Award Date

5-15-2026

Degree Type

Thesis

Degree Name

Master of Science in Engineering (MSE)

Department

Electrical and Computer Engineering

First Committee Member

Brendan Morris

Second Committee Member

Mei Yang

Third Committee Member

Shengjie Zhai

Fourth Committee Member

Fatma Nasoz

Number of Pages

160

Abstract

Accurate identification of the tissue perfusion phase from hand images can aid doctors in decision-making with non-invasive techniques. The present study proposes a multimodal deep learning model for classifying the tissue perfusion phase using infrared, thermal, and visible spectrum images of the human hand. The proposed model consists of various preprocessing techniques such as manipulation, homography alignments, and masking. The significant contribution of this thesis is the interpretability analysis of deep learning models, achieved through the analysis of saliency maps and the Gradient-weighted Class Activation Mapping (Grad-CAM) methods. The purpose of this method is to find out how the convolutional neural networks (CNNs) are able to find discriminative regions in di!erent types of data during the classification of the perfusion phase. Saliency maps provide support for determining whether the model focuses on the physiologically relevant regions of the hand, such as patterns and temperature distribution, rather than the background, which is irrelevant to the data. This interpretability analysis provides important information about the clarity and reliability of the multimodal classification framework.Experiments were performed on unimodal, multimodal, and late fusion configurations of the proposed model, both for known subject (Pass2) and unseen subject evaluation conditions. The results show that the proposed multimodal combinations improve the classification performance over unimodal configurations, while preprocessing operations such as alignment and masking improve robustness by eliminating noise and spatial inconsistencies in the data. Moreover, the saliency maps and Grad-CAM visualization confirm that the proposed models focus on hand-centered physiological features for making predictions. In conclusion, this study proves that the proposed approach of combining multimodal imaging with deep learning can improve the performance of tissue perfusion classification systems while increasing their interpretability with the help of deep learning techniques such as saliency map and Grad-CAM visualization.

Keywords

Deep Learning; Grad-CAM; Late Fusion; Multimodal Imaging; Saliency Maps; Tissue Perfusion

Disciplines

Artificial Intelligence and Robotics | Computer Engineering | Electrical and Computer Engineering

File Format

PDF

File Size

4700 KB

Degree Grantor

University of Nevada, Las Vegas

Language

English

Rights

IN COPYRIGHT. For more information about this rights statement, please visit http://rightsstatements.org/vocab/InC/1.0/

Download


Share

COinS