Data-Centric Evaluation of Image Segmentation Under Dataset Composition Shift
Resumo
Lung segmentation in chest X-ray images is a key step in computer-aided diagnosis, but segmentation quality often degrades under domain shift, particularly in unseen conditions. This work investigates the impact of dataset composition on U-Net-based models trained and evaluated across two public datasets. A basic U-Net and encoder-based variants (ResNet-34, ResNet-50, and VGG-16) are compared under different protocols. Results indicate that segmentation performance is highly sensitive to dataset composition and evaluation protocol, especially in pathological samples. These findings suggest that pathological variability is a major factor in degradation and highlight the importance of data-centric evaluation for reliable benchmarking in medical image analysis.
Palavras-chave:
Image segmentation, Data-centric, Dataset composition, U-Net-based models, Chest X-ray
Referências
Abueed, O. et al. (2026). Automatic semantic segmentation in chest x-ray images using deep learning approaches: a literature review. Neural Computing and Applications, 38(4):70.
Candemir, S. et al. (2013). Lung segmentation in chest radiographs using anatomical atlases with nonrigid registration. IEEE transactions on medical imaging, 33(2):577–590.
Chen, Y., , et al. (2023). Ldanet: Automatic lung parenchyma segmentation from ct images. Computers in Biology and Medicine, 155:106659.
Chowdhury, M. E. H. et al. (2020). Can ai help in screening viral and covid-19 pneumonia? IEEE Access, 8:132665–132676.
Full, P. M. et al. (2021). Studying robustness of semantic segmentation under domain shift in cardiac mri. In Puyol Anton, E., Pop, M., Sermesant, M., Campello, V., Lalande, A., Lekadir, K., Suinesiaputra, A., Camara, O., and Young, A., editors, Statistical Atlases and Computational Models of the Heart. M&Ms and EMIDEC Challenges, pages 238–249, Cham. Springer International Publishing.
Guo, B. et al. (2026). The impact of scanner domain shift on deep learning performance in medical imaging: an experimental study. International Journal of Computer Assisted Radiology and Surgery.
Jaeger, S. et al. (2013). Automatic tuberculosis screening using chest radiographs. IEEE transactions on medical imaging, 33(2):233–245.
Liu, W. et al. (2022). Automatic lung segmentation in chest x-ray images using improved u-net. Scientific Reports, 12:8649.
Peng, T. et al. (2022). H-segnet: hybrid segmentation network for lung segmentation in chest radiographs using mask region-based convolutional neural network and adaptive closed polyline searching method. Physics in Medicine Biology, 67(7):075006.
Rahman, T. et al. (2021). Exploring the effect of image enhancement techniques on covid-19 detection using chest x-ray images. Computers in Biology and Medicine, 132:104319.
Ronneberger, O. et al. (2015). U-net: Convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer-Assisted Intervention (MICCAI), volume 9351 of LNCS, pages 234–241. Springer.
Saifullah, S. et al. (2025). Automatic brain tumor segmentation: advancing u-net with resnet50 encoder for precise medical image analysis. IEEE Access, 13:43473–43489.
Stirenko, S. et al. (2018). Chest x-ray analysis of tuberculosis by deep learning with segmentation and augmentation. In 2018 IEEE 38th International Conference on Electronics and Nanotechnology (ELNANO), pages 422–428.
Wang, X. et al. (2020). Automatic segmentation of pneumothorax in chest radiographs based on a two-stage deep learning method. IEEE Transactions on Cognitive and Developmental Systems, 14(1):205–218.
Yan, W. et al. (2019). The domain shift problem of medical image segmentation and vendor-adaptation by unet-gan. In Shen, D., Liu, T., Peters, T. M., Staib, L. H., Essert, C., Zhou, S., Yap, P.-T., and Khan, A., editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2019, pages 623–631, Cham. Springer International Publishing.
Yoon, J. S. et al. (2024). Domain generalization for medical image analysis: A review. Proceedings of the IEEE, 112(10):1583–1609.
