Explainable AI in Medical Imaging: An Interdisciplinary Translational Approach

<p>Advances in computer vision and image processing have made a clear impact on many fields, from healthcare diagnostics to autonomous driving. However, as these models become more complex, understanding their decision-making processes has grown increasingly challenging, making explainable AI (XAI) a crucial component of modern AI systems. The focus of this work is to integrate these new technologies alongside foundational methods of image processing to create tools that can be used by domain experts who are not programmers. Prior to delving into the projects which investigate these concepts, the methodologies, background, and the overall frameworks are discussed. In the first study, I developed an image processing and ML classification pipeline and web application for identifying the brain region associated with the neuropathology of schizophrenia in a disease agnostic method. In the second study, I used a masked autoencoder for pre-training, and then a vision transformer for the classification on a small dataset of alcoholic patients from healthy controls. In the third study, I used a transformer with spatiotemporal attention, where the data was updated from time-series data to a set of images, and became an image completion problem.</p>

Name thesis:degree_name

Doctor of Philosophy (PhD)

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computational and Data Sciences

Year dc:date.available

2024

Author dc:creator

• Chavez, Caitlyn

Contributors dc:contributor

• Uri Maoz, Ph.D., Chair
• Erik Linstead, Ph.D.
• Kyongsik Yun, Ph.D.

• Data Science

Repository record dc:identifier

https://digitalcommons.chapman.edu/cads_dissertations/48

OAI identifier oai:identifier

oai:digitalcommons.chapman.edu:cads_dissertations-1049