Exploring Bias in Pre-Trained CNNs: Fairness Assessment in Sjögren’s Disease Detection

• Bruna Ferreira dos Santos UNIFESP
• Lilian Berton UNIFESP

DOI: https://doi.org/10.5753/eniac.2025.13767

Resumo

Sjögren’s disease is an autoimmune disorder that primarily affects moisture-producing glands, leading to symptoms such as dry eyes, dry mouth, and systemic complications. Accurate diagnosis remains a challenge, particularly in medical imaging applications. In this study, we employ various Convolutional Neural Networks (CNNs) for Sjögren’s disease detection, using a dataset with inherent biases. The goal is to analyze the impact of transfer learning on fairness, investigating whether pre-trained models amplify or mitigate biases in medical image classification. We compare different CNN architectures and evaluate fairness metrics to assess performance disparities across demographic subgroups. The findings contribute to the development of bias-aware AI models, ensuring more equitable and reliable deep learning applications in health.

Referências

Álvarez Troncoso, J., Ruiz-Bravo, E., Soto Abánades, C., Dumusc, A., López-Janeiro, Á., and Hügle, T. (2024). Classification of salivary gland biopsies in sjögren’s syndrome by a convolutional neural network using an auto-machine learning platform. BMC rheumatology, 8(1):60.

Guan, H. and Liu, M. (2021). Domain adaptation for medical image analysis: a survey. IEEE Transactions on Biomedical Engineering, 69(3):1173–1185.

Mei, X., Liu, Z., Robson, P. M., Marinelli, B., Huang, M., Doshi, A., Jacobi, A., Cao, C., Link, K. E., Yang, T., Wang, Y., Greenspan, H., Deyer, T., Fayad, Z. A., and Yang, Y. Radimagenet: An open radiologic deep learning research dataset for effective transfer learning. Radiology: Artificial Intelligence, 0(ja):e210315.

Nocturne, G. and Mariette, X. (2013). Advances in understanding the pathogenesis of primary sjögren’s syndrome. Nature Reviews Rheumatology, 9(9):544–556.

Olivier, A., Hoffmann, C., Jousse-Joulin, S., Mansour, A., Bressollette, L., and Clement, B. (2023). Machine and deep learning approaches applied to classify gougerot–sjögren syndrome and jointly segment salivary glands. Bioengineering, 10(11):1283.

Rabonato, R. T. and Berton, L. (2024). A systematic review of fairness in machine learning. AI and Ethics, pages 1–12.

Schumann, C., Wang, X., Beutel, A., Chen, J., Qian, H., and Chi, E. H. (2019). Transfer of machine learning fairness across domains. arXiv preprint arXiv:1906.09688.

Teo, C. T., Abdollahzadeh, M., and Cheung, N.-M. (2023). Fair generative models via transfer learning. In Proceedings of the AAAI conference on artificial intelligence, volume 37, pages 2429–2437.

Vyas, B., Khatiashvili, A., Galati, L., Ngo, K., Gildener-Leapman, N., Larsen, M., and Lednev, I. K. (2024). Raman hyperspectroscopy of saliva and machine learning for sjögren’s disease diagnostics. Scientific Reports, 14(1):11135.

Wu, R., Chen, Z., Yu, J., Lai, P., Chen, X., Han, A., Xu, M., Fan, Z., Cheng, B., Jiang, Y., et al. (2024). A graph-learning based model for automatic diagnosis of sjögren’s syndrome on digital pathological images: a multicentre cohort study. Journal of Translational Medicine, 22(1):748.

Zabotti, A., Callegher, S., Tullio, A., Vukicevic, A., Hocevar, A., Milic, V., Cafaro, G., Carotti, M., Delli, K., De Lucia, O., Ernst, D., Ferro, F., Gattamelata, A., Germanò, G., Giovannini, I., Hammenfors, D., Jonsson, M., Jousse-Joulin, S., Macchioni, P., and Vita, S. (2020). Salivary gland ultrasonography in sjögren’s syndrome: A european multicenter reliability exercise for the harmonicss project. Frontiers in Medicine, 7:581248.