Machine Learning Strategies for Enhancing Reconstruction and Classification in Diffuse Optical Tomography and Photoacoustic Tomography for Ovarian and Breast Cancer - Dissertation

Aug 5, 2026

Upload by Asher

Machine learning strategies are developed to improve reconstruction and classification performance in diffuse optical tomography (DOT) and photoacoustic tomography (PAT) for ovarian and breast cancer applications. The work introduces models that incorporate physical constraints for DOT, combines ultrasound-enhanced learning to enable quantitative PAT of ovarian lesions, and proposes a neural radiance field framework (PA-NeRF) to reconstruct 3D PAT from limited B-scan data. Quantitative evaluations, phantom experiments, and clinical examples assess image quality, recovered anatomical parameters, and diagnostic relevance across scenarios.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages124

Reading Time5 h 12 min

AccessFree to download

List of Figures

List of Tables

Acknowledgments

Abstract

Chapter 1: Introduction

Motivation

Optical Imaging and Deep Learning Techniques

Organization of the Dissertation

Chapter 2: Machine Learning Model with Physical Constraints for Diffuse Optical Tomography

Background

Dataset and System

Born-CGD method

ML-PC model

Quantitative Evaluation

Results

Conclusions

Chapter 3: Ultrasound-enhanced Unet model for quantitative photoacoustic tomography of ovarian lesions

Introduction

Methodology and Materials

Results and Analysis

Summary and Discussion

Chapter 4: PA-NeRF, a neural radiance field model for 3D photoacoustic tomography reconstruction from limited Bscan data

Introduction

Methodology

Results

Discussion

Chapter 5: Ultrasound and Diffused Optical Tomography-Transformer Model for Assessing Pathological Complete Response to Neoadjuvant Chemotherapy in Breast Cancer

Washington University in St. Louis Washington University Open Scholarship | McKelvey School of Engineering Theses & Dissertations | McKelvey School of Engineering | | --- | --- | | 10-22-2024
Machine Learning Strategies for Enhancing Reconstruction and Classification in Diffuse Optical Tomography and Photoacoustic Tomography for Ovarian and Breast Cancer
Yun Zou
Washington University – McKelvey School of Engineering
Follow this and additional works at: [https://openscholarship.wustl.edu/eng_etds](https://openscholarship.wustl.edu/eng_etds) | | Recommended Citation Zou, Yun, "Machine Learning Strategies for Enhancing Reconstruction and Classification in Diffuse Optical Tomography and Photoacoustic Tomography for Ovarian and Breast Cancer" (2024) . McKelvey School of Engineering Theses & Dissertations. 1106. [https://openscholarship.wustl.edu/eng_etds/1](https://openscholarship.wustl.edu/eng_etds/1)106 This Dissertation is brought to you for free and open access by the McKelvey School of Engineering at Washington University Open Scholarship. It has been accepted for inclusion in McKelvey School of Engineering Theses & Dissertations by an authorized administrator of Washington University Open Scholarship. For more information, please contact [digital@wumail.wustl.edu](digital@wumail.wustl.edu). WASHINGTON UNIVERSITY IN ST. LOUIS McKelvey School of Engineering Department of Biomedical Engineering Dissertation Examination Committee: Quing Zhu, Chair Adam Bauer Song Hu Abhinav K. Jha Neal Patwari Machine Learning Strategies for Enhancing Reconstruction and Classification in Diffuse Optical Tomography and Photoacoustic Tomography for Ovarian and Breast Cancer by Yun Zou A dissertation presented to the McKelvey School of Engineering of Washington University in partial fulfillment of the requirements for the degree of Doctor of Philosophy December 2024 St. Louis, Missouri © 2024, Yun Zou Table of Contents List of Figures ..................................... v List of Tables ...................................... x Acknowledgments ................................... xi Abstract ......................................... xii Chapter 1: Introduction ............................... 1 1.1 Motivation ..................................... 1 1.2 Optical Imaging and Deep Learning Techniques ................ 3 1.3 Organization of the Dissertation ......................... 4 Chapter 2: Machine Learning Model with Physical Constraints for Diffuse Optical Tomography ................................ 6 2.1 Background .................................... 6 2.2 Dataset and System ................................ 8 2.2.1 Simulation Data .............................. 9 2.2.2 DOT system, phantom and clinical data ................ 9 2.3 Born-CGD method ................................ 11 2.4 ML-PC model ................................... 13 2.4.1 Training .................................. 14 2.4.2 Testing and Physical Constraints for ML-PC .............. 15 2.5 Quantitative Evaluation ............................. 17 2.5.1 Reconstructed Target Diameter ..................... 17 2.5.2 Reconstructed Depth Profile ....................... 17 2.5.3 Reconstructed Maximum Absorption Coefficient ............ 18 2.6 Results ....................................... 18 2.6.1 Phantom Reconstruction Results .................... 18 2.6.2 Phantom Quantitative Results ...................... 20 2.6.3 Clinical Examples ............................. 22 2.7 Conclusions .................................... 24 Chapter 3: Ultrasound-enhanced Unet model for quantitative photoacoustic tomography of ovarian lesions .......................... 27 3.1 Abstract ...................................... 27 3.2 Introduction .................................... 28 3.3 Methodology and Materials ........................... 30 3.3.1 Ultrasound-enhanced Unet Model .

FAQs

What imaging modalities and cancer types are covered in the dissertation?

The dissertation focuses on diffuse optical tomography and photoacoustic tomography for ovarian cancer and breast cancer applications.

How does the work improve DOT reconstruction using machine learning?

It proposes machine learning models that embed physical constraints, including a Born-CGD method and an ML-PC approach, and evaluates them quantitatively.

What is the purpose of the PA-NeRF model in photoacoustic tomography?

PA-NeRF is presented as a neural radiance field model to enable 3D photoacoustic tomography reconstruction from limited B-scan data, supported by simulations, phantom results, and clinical examples.

Aug 5, 2026

Upload by Asher