Explaining AI decisions in personalized healthcare: towards integration of deep learning into diagnosis and treatment planning for Multiple Sclerosis

This is a Haslerfoundation funded project under the “Responsible AI” program. Our ultimate goal is to evaluate the trustworthy behaviour of ML-based neuroimage analysis to support diagnosis and prognosis of Multiple Sclerosis patients. To this end we have established three objectives centered around engineering, physicians/end-users and patients:

1. New insights of DL-decisions will help us identifying any bias within our models and thus further improve them by reducing the bias.

2. Increasing confidence, quality of decision making and clinical impact with new interpretability and explainability strategies to be joined by uncertainty measures of DL-models.

3. Better understanding of disease progression using biologically interpretable measures of inflammation and degeneration.


MSxplain is an interdisciplinary and inter-cantonal project lead by Dr. M. Bach Cuadra (CIBM, CHUV-UNIL), Prof Cristina Granziera (ThINK, University of Basel), Prof Henning Müller and Prof. Adrien Depeursinge (medGIFT, HES-SO Valais). The project funds two PhD candidates Ms Nataliia Molchanova (UNIL, HES-SO) and Mr Federico Spagnolo (Basel, HES-SO) and part-time software engineer Mr Roger Schaer (HES-SO). The research is conducted jointly with Dr Mara Graziani (HES-SO).

Collaborators: F. La Rosa (Mount Sinai Hospital), V. Raina, Prof M. Gales (Oxford University), Dr. A. Malinin (Yandex).

Talks and publications

[custom-twitter-feeds hashtag=#MSxplain]

Partner institutions