"Advancing Precision and Autonomy in Agriculture and Medical Imaging th" by Tan-Hanh Pham

Date of Award

5-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical and Civil Engineering

First Advisor

Kim-Doang Nguyen

Second Advisor

Xianqi Li

Third Advisor

Hector M. Gutierrez

Fourth Advisor

Ryan T. White

Abstract

Deep learning has revolutionized numerous fields by enhancing precision, automation, and decision-making capabilities. This dissertation explores its applications in agriculture and medical image processing, introducing novel methodologies to improve accuracy and efficiency in these domains. These fields hold critical societal importance -- agriculture underpins global food security and sustainability, while medical imaging drives advancements in diagnostics and personalized healthcare, both benefiting significantly from data-driven innovations. In agriculture, deep learning is applied to precision spray systems through droplet analysis. Specifically, a generative model is designed to create synthetic droplet images, addressing the challenge of limited training samples, which are expensive and time-consuming to collect. Additionally, a spectral image-based soil sampling localization method is introduced to enhance resource management and crop monitoring, enabling targeted soil sampling instead of random collection. Furthermore, a deep learning pipeline for automated rice tiller angle and plant width measurement is proposed, leveraging keypoint detection models to accurately estimate tiller angles from UAV-collected imagery. In medical image processing, a Transformer-based model is developed for colon polyp detection and organ segmentation. This method fully utilizes Transformer capabilities to extract meaningful features while reducing computational complexity through spatial reduction attention. Extending this approach, a 3D version of the model is implemented for automatic brain image segmentation, demonstrating its scalability from 2D to 3D. Beyond image segmentation, this dissertation explores deep learning for reconstructing missing voxels or pixels in brain scans obtained from multiple imaging modalities. The study focuses on 3D Magnetic Resonance Spectroscopic Imaging (MRSI), a valuable technique for analyzing approximately 20 human brain metabolites. The results demonstrate that deep learning outperforms traditional methods, such as image filtering or interpolation, particularly for 3D volumetric data. The findings of this dissertation highlight deep learning’s potential to enhance precision, adaptability, and automation across multiple fields. The proposed methods contribute to advancements in medical diagnostics and sustainable agriculture, paving the way for further interdisciplinary innovations.

Share

COinS