Moritz Bednorz
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Moritz Bednorz has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Alexander Meyer
DZHK (German Centre for Cardiovascular Research), partner site Berlin, Berlin, Germany
1 shared publication- Andreas Leha
Department of Medical Bioinformatics, University Medical Center Göttingen
1 shared publication - Anja Hennemuth
Deutsches Herzzentrum der Charite
1 shared publication - Clemens Scherer
Department of Medicine I, LMU University Hospital, LMU Munich, Munich, Germany
1 shared publication Eike Nagel
DZHK (German Centre for Cardiovascular Research), partner site RhineMain, Frankfurt, Germany
1 shared publicationFlorian André
DZHK (German Centre for Cardiovascular Research), Partner site Heidelberg/Mannheim, Heidelberg, Germany
1 shared publication
Latest Publications
- Open Access
Reconstruction of Respiratory Flow from an Impedance-Based Contactless Sensor System
Conventional respiratory monitoring is often invasive, while most non-contact technologies like radar or cameras are limited to estimating respiratory rate, failing to reconstruct the detailed waveform of the respiratory flow itself. This gap limits their clinical utility for advanced diagnostics. We introduce a novel system that bridges this gap by combining a contactless, impedance-based sensor (the Thoraxmonitor) with a dedicated machine learning framework to directly reconstruct the full respiratory flow signal. Operating at 433 , the system’s antenna array detects subtle changes in thoracic impedance, which are then translated into a quantitative flow signal by a Multilayer Perceptron Regressor. Based on data from 17 subjects benchmarked against a gold-standard flowmeter, our system accurately detected 98% of respiratory cycles. It achieved remarkable precision in timing respiratory events, with mean deviations of + 60 (±79 ) for inspiration and + 50 (±63 ) for expiration, making it suitable for time-critical applications. While a systematic bias in absolute tidal volume prediction currently limits inter-subject comparisons, the system excels at tracking relative intra-subject changes. Crucially, our model quantifies its own reliability, providing an intrinsic self-assessment mechanism. This work demonstrates a significant step beyond simple rate detection towards comprehensive, comfortable, and reliable respiratory analysis in clinical and everyday settings.Open AccessReconstruction of Respiratory Flow from an Impedance-Based Contactless Sensor System
Conventional respiratory monitoring is often invasive, while most non-contact technologies like radar or cameras are limited to estimating respiratory rate, failing to reconstruct the detailed waveform of the respiratory flow itself. This gap limits their clinical utility for advanced diagnostics. We introduce a novel system that bridges this gap by combining a contactless, impedance-based sensor (the Thoraxmonitor) with a dedicated machine learning framework to directly reconstruct the full respiratory flow signal. Operating at 433 , the system’s antenna array detects subtle changes in thoracic impedance, which are then translated into a quantitative flow signal by a Multilayer Perceptron Regressor. Based on data from 17 subjects benchmarked against a gold-standard flowmeter, our system accurately detected 98% of respiratory cycles. It achieved remarkable precision in timing respiratory events, with mean deviations of + 60 (±79 ) for inspiration and + 50 (±63 ) for expiration, making it suitable for time-critical applications. While a systematic bias in absolute tidal volume prediction currently limits inter-subject comparisons, the system excels at tracking relative intra-subject changes. Crucially, our model quantifies its own reliability, providing an intrinsic self-assessment mechanism. This work demonstrates a significant step beyond simple rate detection towards comprehensive, comfortable, and reliable respiratory analysis in clinical and everyday settings. - Top Cited PapersOpen Access
- 6 February 2025
- Published by Springer Nature in npj Digital Medicine
- Vol. 8 (1) , 1-14
Real world federated learning with a knowledge distilled transformer for cardiac CT imaging
Federated learning is a renowned technique for utilizing decentralized data while preserving privacy. However, real-world applications often face challenges like partially labeled datasets, where only a few locations have certain expert annotations, leaving large portions of unlabeled data unused. Leveraging these could enhance transformer architectures’ ability in regimes with small and diversely annotated sets. We conduct the largest federated cardiac CT analysis to date (n = 8, 104) in a real-world setting across eight hospitals. Our two-step semi-supervised strategy distills knowledge from task-specific CNNs into a transformer. First, CNNs predict on unlabeled data per label type and then the transformer learns from these predictions with label-specific heads. This improves predictive accuracy and enables simultaneous learning of all partial labels across the federation, and outperforms UNet-based models in generalizability on downstream tasks. Code and model weights are made openly available for leveraging future cardiac CT analysis.
- 6 February 2025
- Published by Springer Nature in npj Digital Medicine
- Vol. 8 (1) , 1-14
Top Cited PapersOpen AccessReal world federated learning with a knowledge distilled transformer for cardiac CT imaging
Federated learning is a renowned technique for utilizing decentralized data while preserving privacy. However, real-world applications often face challenges like partially labeled datasets, where only a few locations have certain expert annotations, leaving large portions of unlabeled data unused. Leveraging these could enhance transformer architectures’ ability in regimes with small and diversely annotated sets. We conduct the largest federated cardiac CT analysis to date (n = 8, 104) in a real-world setting across eight hospitals. Our two-step semi-supervised strategy distills knowledge from task-specific CNNs into a transformer. First, CNNs predict on unlabeled data per label type and then the transformer learns from these predictions with label-specific heads. This improves predictive accuracy and enables simultaneous learning of all partial labels across the federation, and outperforms UNet-based models in generalizability on downstream tasks. Code and model weights are made openly available for leveraging future cardiac CT analysis.