DARSA (Deep-learning Assisted Radiological Software Application):Innovative Machine Learning approaches for Detecting Pathology inImages

Many aspects of healthcare are time consuming and error prone. Recently there has been great progress in using artificial intelligence to solve a number of problems. One of the best examples of this is image labelling using a type of neural network approach called deep learning. Recent research has shown that deep learning approaches can outperform expert human radiologists when diagnosing disease in chest x-rays, in some situations. In this project we use a large set of chest x-rays as a test bed and develop a new method for software based radiological diagnosis using deep learning models. We will compare the predictive quality of our models with other machine learning approaches. The outcome of this research should be a new software-based radiological diagnosis system that is trained on chest x-rays but can also be extended to other body parts.

Faculty Supervisor:

Mark Chignell

Student:

Yilun Zhang

Partner:

GenomeMe Lab Inc

Discipline:

Engineering - mechanical

Sector:

Health care and social assistance

University:

University of Toronto