The webinar on the 7^th of December will be about:

Explainability models for the early detection of Alzheimer’s by Christoph Friedrich (FH Dortmund)

Zoom-Link: https://zhaw.zoom.us/j/67277622484?pwd=NFJwZFNrNTlsdkdFUTQ5aHB6UURZZz09

Abstract: Hard-to-interpret black-box Machine Learning (ML) has been used regularly for early Alzheimer’s Disease (AD) detection. In this talk,
methods to help the interpretation of different machine learning models will be discussed. All models were trained on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset and evaluated for an independent ADNI test set, as well as the external Australian Imaging and Lifestyle flagship study of Ageing (AIBL), and Open Access Series of Imaging Studies (OASIS) datasets.

Shapley values will be compared to intuitively interpretable Decision Trees, and Logistic Regression, as well as natural and permutation-based feature importances. When features are correlated, explanations are no longer valid. Forward selection and aspect consolidation will be presented to mitigate this problem. Some black-box models outperformed DTs and LR. The forward selected features correspond to brain areas previously associated with AD. Shapley values identified biologically plausible associations with moderate-to-strong correlations with feature importance. The most important RF features to predict AD conversion were the volume of the amygdalae and a cognitive test score. The presented methods can be transferred easily to other indications.

Short biography: Christoph M. Friedrich received a diploma in computer science from the University of Dortmund, Germany, in 1996 and a PhD degree in life science engineering from the University of Witten/Herdecke, Germany, in 2006. He joined the Computer Science Department, University of Applied Sciences and Arts Dortmund, Germany, as a biomedical computer science professor, in 2010. Since 2021, Professor Friedrich is also leading the medical informatics group at the Institute for Medical Informatics, Biometry and Epidemiology (IMIBE) at the University Hospital Essen, Germany. From 2005 to 2010, he led the Data Mining Group in the Bioinformatics Department, Fraunhofer Institute for Algorithms and Scientific Computing (SCAI). He is author or co-author of more than 120 publications in international journals or conference proceedings. His main interests are machine learning, text mining, biomedical applications of computer vision, and bioinformatics.