Prediction of Cardiovascular Disease Based on Deep Learning Approaches

Yu, Shixiang

User: Guest

Shixiang Yu

If you experience problems opening the document, please try this link.

Original title:: Prediction of Cardiovascular Disease Based on Deep Learning Approaches
Translated title:: Vorhersage von Herz-Kreislauf-Erkrankungen auf der Grundlage von Deep-Learning-Verfahren
Author:: Yu, Shixiang
Year:: 2026
Document type:: Dissertation
Faculty/School:: TUM School of Medicine and Health
Institution:: Lehrstuhl für Translationale Genomik (HMGU) (Prof. Zeggini)
Advisor:: Zeggini, Eleftheria (Prof. Dr.)
Referee:: Zeggini, Eleftheria (Prof. Dr.); Rückert, Daniel (Prof. Dr.)
Language:: en
Subject group:: MED Medizin
TUM classification:: BIO 180; CIT 972
Abstract:: This study addresses the class imbalance in cardiovascular disease (CVD) datasets using generative models. GANs and diffusion models were used to create synthetic clinical data. A multimodal deep learning model integrating phenotypic and metabolomic features was developed, enhancing prediction of myocardial infarction and broader CVD risk. SHAP-based interpretation revealed key contributing features, supporting early, accurate risk assessment.
Translated abstract:: Diese Studie adressiert das Klassenungleichgewicht in Datensätzen zu Herz-Kreislauf-Erkrankungen (CVD) mithilfe generativer Modelle. Es wurden GANs und Diffusionsmodelle eingesetzt, um synthetische klinische Daten zu erzeugen. Ein multimodales Deep-Learning-Modell, das phänotypische und metabolomische Daten integriert, wurde entwickelt, um das Risiko eines Myokardinfarkts und anderer CVDs präzise vorherzusagen.
WWW:: https://mediatum.ub.tum.de/?id=1788262
Date of submission:: 21.07.2025
Oral examination:: 21.03.2026
File size:: 8786648 bytes
Pages:: 138
Urn (citeable URL):: https://nbn-resolving.org/urn:nbn:de:bvb:91-diss-20260321-1788262-0-4
Published:: 05.05.2026
BibTeX