Zhang, L. et al. (2020). Generalizing deep learning for medical image segmentation to unseen domains via deep stacked transformation. IEEE Transactions on Medical Imaging, 39(7):2531–2540.
Candemir, S. et al. (2013). Lung segmentation in chest radiographs using anatomical atlases with nonrigid registration. IEEE transactions on medical imaging, 33(2):577–590.
Chen, Y., , et al. (2023). Ldanet: Automatic lung parenchyma segmentation from ct images. Computers in Biology and Medicine, 155:106659.
Chowdhury, M. E. H. et al. (2020). Can ai help in screening viral and covid-19 pneumonia? IEEE Access, 8:132665–132676.
Full, P. M. et al. (2021). Studying robustness of semantic segmentation under domain shift in cardiac mri. In Puyol Anton, E., Pop, M., Sermesant, M., Campello, V., Lalande, A., Lekadir, K., Suinesiaputra, A., Camara, O., and Young, A., editors, Statistical Atlases and Computational Models of the Heart. M&Ms and EMIDEC Challenges, pages 238–249, Cham. Springer International Publishing.
Guo, B. et al. (2026). The impact of scanner domain shift on deep learning performance in medical imaging: an experimental study. International Journal of Computer Assisted Radiology and Surgery.
Jaeger, S. et al. (2013). Automatic tuberculosis screening using chest radiographs. IEEE transactions on medical imaging, 33(2):233–245.
Liu, W. et al. (2022). Automatic lung segmentation in chest x-ray images using improved u-net. Scientific Reports, 12:8649.
Peng, T. et al. (2022). H-segnet: hybrid segmentation network for lung segmentation in chest radiographs using mask region-based convolutional neural network and adaptive closed polyline searching method. Physics in Medicine Biology, 67(7):075006.
Rahman, T. et al. (2021). Exploring the effect of image enhancement techniques on covid-19 detection using chest x-ray images. Computers in Biology and Medicine, 132:104319.
Ronneberger, O. et al. (2015). U-net: Convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer-Assisted Intervention (MICCAI), volume 9351 of LNCS, pages 234–241. Springer.
Saifullah, S. et al. (2025). Automatic brain tumor segmentation: advancing u-net with resnet50 encoder for precise medical image analysis. IEEE Access, 13:43473–43489.
Stirenko, S. et al. (2018). Chest x-ray analysis of tuberculosis by deep learning with segmentation and augmentation. In 2018 IEEE 38th International Conference on Electronics and Nanotechnology (ELNANO), pages 422–428.
Wang, X. et al. (2020). Automatic segmentation of pneumothorax in chest radiographs based on a two-stage deep learning method. IEEE Transactions on Cognitive and Developmental Systems, 14(1):205–218.
Yan, W. et al. (2019). The domain shift problem of medical image segmentation and vendor-adaptation by unet-gan. In Shen, D., Liu, T., Peters, T. M., Staib, L. H., Essert, C., Zhou, S., Yap, P.-T., and Khan, A., editors, Medical Image Computing and Computer Assisted Intervention – MICCAI 2019, pages 623–631, Cham. Springer International Publishing.
Yoon, J. S. et al. (2024). Domain generalization for medical image analysis: A review. Proceedings of the IEEE, 112(10):1583–1609.
Zhang, L. et al. (2020). Generalizing deep learning for medical image segmentation to unseen domains via deep stacked transformation. IEEE Transactions on Medical Imaging, 39(7):2531–2540.
Publicado
08/09/2026
Como Citar
FAVORETO, Graciella S.; KOENIGKAM SANTOS, Marcel; TRAINA, Agma J. M..
Data-Centric Evaluation of Image Segmentation Under Dataset Composition Shift. In: SIMPÓSIO BRASILEIRO DE BANCO DE DADOS (SBBD), 41. , 2026, São Carlos/SP.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 973-979.
ISSN 2763-8979.
DOI: https://doi.org/10.5753/sbbd.2026.249643.
