Education and Training
Dr. Bettina Katalin Budai is a postdoctoral researcher at DIR UKHD, leading a newly forming interdisciplinary research group focused on the development of deep learning-based image analysis tools to support automated assessment targeting a diverse range of pathologies in the context of body imaging. She holds an MD degree, and a PhD degree in Clinical Science with a specialization in artificial intelligence and radiomics analysis in abdominal imaging from Semmelweis University, Hungary. In 2024, she was awarded a fellowship in the Medical Data Scientist Program at the Medical Faculty of the University of Heidelberg.
Scientific focus and expertise
Our research group aims to develop deep learning-based image analysis tools for medical applications in the field of body imaging. Our interdisciplinary team consists of computer scientists, experienced radiologists, and radiology residents. The research projects within the TLRC focus on AI-based body composition analysis, AI-based spine analysis for the assessment of comorbidities, and deep learning and radiomics analyses of lung cancer on CT. In 2025, the UKHD was selected to join the EUCAIM consortium with a project led by Dr. Bettina Katalin Budai and Prof. Dr. med. Hans-Ulrich Kauczor, which aims for the multicentric validation of the research group’s AI tool that combines deep learning and radiomics for the assessment of advanced NSCLC on CT scans.
- The role of AI-based CT body composition in COPD
- AI-based spine analysis for the assessment of osteoporosis and compression fractures in COPD patients
- AI-based analysis of advanced-stage lung cancer on CT scans using deep learning and radiomics analysis
- COPD
- Lung Cancer
PD Dr. med. Philipp Mayer | Senior physician, Specialist in radiology | Philipp.Mayer@med.uni-heidelberg.de |
Dr. med. Viktoria Palm | Senior physician,Specialist in radiology | Viktoria.Palm@med.uni-heidelberg.de |
Dr. med. Tobias Nonnenmacher | Senior physician | Tobias.Nonnenmacher@med.uni-heidelberg.de |
Dr. med. univ. Róbert Stollmayer | Assisent Physician | Robert.Stollmayer@med.uni-heidelberg.de |
Ondrej Havlicek | PhD Stundent | Ondrej.Havlicek@med.uni-heidelberg.de |
Hanyi Zhang | PhD Student | Hanyi.Zhang@med.uni-heidelberg.de |
Lung Research - Projects
1. The role of AI-based CT body composition in COPD
This study investigates the use of AI-based body composition analysis on COPD patients’ chest CT scans. The aim is to test whether AI-supported CT body composition analysis (BCA) is a suitable alternative to bioelectrical impedance analysis (BIA) as a clinical reference for identifying patients with an increased risk of sarcopenia. Moreover, the AI-based BCA measures are investigated for body phenotyping, for prognosis prediction, and for their potential association with comorbidities such as osteoporosis and vascular calcification in COPD patients.
2. AI-based spine analysis for the assessment of osteoporosis and compression fractures in COPD patients
This project aims to develop an AI-based tool for the spine that is robust against institutional differences in clinical CT protocols (e.g., different reconstruction kernels, slice thicknesses, contrast agent use) and anatomical variances. A special focus is paid to spinal anomalies, artifacts, implants, and osteoporosis-mimicking diseases, confounding factors that previously published models did not address specifically. The goal is to develop a user-friendly analysis tool that enables robust evaluation of routine CT scans, facilitating the detection and assessment of osteoporotic compression fractures in COPD patients.
3. AI-based analysis of advanced-stage lung cancer on CT scans using deep learning and radiomics analysis
The currently available AI-based algorithms are limited to the assessment of lung nodules and are not applicable to patients with large tumor masses or tumors localized around large blood vessels. However, patients with NSCLC are usually diagnosed at more advanced stages. Accurate detection and segmentation of these advanced-stage tumors could allow automated extraction of radiomics features that correlate with genetic mutations, paving the way for radiomics-based "virtual biopsies". This study aims to develop our AI tool to automatically segment and assess these advanced-stage lung cancers. The project investigates the generalizability on a multicentric dataset of the EUCAIM database.