Concept Bottleneck Models to detect hidden features or avoid memorization

Concept Bottleneck Models [1] are designed to leverage high-level concepts. They revisit the classic idea of first predicting concepts that are providing at training time, and then using these concepts to predict the label. By construction, it is possible to intervene on these concept bottleneck models by editing their predicted concept values and propagating these changes to the final prediction.

The goal of this project is to leverage the concept block to detect possible spurious correlations (shortcuts) [2,3], memorization [4] or unfair predictions [5]. For this, we plan to leverage the meta-data available (chest drains, age, sex, race, etc.)

Datasets that could be used: * Chest X-rays: CheXpert, NIH-CXR14, PadChest, etc. * Skin lesions: ISIC, PAD-UFES-20, etc. * Other datasets can be found at https://grand-challenge.org/challenges/

Github repository: https://github.com/yewsiang/ConceptBottleneck

Projects related to the use of meta-data, meta-learning and transfer learning are possible. More precise research questions will be decided together with the student. It requires some background knowledge and experience in deep learning. This project is co-supervised by Veronika Cheplygina.

References

1. Kim, C., Gadgil, S. U., DeGrave, A. J., Omiye, J. A., Cai, Z. R., Daneshjou, R., & Lee, S. I. (2024). Transparent medical image AI via an image–text foundation model grounded in medical literature. Nature Medicine, 1-12.
2. Koh, P. W., Nguyen, T., Tang, Y. S., Mussmann, S., Pierson, E., Kim, B., & Liang, P. (2020, November). Concept bottleneck models. In International Conference on Machine Learning (pp. 5338-5348). PMLR.
3. Oakden-Rayner, L., Dunnmon, J., Carneiro, G., & Ré, C. (2020, April). Hidden stratification causes clinically meaningful failures in machine learning for medical imaging. In Proceedings of the ACM conference on health, inference, and learning (pp. 151-159).
4. Jiménez-Sánchez, A., Juodelye, D., Chamberlain, B., & Cheplygina, V. (2022). Detecting Shortcuts in Medical Images-A Case Study in Chest X-rays. arXiv preprint arXiv:2211.04279.
5. Gichoya, J. W., Banerjee, I., Bhimireddy, A. R., Burns, J. L., Celi, L. A., Chen, L. C., … & Zhang, H. (2022). AI recognition of patient race in medical imaging: a modelling study. The Lancet Digital Health, 4(6), e406-e414.
6. Larrazabal, A. J., Nieto, N., Peterson, V., Milone, D. H., & Ferrante, E. (2020). Gender imbalance in medical imaging datasets produces biased classifiers for computer-aided diagnosis. Proceedings of the National Academy of Sciences, 117(23), 12592-12594.