Name: SABRINA SIQUEIRA PANCERI
Publication date: 27/01/2025
Examining board:
Name |
Role |
|---|---|
| ALBERTO FERREIRA DE SOUZA | Presidente |
| CLAUDINE SANTOS BADUE | Examinador Interno |
| FRANCISCO DE ASSIS BOLDT | Examinador Externo |
| JOSE GERALDO MILL | Examinador Interno |
| MARIELLA BERGER ANDRADE | Examinador Externo |
Summary: Breast cancer is the second most common type of cancer among women worldwide, with more than 2.3 million new cases registered, accounting for 11.6% of all global cancer diagnoses in 2022. Furthermore, it remains one of the leading causes of mortality among women, making early detection essential to improve prognosis and survival rates. In recent years, deep learning techniques have shown great potential in automating and enhancing breast cancer diagnosis through mammography analysis. This thesis explores the application of deep convolutional neural networks (CNNs) in the analysis of mammographic images, focusing on improving accuracy, sensitivity, and specificity in the detection of radiological findings. To this end, we developed a digital mammography dataset named HUCAMammo, which was used to train and validate the CNN models. This dataset contains the necessary features for the models to learn complex patterns associated with radiological findings. Our approach employs transfer learning and data augmentation techniques to address the challenges posed by imbalanced datasets and the inherent variability in breast tissue density. The models were evaluated using various performance metrics, such as accuracy, sensitivity, and precision, demonstrating competitive results when compared to traditional radiologist assessments. The integration of these AI-based methodologies has the potential to reduce radiologists' workload, enhance diagnostic consistency, and ultimately contribute to more effective breast cancer screening programs. Future work will focus on optimizing these models for clinical deployment and expanding their applicability to other imaging modalities.
