Defense Date

2022

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Electrical & Computer Engineering

First Advisor

Dr. Yanxiao Zhao

Second Advisor

Dr. Murat Kuzlu

Third Advisor

Dr. Ruixin Niu

Fourth Advisor

Dr. Sherif Abdelwahed

Fifth Advisor

Dr. Changqing Luo

Abstract

Artificial Intelligence (AI) has witnessed increased application and widespread adoption over the past decade. AI applications to medical images have the potential to assist caregivers in deciding on a proper chronic wound treatment plan by helping them to understand wound and tissue classification and border segmentation, as well as visual image synthesis.

This dissertation explores chronic wound management using AI methods, such as Generative Adversarial Networks (GAN) and Explainable AI (XAI) techniques. The wound images are collected, grouped, and processed. One primary objective of this research is to develop a series of AI models, not only to present the potential of AI in wound management but also to develop the building blocks of human digital twins.

First of all, motivations, contributions, and the dissertation outline are summarized to introduce the aim and scope of the dissertation. The first contribution of this study is to build a chronic wound classification and its explanation utilizing XAI. This model also benefits from a transfer learning methodology to improve performance. Then a novel model is developed that achieves wound border segmentation and tissue classification tasks simultaneously. A Deep Learning (DL) architecture, i.e., the GAN, is proposed to realize these tasks. Another novel model is developed for creating lifelike wounds. The output of the previously proposed model is used as an input for this model, which generates new chronic wound images. Any tissue distribution could be converted to lifelike wounds, preserving the shape of the original wound.

The aforementioned research is extended to build a digital twin for chronic wound management. Chronic wounds, enabling technologies for wound care digital twins, are examined, and a general framework for chronic wound management using the digital twin concept is investigated. The last contribution of this dissertation includes a chronic wound healing prediction model using DL techniques. It utilizes the previously developed AI models to build a chronic wound management framework using the digital twin concept. Lastly, the overall conclusions are drawn. Future challenges and further developments in chronic wound management are discussed by utilizing emerging technologies.

Rights

© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission

8-12-2022

Share

COinS