Date of Award

Winter 12-10-2022

Author's School

McKelvey School of Engineering

Author's Department

Electrical & Systems Engineering

Degree Name

Master of Science (MS)

Degree Type



Traumatic events such as dislocation, breaks, and arthritis of musculoskeletal joints can cause the development of post-traumatic joint contracture (PTJC). Clinically, noninvasive techniques such as Magnetic Resonance Imaging (MRI) scans are used to analyze the disease. Such procedures require a patient to sit sedentary for long periods of time and can be expensive as well. Additionally, years of practice and experience are required for clinicians to accurately recognize the diseased anterior capsule region and make an accurate diagnosis. Manual tracing of the anterior capsule is done to help with diagnosis but is subjective and timely. As a result, there is a need to acquire tissue-level information from low-resolution MRI scans and segment the capsule in a timely, cost-efficient, and unbiased manner. Previous studies have utilized similar deep learning architecture to super-resolve, or segment areas of interest within an MRI image, while others have used traditional machine learning models to classify an image based on high-level pixel or image statistics. However, no study has utilized these methods in a complete imaging pipeline to analyze and diagnose PTJC. This research proposes a four-part fully automated imaging pipeline to (1) enhance the MRI resolution, (2) segment the capsule region, (3) unbiasedly detect unique tissue within that region, and (4) classify the image as healthy or injured. We found that the super-resolution model was able to improve the low-resolution MRI scans by an average of 6.58 PSNR but was not able to capture enough tissue detail for the segmentation model to predict an accurate mask. Despite this, the tissue clustering model was able to determine unique tissue representations within the anterior capsule as well as show the distribution of tissue is distinctly different between healthy and injured elbows. When connecting the entire pipeline together, we were able to accurately classify high-resolution MRI scans of elbows as healthy or injured with an accuracy of 72.1%, precision of 66.7%, and recall of 58.8%.


English (en)


Ulugbek S. Kamilov

Committee Members

Joseph A. O’Sullivan Tao Ju