Liver Tumor Segmentation in CT Scans Using Deep Learning: A U-Net Approach with Transfer Learning
Resumo
Este trabalho apresenta um pipeline para segmentação 2D de fígado e tumores a partir de tomografia computadorizada (NIfTI). Utilizamos o dataset público Liver Tumor Segmentation com 131 volumes, do qual foram extraídas fatias 2D e convertidas em imagens JPEG (com compressão mínima, sem perda perceptível) para treinamento da U-Net com backbone ResNet50 (transfer learning) implementando a biblioteca FastAI. O pré-processamento inclui leitura dos volumes com NiBabel, aplicação de janelas DICOM específicas para fígado, normalização e redimensionamento para 128×128. Como função de perda foi adotada CrossEntropy (multiclasse) e métricas customizadas de acurácia sobre os pixels de interesse (foreground).
Referências
Dice, L. R. (1945). Measures of the amount of ecologic association between species. Ecology, 26(3):297–302.
He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778.
Howard, J. and Gugger, S. (2020). Deep Learning for Coders with FastAI and PyTorch: AI Applications Without a PhD. O’Reilly Media.
Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J., and Maier-Hein, K. H. (2021). nnu-net: Self-adapting framework for u-net-based medical image segmentation. Nature Methods, 18:203–211.
Milletari, F., Navab, N., and Ahmadi, S.-A. (2016). V-net: Fully convolutional neural networks for volumetric medical image segmentation. 2016 Fourth International Conference on 3D Vision (3DV), pages 565–571.
Mvd, A. (2017). Liver tumor segmentation challenge (lits). [link].
Ronneberger, O., Fischer, P., and Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention (MICCAI), 9351:234–241.
He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778.
Howard, J. and Gugger, S. (2020). Deep Learning for Coders with FastAI and PyTorch: AI Applications Without a PhD. O’Reilly Media.
Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J., and Maier-Hein, K. H. (2021). nnu-net: Self-adapting framework for u-net-based medical image segmentation. Nature Methods, 18:203–211.
Milletari, F., Navab, N., and Ahmadi, S.-A. (2016). V-net: Fully convolutional neural networks for volumetric medical image segmentation. 2016 Fourth International Conference on 3D Vision (3DV), pages 565–571.
Mvd, A. (2017). Liver tumor segmentation challenge (lits). [link].
Ronneberger, O., Fischer, P., and Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention (MICCAI), 9351:234–241.
Publicado
12/11/2025
Como Citar
ALVES, Marcos L.; SALEM, Murilo C.; BARRETOS, Daniel H. S. P.; FERRUGEM, Anderson P..
Liver Tumor Segmentation in CT Scans Using Deep Learning: A U-Net Approach with Transfer Learning. In: ESCOLA REGIONAL DE APRENDIZADO DE MÁQUINA E INTELIGÊNCIA ARTIFICIAL DA REGIÃO SUL (ERAMIA-RS), 1. , 2025, Porto Alegre/RS.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2025
.
p. 432-435.
DOI: https://doi.org/10.5753/eramiars.2025.16788.
