h1

%0 Thesis
%A Han, Tianyu
%T Generative modeling for medical image analysis
%I RWTH Aachen University
%V Dissertation
%C Aachen
%M RWTH-2024-00999
%P 1 Online-Ressource : Illustrationen
%D 2024
%Z Veröffentlicht auf dem Publikationsserver der RWTH Aachen University
%Z Dissertation, RWTH Aachen University, 2024
%X Radiology is undergoing a transformation with the increasing integration of machine learning into medical imaging. This promises enhanced diagnostic precision and faster diagnosis. Yet, creating potent machine learning algorithms for this domain often demands vast volumes of labeled data. In intricate radiological scenarios, annotating these images can be both time-intensive and costly. To counteract these hurdles, we advocate for unsupervised learning methodologies in medical imaging analysis. We place emphasis on the deployment of a self-supervised generative model. This model is tailored to decipher latent temporal trajectories from longitudinal radiographs of patients, thereby facilitating model-driven forecasts regarding the emergence and evolution of osteoarthritis. Further, we introduce an innovative paradigm aimed at publicly sharing medical imaging data for algorithm training. This ensures the safeguarding of sensitive patient information. Alongside, we underline the imperative to elevate the transparency of deep learning in a clinical setting and bolster the algorithms’ resilience to adversarial onslaughts. In essence, by capitalizing on unsupervised learning strategies and pioneering ways to make medical imaging data both accessible and secure, we stand on the cusp of redefining radiological research, all while elevating patient care outcomes.
%F PUB:(DE-HGF)11
%9 Dissertation / PhD Thesis
%R 10.18154/RWTH-2024-00999
%U https://publications.rwth-aachen.de/record/977930