A carregar...
Publicação
Image Reconstruction aided by Machine Learning for Breast Microwave Imaging
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 18.35 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Breast cancer remains the most prevalent malignancy among women worldwide, and early detection is essential to improve treatment outcomes and survival rates. However, current imaging modalities, such as mammography, ultrassound, and MRI, present limitations in terms of cost, comfort, availability, and diagnostic sensitivity, particularly in women with dense breast tissue. In this context, Microwave Imaging (MWI) has emerged as a promising alternative, offering non-ionizing, cost-effective, and portable solutions for breast cancer detection. Nonetheless, traditional MWI reconstruction techniques such as Delay-And-Sum (DAS) are constrained by low resolution, noise sensitivity, and limited image quality. To address these challenges, this dissertation explores the application of Machine Learning (ML) algorithms, namely K-Nearest Neighbors (KNN), Decision Trees (DT), and Support Vector Machines (SVM), to undergo image reconstruction in breast MWI. A scoping literature review was conducted, highlighting current trends, challenges, and research gaps in the application of ML to MWI. A methodology was developed, involving signal acquisition using numerical breast phantoms, Feature Engineering (FE) pipelines (with optional normalization and window trimming), and model optimization via crossvalidation. The classification performance was assessed using metrics such as sensitivity, specificity, accuracy, precision, and F1-score, while reconstruction quality was evaluated using Intersection over Union (IoU), Dice Similarity Coefficient (DSC), Relative Volume Error (RVE), and Matthews Correlation Coefficient (MCC). Results show that ML-based reconstructions outperform DAS in visual assessments, demonstrating improved tumor localization and reduced noise. Among the models tested, SVM generally achieved superior classification and reconstruction performance across multiple datasets. This work confirms the potential of integrating ML techniques in MWI for breast cancer imaging and provides a reproducible pipeline that can be further expanded with more advanced models and experimental data. Overall, this dissertation contributes to advancing reliable, accessible, and non-invasive imaging methods for breast cancer detection.
Descrição
Tese de Mestrado, Engenharia Biomédica e Biofísica, 2025, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Microwave Imaging Breast Cancer Machine Learning Feature Engineering