SSSusanne Suter

Susanne Suter

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Susanne Suter has a total of 77 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Wearables in der Patientenversorgung: Vernetzt, personalisiert, durchgängig

    by Samuel Wehrli, Dominik Kunz, Ayla Durrer, Sven Hirsch, Maren Leifke, Bianca Hölz, Jens Eckstein, Joris Kirchberger, Guido Perrot, Cornelia Gujer, et al.

  • Identifying Severities in Osteoarthritis

    by Caitlyn Zuppinger, Vincent Stadelmann, Susanne Suter

    Knee osteoarthritis (OA) severity is commonly assessed using the Kellgren-Lawrence (KL) grading system, but KL-scores are discrete, subjective, and variable in early disease, making them a noisy target for clinical stratification and machine learning. In addition, datasets used for OA classification are often strongly class-imbalanced, with a predominance of healthy to mid-range cases. We propose a KL-independent OA severity representation with unsupervised clustering of patient-reported outcomes and demographic features. Using a dataset of knee radiographs and clinical data, we first derive a radiographic anomaly score (XA-score) using a Siamese X-ray anomaly detection model. Questionnaire features are then clustered using UMAP and HDBSCAN, and cluster structure is evaluated against the XA-score. A multi-modal variant incorporating the XA-score during clustering is also explored. Radiograph-derived representations separate advanced OA well, questionnaire features yield clearer cluster structure, and the multi-modal configuration provides the most consistent clustering, producing coherent symptom profiles and strong separation of severe OA, though early disease remains difficult to distinguish under KL-based benchmarks. Expert inspection further supports the visual coherence of questionnaire-derived clusters. Overall, these findings support multi-modal, unsupervised stratification as a promising direction toward clinically meaningful OA phenotyping beyond traditional KL-based severity grading.

  • Open Access

    Evaluating a Telemedical Follow-Up Program for Continuity of Care After Hospital Discharge: Prospective Clinical Intervention Study

    by Jasmin Stéphanie Gram, Rebecca Meier, Bianca Hölz, Khaled Abdelhamid, Tobias Lang, Linus Imfeld, Henry Rotte, Iris Warthmann, Simone Zweipfenning, Kerstin Casper, et al.

    Background: The transition from hospital to primary care is a vulnerable period for patients. Telemedicine may enhance continuity of care; however, evidence on its role in supporting hospital-to-primary-care transitions remains limited. Objective: We implemented and evaluated a telephone-based follow-up program to support the transition from hospital to home and primary care, focusing on patient health as well as patient and health care provider satisfaction. Methods: In this prospective, single-center intervention study (University Hospital Basel, September 2022-December 2024), 234 patients discharged from the emergency unit or internal medicine wards received structured telemedical follow-up for up to 10 days until their first primary care appointment. Primary outcomes were patient health and satisfaction; secondary outcomes were health care provider satisfaction. Data were collected using patient-reported outcome measures, patient-reported experience measures, and provider assessments and analyzed descriptively and analytically ( P =.05). Results: Patient-reported outcome measure and patient-reported experience measure scores changed during follow-up, while no deterioration was observed. Health care provider satisfaction varied, with telemedical physicians reporting the highest ratings, hospital physicians intermediate ratings, and general practitioners the lowest, citing challenges in information transfer and perceived added value. Conclusions: This study outlines both the potential benefits and the practical challenges of implementing a telephone follow-up program after hospital discharge. Variations in physician satisfaction highlight the need for more user-friendly technical infrastructure and clearer role definitions. Future multicenter studies with broader patient samples, usual care controls, and simplified recruitment processes are required to strengthen the feasibility and generalizability of this approach.


  • Semantic or Instance Segmentation? A Dental Imaging Perspective

    by Simone Bonato, Melanie Mathys, Tobias Minder, Julia Hartmann, Florian Schärer, Lisa Kropf, Martin Melchior, Susanne Suter

