Regensburg Center of Biomedical Engineering - RCBE
Refine
Year of publication
Document Type
- Article (105)
- conference proceeding (article) (95)
- conference talk (43)
- conference proceeding (presentation, abstract) (34)
- Part of a Book (4)
- conference proceeding (volume) (4)
- Doctoral Thesis (2)
- Moving Images (2)
- Other (2)
- Preprint (2)
Is part of the Bibliography
- no (294)
Keywords
- Simulation (12)
- Deep Learning (10)
- Biomechanics (9)
- Biomechanik (9)
- Bildverarbeitung (8)
- Bildgebendes Verfahren (7)
- Hüftgelenkprothese (7)
- Lernprogramm (7)
- Biomechanische Analyse (6)
- Handchirurgie (6)
Institute
- Regensburg Center of Biomedical Engineering - RCBE (294)
- Fakultät Maschinenbau (234)
- Labor Biomechanik (LBM) (148)
- Regensburg Center of Health Sciences and Technology - RCHST (62)
- Fakultät Informatik und Mathematik (54)
- Labor Regensburg Medical Image Computing (ReMIC) (46)
- Labor Innovation & Regulatory Affairs (IRA) (33)
- Labor Biofluidmechanik (31)
- Labor Medizinprodukte (20)
- Labor eHealth (eH) (5)
Begutachtungsstatus
- peer-reviewed (115)
- begutachtet (12)
Person-specific evaluation method for occupational exoskeletons - Biomechanical body heat map
(2025)
Human-centred and ergonomic work design is one of the most important drivers for increasing the competitiveness of the European Union. As a flexible, person-specific occupational measure, exoskeletons promise great potential for effectively reducing individual ergonomic stress. Digital human models can provide important insights and offer great potential for systematising the effect and targeted use of exoskeletons, supporting their effective implementation in practice. In this article, digital human models are applied on two levels. Firstly, a realistic industrial logistics scenario in which boxes had to be relocated is designed with the help of a digital human model for workplace and process planning and secondly, a new biomechanical evaluation methodology to analyse intended and unintended effects on internal stress on the human body is demonstrated by applying musculoskeletal exoskeleton human models of four test subjects. Finally, the modelled biomechanical support tendencies of one exoskeleton are preliminary validated using EMG measurement data of the back muscles collected from the four male workers. The preliminary analysis of two back-support exoskeletons to demonstrate the new methodological approach confirms the expected, intended effects in the lower back and reveals unintended effects, such as e.g. changes in knee kinetics when applying a soft or hard-frame exoskeleton. Furthermore, the exemplary results to demonstrate the methodological approach expose notable differences between the test subjects, which underlines the relevance of person-specific evaluation and consideration of exoskeleton support. The preliminary validation shows a correlation between the modelled and the EMG-measured biomechanical exoskeleton support of the considered back muscles.
Intra-abdominal pressure is a significant physiological parameter influencing spinal stability and pelvic floor health. This study investigates the potential of a transformer encoder model to predict IAP using motion capture data and musculoskeletal modeling. Data from 211 subjects performing walking, fast walking, and running were used to train a transformer encoder. The model showed promising results with an overall Mean Absolute Percentage Error of 13.5% and a Pearson correlation coefficient of 0.85. Predictions for fast walking and running proved to be more challenging compared to walking, which was attributed to the greater variability and complexity of faster movements.
Purpose
Recognizing previously unseen classes with neural networks is a significant challenge due to their limited generalization capabilities. This issue is particularly critical in safety-critical domains such as medical applications, where accurate classification is essential for reliability and patient safety. Zero-shot learning methods address this challenge by utilizing additional semantic data, with their performance relying heavily on the quality of the generated embeddings.
Methods
This work investigates the use of full descriptive sentences, generated by a Sentence-BERT model, as class representations, compared to simpler category-based word embeddings derived from a BERT model. Additionally, the impact of z-score normalization as a post-processing step on these embeddings is explored. The proposed approach is evaluated on a multi-label generalized zero-shot learning task, focusing on the recognition of surgical instruments in endoscopic images from minimally invasive cholecystectomies.
Results
The results demonstrate that combining sentence embeddings and z-score normalization significantly improves model performance. For unseen classes, the AUROC improves from 43.9% to 64.9%, and the multi-label accuracy from 26.1% to 79.5%. Overall performance measured across both seen and unseen classes improves from 49.3% to 64.9% in AUROC and from 37.3% to 65.1% in multi-label accuracy, highlighting the effectiveness of our approach.
Conclusion
These findings demonstrate that sentence embeddings and z-score normalization can substantially enhance the generalization performance of zero-shot learning models. However, as the study is based on a single dataset, future work should validate the method across diverse datasets and application domains to establish its robustness and broader applicability.
Reliable recognition and localization of surgical instruments in endoscopic video recordings are foundational for a wide range of applications in computer- and robot-assisted minimally invasive surgery (RAMIS), including surgical training, skill assessment, and autonomous assistance. However, robust performance under real-world conditions remains a significant challenge. Incorporating surgical context - such as the current procedural phase - has emerged as a promising strategy to improve robustness and interpretability. To address these challenges, we organized the Surgical Procedure Phase, Keypoint, and Instrument Recognition (PhaKIR) sub-challenge as part of the Endoscopic Vision (EndoVis) challenge at MICCAI 2024. We introduced a novel, multi-center dataset comprising thirteen full-length laparoscopic cholecystectomy videos collected from three distinct medical institutions, with unified annotations for three interrelated tasks: surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation. Unlike existing datasets, ours enables joint investigation of instrument localization and procedural context within the same data while supporting the integration of temporal information across entire procedures. We report results and findings in accordance with the BIAS guidelines for biomedical image analysis challenges. The PhaKIR sub-challenge advances the field by providing a unique benchmark for developing temporally aware, context-driven methods in RAMIS and offers a high-quality resource to support future research in surgical scene understanding.
