgery. The problem highlighted is the limited access to neuropathologists for immediate consultation during brain tumor surgeries. The objective is to develop a deep machine learning algorithm that can analyze standard intraoperative frozen sections to provide a whole slide diagnosis. The speaker describes the process of training a convolutional neural network to classify individual small fields of view, which are then summated into a single image. The algorithm was tested on 40 whole slide images and achieved an overall accuracy of 92.5%. Prospective multicenter validation is currently underway. The research was supported by an NIH T32 grant.

Asset Subtitle

Siri Sahib Singh Khalsa, MD

Keywords

machine learning

brain tumor

neurosurgery

diagnosis

convolutional neural network