Banca de QUALIFICAÇÃO: PAULO SERGIO DA SILVA JUNIOR

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : PAULO SERGIO DA SILVA JUNIOR
DATE: 21/07/2025
TIME: 09:30
LOCAL: Online
TITLE:

4D Cardiac Segmentation Using Neural Networks

KEY WORDS:

Cardiac Segmentation, UNet 3D, Transformer, Medical Imaging, Deep Learning

PAGES: 77
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
SPECIALTY: Processamento Gráfico (Graphics)
SUMMARY:

Cardiovascular diseases are the leading cause of death worldwide, accounting for approximately 17.9 million deaths annually, according to the World Health Organization (WHO). This scenario highlights the urgent need for advanced methods to enable accurate diagnosis and monitoring. Automatic cardiac image segmentation plays a crucial role in this context by allowing the extraction of essential clinical parameters with greater speed and consistency.

This work investigated deep learning-based approaches for cardiac segmentation in magnetic resonance imaging (MRI). The initial proposal involved the implementation of a 3D UNet architecture, widely adopted in volumetric medical image segmentation. However, persistent overfitting was observed during training, even after applying regularization techniques, data augmentation, and hyperparameter tuning. This limitation hindered the model’s generalization capacity, especially in segmenting complex structures such as the myocardium and right ventricle.

As a result, a methodological shift was made towards Transformer-based architectures, motivated by their greater ability to capture long-range spatial dependencies and recent advances in medical image segmentation. The final algorithm, developed in Python, enables 4D analysis of the cardiac cycle and was evaluated on a realistic dataset using metrics such as Dice and IoU. The results demonstrate that the new approach achieved significant improvements in accuracy and robustness.

COMMITTEE MEMBERS:
Presidente - 2177445 - BRUNO MOTTA DE CARVALHO
Interna - 1350250 - ANNE MAGALY DE PAULA CANUTO
Externo ao Programa - 1153006 - LUIZ AFFONSO HENDERSON GUEDES DE OLIVEIRA - UFRN