Many dynamic hand orthoses use one degree of freedom joints, such as hinge joints. Therefore, these orthoses can only partially replicate the complex, multi-axis movement of the hand. A possible solution for this is the use of prestressed compliant structures as the basis for orthoses. Determining the joint forces in the wrist and optimizing the dynamic orthosis to influence these forces as well as acting muscle forces are important steps in the development of these orthoses. For this reason, in this work multi-body simulation models of an orthosis with human hand models are presented. Based on these theoretical investigations, more detailed orthosis models as well as initial prototypes of prestressed compliant dynamic hand orthoses can be developed.
AIM:
To develop a 3D-printed model enabling treatment simulation of trauma-related endodontic and restorative procedures like revitalization or placement of an apical plug on teeth with incomplete root formation.
METHODOLOGY:
To generate a realistic training model suitable for trauma-related simulations, CT-Data sets were segmented, combined, adjusted and optimized using computer-aided design features. Specific focus was on optimally reflecting characteristics of immature teeth like open apical foramina and thin root walls. For reflection of the revitalization process, a set-up including a tube filled with red liquid under pressure was developed to be perforated during the procedure by the trainee, simulating bleeding from the apical papilla into the root canal system. Design was based on a combination of cost-effective simple parts combined with 3D-printed components to achieve maximal accessibility, exchangeability and re-usability. Model assembly and preparation were described in step-by-step instructions.
RESULTS:
As a result of the developmental process, the presented model qualifies for the training of the endodontic management of complications on immature teeth after dental trauma. Step-by-step descriptions aligned with the clinical procedures are performable for a revitalization procedure, for pulpotomy followed by fragment re-attachment and splint application and for placement of an apical plug using hydraulic calcium silicate cements.
CONCLUSIONS:
The model successfully achieved its intended objectives, so that it may be used in the future in various dental trauma treatment simulations for undergraduate and postgraduate education on the endodontic management of trauma-related therapies in immature teeth. Its common manufacturing process, affordability and re-usability support accessibility and sustainability.
BACKGROUND: A widely accepted tool to assess hemodynamics, one of the most important factors in aneurysm pathophysiology, is Computational Fluid Dynamics (CFD). As current workflows are still time consuming and difficult to operate, CFD is not yet a standard tool in the clinical setting. There it could provide valuable information on aneurysm treatment, especially regarding local risks of rupture, which might help to optimize the individualized strategy of neurosurgical dissection during microsurgical aneurysm clipping.
METHOD: We established and validated a semi-automated workflow using 3D rotational angiographies of 24 intracranial aneurysms from patients having received aneurysm treatment at our centre. Reconstruction of vessel geometry and generation of volume meshes was performed using AMIRA 6.2.0 and ICEM 17.1. For solving ANSYS CFX was used. For validational checks, tests regarding the volumetric impact of smoothing operations, the impact of mesh sizes on the results (grid convergence), geometric mesh quality and time tests for the time needed to perform the workflow were conducted in subgroups.
RESULTS: Most of the steps of the workflow were performed directly on the 3D images requiring no programming experience. The workflow led to final CFD results in a mean time of 22 min 51.4 s (95%-CI 20 min 51.562 s-24 min 51.238 s, n = 5). Volume of the geometries after pre-processing was in mean 4.46% higher than before in the analysed subgroup (95%-CI 3.43-5.50%). Regarding mesh sizes, mean relative aberrations of 2.30% (95%-CI 1.51-3.09%) were found for surface meshes and between 1.40% (95%-CI 1.07-1.72%) and 2.61% (95%-CI 1.93-3.29%) for volume meshes. Acceptable geometric mesh quality of volume meshes was found.
CONCLUSIONS: We developed a semi-automated workflow for aneurysm CFD to benefit from hemodynamic data in the clinical setting. The ease of handling opens the workflow to clinicians untrained in programming. As previous studies have found that the distribution of hemodynamic parameters correlates with thin-walled aneurysm areas susceptible to rupture, these data might be beneficial for the operating neurosurgeon during aneurysm surgery, even in acute cases.
Advances in genomic research have created new privacy challenges. This scoping review analyzes the risks associated with the processing, storage, and sharing of genomic data including epigenetics, and examines current privacy protection strategies. It also attempts to identify research gaps in this area. Using the PRISMA methodology, 37 relevant studies were identified and analyzed. The results of the risk assessment can be grouped into four main themes: Risks posed by processing of functional genomic data, sharing of genomic data, patient (re-)identification, and dividuality, i.e. the extending of privacy risks to blood relatives. The identified risk mitigation strategies were systematically categorized into five classes: pre-release measures, governance, secure data processing and exchange, access restriction and transparency, anonymization and masking. However, there are some important research gaps that still need to be addressed. The current literature neglects to assess the likelihood of potential breaches and tends to focus only on assessing possible scenarios of privacy risks. It also mainly fails to assess the role of contextualized data and the effectiveness of policies and governance systems with respect to privacy risks.