    Dental X-rays play a crucial role in assessing oral health and supporting the diagnosis of conditions affecting teeth and bone structures. We investigated the automated identification of areas of interest for potential caries lesion candidates in Xrays to support the human caries detection process. To achieve this, we investigated whether computationally efficient semantic segmentation models can rival the performance of instance segmentation models. We challenged a highly accurate two-stage instance segmentation approach by comparing it with semantic segmentation enhanced with advanced training strategies that use unlabeled data and tooth detection information. Compared to most previous work, we trained our models on a larger dataset (4,000 bitewings). Our top semantic model, the multitask Dynamic U-Net (F1 0.69, recall 0.66, precision 0.73), scored comparably to our Mask R-CNN instance model (F1 0.65, recall 0.80, precision 0.54). Both models achieved scores comparable to or higher than those of previous studies that used object detection metrics and larger datasets. Qualitative expert evaluation confirmed that semantic segmentation models perform similar to instance segmentation. Mask R-CNN training took two to three times longer than for Dynamic U-Net and DeepLabV3+, respectively. These findings suggest that semantic segmentation models provide a computationally efficient automation approach to visualize areas of interest for caries lesion candidates.

  • Using Acoustic Signal to Predict Grain Size of Bedload Particles

    by Firat Saritas, Simon Staehli, Lorenz Ammann, Dieter Rickenmann, Rocco Custer, Susanne Suter

    Bedload transport is a potentially hazardous environmental process that occurs in rivers during high-flow conditions. Indirect bedload transport measuring systems relying on acoustic impact sensors help to obtain measurements of high temporal resolution. In this work, we evaluated the use of machine learning approaches to predict the size of bedload particles based on indirect acoustic sensor measurements. Data from controlled field experiments was available for three measuring systems from the Swiss Federal Research Institute WSL: Swiss plate geophone (SPG), miniplate accelerometer (MPA), and square pipe system (SPS). A simple and a complex feature set (9 vs. 44 features for the SPG) were created from raw signals for all measuring systems. Out of 9 evaluated model types, CatBoost models in combination with the complex feature set performed best and achieved R2-scores above 0.8 for all measuring systems. Models trained on data of the most recently developed measuring system (SPS) achieved the best results with an R2-score of 0.84. Finally, a feature relevance analysis revealed that the spectral roll-off and spectral centroid contributed most to the model predictions.


  • ICU Cockpit: a platform for collecting multimodal waveform data, AI-based computational disease modeling and real-time decision support in the intensive care unit

    by Jens Michael Boss, Gagan Narula, Christian Straessle, Jan Willms, Jan Azzati, Dominique Brodbeck, Rahel Luethy, Susanne Suter, Christof Buehler, Carl Muroi, et al.

    ICU Cockpit: a secure, fast, and scalable platform for collecting multimodal waveform data, online and historical data visualization, and online validation of algorithms in the intensive care unit. We present a network of software services that continuously stream waveforms from ICU beds to databases and a web-based user interface. Machine learning algorithms process the data streams and send outputs to the user interface. The architecture and capabilities of the platform are described. Since 2016, the platform has processed over 89 billion data points (N = 979 patients) from 200 signals (0.5–500 Hz) and laboratory analyses (once a day). We present an infrastructure-based framework for deploying and validating algorithms for critical care. The ICU Cockpit is a Big Data platform for critical care medicine, especially for multimodal waveform data. Uniquely, it allows algorithms to seamlessly integrate into the live data stream to produce clinical decision support and predictions in clinical practice.


  • Open Access

    Unraveling the deep learning gearbox in optical coherence tomography image segmentation towards explainable artificial intelligence

    by Peter M. Maloca, Philipp L. Müller, Aaron Y. Lee, Adnan Tufail, Konstantinos Balaskas, Stephanie Niklaus, Pascal Kaiser, Susanne Suter, Javier Zarranz-Ventura, Catherine Egan, et al.

    Machine learning has greatly facilitated the analysis of medical data, while the internal operations usually remain intransparent. To better comprehend these opaque procedures, a convolutional neural network for optical coherence tomography image segmentation was enhanced with a Traceable Relevance Explainability (T-REX) technique. The proposed application was based on three components: ground truth generation by multiple graders, calculation of Hamming distances among graders and the machine learning algorithm, as well as a smart data visualization (‘neural recording’). An overall average variability of 1.75% between the human graders and the algorithm was found, slightly minor to 2.02% among human graders. The ambiguity in ground truth had noteworthy impact on machine learning results, which could be visualized. The convolutional neural network balanced between graders and allowed for modifiable predictions dependent on the compartment. Using the proposed T-REX setup, machine learning processes could be rendered more transparent and understandable, possibly leading to optimized applications.