Application for detecting breast cancer in mammography images using neural network technology(Aplicação para detectar câncer de mama em imagens de mamografias com a técnica de redes neurais)

Resumo

https://doi.org/10.5281/zenodo.20600286

ABSTRACT

Given the alarming and growing rate of breast cancer cases, a computational model was developed to assist healthcare professionals and optimize the tumor diagnosis process. When detected early, breast cancer has a higher likelihood of effective treatment and better prognoses. However, the manual analysis of medical exams is time-consuming and prone to human error.

According to the Brazilian National Cancer Institute (INCA), breast cancer is the most prevalent type among women in Brazil, with an estimated 73,610 new cases per year for the 2023–2025 period (INCA, 2024). International studies indicate that the use of artificial intelligence techniques in mammography interpretation can increase early tumor detection by up to 47%, while also reducing diagnostic errors and medical analysis time (Nature Medicine, 2023).

This study aims to evaluate machine learning and computer vision models applied to breast cancer detection in mammograms. Four distinct models were implemented and compared: YOLO (You Only Look Once), designed for automatic detection of suspicious regions in images; Random Forest, used for classifying patterns extracted from the images; CNN (Convolutional Neural Network), applied for analyzing complex visual features; and ResNet (Residual Neural Network), employed for hierarchical feature extraction and accuracy enhancement.

Among the evaluated models, ResNet achieved the best performance in terms of accuracy and generalization capability, demonstrating to be the most promising approach for automated support in early breast cancer diagnosis.

Thus, this study seeks not only to accelerate the detection of breast abnormalities but also to provide intelligent support for clinical decision-making, contributing to more precise diagnoses, more effective treatments, and the preservation of lives through the integration of technology and medicine.