MILA seminar - both in person and online on the 24th of March we will be hearing from Anne-Christin Hauschild from the Institute of Medical Informatics, University of Göttingen. She will be presenting on "Addressing the Hurdles in Developing Predictive Models in Healthcare".
https://www.medizin.uni-greifswald.de/medizininformatik/news/singleview/?tx_ttnews%5Btt_news%5D=63&cHash=02c5afda31ea2a437b47c26cb98c1af9
Machine learning has shown promising results in biomedical research by integrating clinical, molecular, and medical image data for disease classification and biomarker identification. However, challenges such as limited sample sizes, data heterogeneity, and lack of model interpretability hinder clinical adoption. To address these, novel architectures and algorithms are developed to enhance diagnostics and therapy optimisation. The FAIrPaCT project uses federated AI with privacy-preserving methods to enable large-scale medical data analysis without sharing raw patient data. Transfer learning is also employed to overcome data scarcity, particularly in model organism research, while explainable AI (XAI) methods improve model transparency and understanding. These innovations aim to enhance personalised medicine and clinical decision-making.