Abstract: Stroke is one of the leading causes of death worldwide. The standard procedure for diagnosis of stroke is by visual analysis of scans, such as CT and MRI scans. MRI scans are much better for visualization and accurate diagnosis of ischemic stroke; however, in practice, CT scans are more often used because of their flexibility and speed. Unfortunately, they are also much less reliable for diagnosing ischemic stroke because their relatively poor resolution makes the affected tissue much harder to visualize. The objective of this project is a neural network algorithm capable of determining the presence of ischemic stroke via a CT scan. A machine learning algorithm would standardize the identification of ischemic strokes without having to rely solely upon the subjectivity of the human eye, thus making CT scans a more reliable diagnostic tool. The results of this project show that designing such an algorithm is indeed possible, as this neural network algorithm was able to successfully make predictions on CT scans with ischemic stroke, thus, essentially doing the job of a radiologist.

Contributors:

Zoe Cohen
cohen.zoe@wustl.edu
BS in Electrical Engineering

 

Kara Dimicco
karadimicco@wustl.edu
BS in Electrical Engineering
BS in Systems Science and Engineering
Dr. Shantanu Chakrabartty

shantanu@seas.wustl.edu

Project Advisor

Professor in the Department of Electrical and Systems Engineering

Oindrila Chatterjee
oindrila.chatterjee@wustl.edu
PhD Student in Electrical Engineering