Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.

Non-WSU users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Access Type

WSU Access

Date of Award

January 2023

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Industrial and Manufacturing Engineering

First Advisor

Ratna B. Chinnam

Second Advisor

Suzan Arslanturk

Abstract

Detecting brain abnormalities in neonates is crucial for early diagnosis and treatment of various neurological disorders. Magnetic Resonance Imaging (MRI) has been shown to be an effective imaging modality for detecting brain abnormalities in neonates. Furthermore, machine learning and deep learning algorithms that utilize MRI's for clinical tasks are increasingly popular as their architectures improve and available computational increases.

In this dissertation, we propose multiple novel unsupervised approaches for detecting brain abnormalities in neonates using multimodal and segmented MRI brain scans. We also define a framework against which the quality of our models can be assessed, and the results of which can be utilized by a neuroradiologists in a clinical setting.

Our approach consists of variational autoencoder (VAE) networks for learning low-dimensional representations of normal brain scans and postprocessing to aid in detecting abnormal brain regions based on the learned representation of normality. We evaluate our approaches on a dataset provided by the Developing Human Connectome Project (dHCP), consisting of T1-weighted and T2-weighted, as well as segmentation masks encapsulating various regions of the brain.

Our experimental results show that our approach is capable of detecting anomalies regions that do not conform to normal neonatal brain physiology, with each approach capable of highlighting different aspects of abnormality. Furthermore, we demonstrate the interpretability of our approach by visualizing the abnormal brain regions detected by our methods. Our approach has the significant potential to be used as a radiologist's aide for detecting brain abnormalities in neonates, which can lead to earlier diagnosis and better treatment outcomes.

Off-campus Download

Share

COinS