Use of neural networks in tissue abnormality detection
Principal Investigator
Name
Rafal Grzeszczuk
Degrees
None
Institution
AGH University of Science and Technology
Position Title
student
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-299
Initial CDAS Request Approval
Apr 11, 2017
Title
Use of neural networks in tissue abnormality detection
Summary
This is a thesis project ("inżynier" degree, BSc equivalent). The main intention of this project is to assess performance of various neural network architectures given the task of classifying lung CT scans into two categories: Patients who are likely to have or already have cancer, and patients who are unlikely to develop cancer in the following year - supposedly based on specific traits of subjects' lung images, such as nodule properties. It is assumed that convolutional neural networks will be most suitable for this task. A subset of the data obtained will be used to train the network and choose the most optimal architecture. Its size will be limited by available computational power.
Aims
The main aim is to develop an optimal neural network that will allow to answer the question stated in project summary with reasonable accuracy. No specific accuracy threshold is defined but generally it should be noted that false positives are less dangerous than false negatives, so it is essential to focus on eliminating the latter.
Collaborators
AGH University of Science and Technology