Hemorrhage classification in traumatic brain injury CT scans

OData support
Dr. Jobbágy Ákos Andor
Department of Measurement and Information Systems

In this semester I studied the different haemorrhages caused by traumatic brain injuries. I have done my thesis at GE Heathcare multinational company for a half year.

This document contains the anatomical background, evolution, type and neuroradiological specificity of haemorrhages. I present a variety of diagnostic imaging procedures and I introduce the most advanced examination techniques for the traumatic brain injuries. In addition I have read a lot of scientific articles in order to detection the different hemorrhages. Results of my classification are validated by radiologist.

The purpose of my master thesis is to develop an algorithm in this semester based my acquired knowledge. This algorithm can detect automatically the different haemorrhages according to type and location.

I have chosen the DIGITS (Deep Learning GPU Training System) that can classify the different haemorrhages. I have created some pictures with their color channels emerge in the light of the various hyperdensities. After a lot of experimentation accuracy of my results that the validaton approximately is 60 percent. Accuracy of recognition the acute blood is 80 percent.

Aim of my research is to help doctors’ diagnosis because they don’t have to pay attention to the hemorrhages just the diagnosis.


Please sign in to download the files of this thesis.