Deep federated learning in healthcare
Professor of Computational Medical Imaging Research at Universitätsklinikum Bonn
- Principal Investigator at Munich School for Data Science (MuDS)
- Affiliated Researcher at Chair for AI in Medicine (Prof. Rueckert), TU Munich
Learning to distill and share knowledge among AI agents
Federated learning (FL) has been recently introduced to enable training deep learning (DL) models or AI agents without sharing the data. In other words, AI agents at local hubs, e.g. hospitals, are trained on their own data and only share the trained parameters with a centralized AI model or other AI agents. Leveraging such a massive amount of data in a privacy-preserved fashion adhering to the General Data Protection Regulation (GDPR) would have a great impact on medical diagnosis, outbreak detection, and other healthcare services.
Yet, principal challenges, to overcome, concern the nature of medical data, namely data heterogeneity; severe class-imbalance, few amounts of annotated data, inter-/intra-scanners variability (domain shift), inter-/intra- observer variability (noisy annotations); system heterogeneity, and explainability and robustness.
The mission of this Helmholtz AI young investigator group is to develop novel algorithms for a groundbreaking new generation of deep federated learning, which can learn to reCognize, AdapT, lEarn, Reason and exPlain, dIstiLl the knowledge and coLlAboRate with other AI agents (CATERPILLAR) in a robust and privacy-preserved fashion, to provide personalized healthcare services.
- Medical Imaging with Deep Learning: We will continue our research directions to develop fully-automated, high accurate solutions that save export labor and efforts, and mitigate the challenges in medical imaging, i.e. i) the availability of a few annotated data, ii) low inter-/intra-observers agreement, iii) high-class imbalance, iv) inter-/intra-scanners variability and v) domain shift. Our research portfolio can be categorized into Learn to Recognize, Adapt, Learn, Reason and Explain, incorporate prior knowledge.
- Federated Learning in Healthcare: We will focus our research on developing innovative deep Federated Learning algorithms that can distill and share the knowledge among AI agents in a robust and privacy-preserved fashion. Research topics include, but not limited to, i) handling distributed DL models with data heterogeneity including non i.i.d, and domain shifts, ii) developing explainability and quality control tools for distributed models, and iii) robustness to poisoning models.
- Affordable AI and Healthcare: In addition, we are also interested in developing affordable AI solutions suitable for poor-quality data generated by low infrastructure and point-of-care diagnosis.
- Rieke, N., Hancox, J., Li, W., Milletari, F., Roth, H.R., Albarqouni, S., Bakas, S., Galtier, M.N., Landman, B.A., Maier-Hein, K. and Ourselin, S., 2020. The future of digital health with federated learning. NPJ digital medicine, 3(1), pp.1-7.
- Baur, C., Denner, S., Wiestler, B., Navab, N. and Albarqouni, S., 2021. Autoencoders for unsupervised anomaly segmentation in brain mr images: A comparative study. Medical Image Analysis, p.101952.
- Tomczak, A., Ilic, S., Marquardt, G., Engel, T., Forster, F., Navab, N. and Albarqouni, S., 2020. Multi-task multi-domain learning for digital staining and classification of leukocytes. IEEE Transactions on Medical Imaging.
- Lahiani, A., Klaman, I., Navab, N., Albarqouni, S*. and Klaiman, E*., 2020. Seamless virtual whole slide image synthesis and validation using perceptual embedding consistency. IEEE journal of biomedical and health informatics.
- Sarhan, M.H., Navab, N., Eslami, A. and Albarqouni, S., 2020, August. Fairness by learning orthogonal disentangled representations. In European Conference on Computer Vision (pp. 746-761). Springer, Cham.
- Bui, M., Birdal, T., Deng, H., Albarqouni, S., Guibas, L., Ilic, S. and Navab, N., 2020, 6D Camera Relocalization in Ambiguous Scenes via Continuous Multimodal Inference. In European Conference on Computer Vision (pp 139-157). Springer, Cham.
- Kazi, A., Shekarforoush, S., Krishna, S.A., Burwinkel, H., Vivar, G., Kortüm, K., Ahmadi, S.A., Albarqouni, S. and Navab, N., 2019, June. InceptionGCN: receptive field aware graph convolutional network for disease prediction. In International Conference on Information Processing in Medical Imaging (pp. 73-85). Springer, Cham.
- Degel, M.A., Navab, N. and Albarqouni, S., 2018, September. Domain and geometry agnostic CNNs for left atrium segmentation in 3D ultrasound. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 630-637). Springer, Cham.
- Albarqouni, S., Fotouhi, J. and Navab, N., 2017, September. X-ray in-depth decomposition: Revealing the latent structures. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 444-452). Springer, Cham.
- Albarqouni, S., Baur, C., Achilles, F., Belagiannis, V., Demirci, S. and Navab, N., 2016. Aggnet: deep learning from crowds for mitosis detection in breast cancer histology images. IEEE transactions on medical imaging, 35(5), pp.1313-1321.
A full list of publication can be found here.
Helmholtz AI Project on Eyesight to AI: Discovery of efficient corrosion modulators via predictive machine learning models in collaboration with Dr. Christian Feiler from Hereon.