arXiv subjects

Oskar Pfeffer

Publications and source records attributed to Oskar Pfeffer.

6 recordsLinked to original sources

arXiv2026

Apnea Burden-Guided Framework: Enhancing Out-of-Distribution Generalization in PPG-Based Sleep Apnea Characterization

Sleep apnea is a common sleep-related breathing disorder associated with substantial cardiovascular and metabolic risk. Although overnight polysomnography remains the reference standard for diagnosis, its complexity and cost limit its suitability for long-term preventive monitoring at home. In wearable systems, arterial blood oxygen saturation (SpO2) is commonly used as the main predictor, whereas additional morphological features of the photoplethysmographic (PPG) pulse wave are usually underexplored. This study proposes a novel apnea burden prediction-based framework for sleep apnea severity assessment and investigates the influence of PPG features on model performance and out-of-distribution (OOD) generalization. The proposed framework first predicts apnea burden as a continuous measure, which is subsequently converted into the clinically relevant apnea-hypopnea index for subject-level classification into four severity groups in both in-distribution (ID) and OOD data. Three artificial neural network architectures were evaluated, and the performance metrics were averaged over five independent runs with different fixed random seeds. During OOD testing, the combination of PPG features and SpO2 led to increases of approximately 15.72% in macro-sensitivity, 9.22% in macro-accuracy, 16.01% in macro-F1-score, 11.08% in Cohen's kappa, and 13.22% in Matthews correlation coefficient, compared with using SpO2 alone. The low-complexity convolutional-recurrent models achieved the highest overall performance. The results indicated that the proposed apnea burden-guided framework, combined with PPG features and SpO2, improves sleep apnea characterization while showing encouraging OOD performance on an independent external dataset. Moreover, simpler hybrid architectures demonstrated strong potential for robust home-based preventive monitoring.

eess.SP↗

arXiv2026

Benchmark Problems and Benchmark Datasets for the evaluation of Machine and Deep Learning methods on Photoplethysmography signals: the D4 report from the QUMPHY project

This report is part of the Qumphy project (22HLT01 Qumphy) that is funded by the European Union and is dedicated to the development of measures to quantify the uncertainties associated with Machine Learning algorithms applied to medical problems, in particular the analysis and processing of Photoplethysmography (PPG) signals. In this report, a list of six medical problems that are related to PPG signals and serve as Benchmark Problems is given. Suitable Benchmark datasets and their usage are described also.

cs.LG↗

arXiv2025

A systematic evaluation of uncertainty quantification techniques in deep learning: a case study in photoplethysmography signal analysis

In principle, deep learning models trained on medical time-series, including wearable photoplethysmography (PPG) sensor data, can provide a means to continuously monitor physiological parameters outside of clinical settings. However, there is considerable risk of poor performance when deployed in practical measurement scenarios leading to negative patient outcomes. Reliable uncertainties accompanying predictions can provide guidance to clinicians in their interpretation of the trustworthiness of model outputs. It is therefore of interest to compare the effectiveness of different approaches. Here we implement an unprecedented set of eight uncertainty quantification (UQ) techniques to models trained on two clinically relevant prediction tasks: Atrial Fibrillation (AF) detection (classification), and two variants of blood pressure regression. We formulate a comprehensive evaluation procedure to enable a rigorous comparison of these approaches. We observe a complex picture of uncertainty reliability across the different techniques, where the most optimal for a given task depends on the chosen expression of uncertainty, evaluation metric, and scale of reliability assessed. We find that assessing local calibration and adaptivity provides practically relevant insights about model behaviour that otherwise cannot be acquired using more commonly implemented global reliability metrics. We emphasise that criteria for evaluating UQ techniques should cater to the model's practical use case, where the use of a small number of measurements per patient places a premium on achieving small-scale reliability for the chosen expression of uncertainty, while preserving as much predictive performance as possible.

cs.LG↗

arXiv2025

Machine-learning for photoplethysmography analysis: Benchmarking feature, image, and signal-based approaches

Photoplethysmography (PPG) is a widely used non-invasive physiological sensing technique, suitable for various clinical applications. Such clinical applications are increasingly supported by machine learning methods, raising the question of the most appropriate input representation and model choice. Comprehensive comparisons, in particular across different input representations, are scarce. We address this gap in the research landscape by a comprehensive benchmarking study covering three kinds of input representations, interpretable features, image representations and raw waveforms, across prototypical regression and classification use cases: blood pressure and atrial fibrillation prediction. In both cases, the best results are achieved by deep neural networks operating on raw time series as input representations. Within this model class, best results are achieved by modern convolutional neural networks (CNNs). but depending on the task setup, shallow CNNs are often also very competitive. We envision that these results will be insightful for researchers to guide their choice on machine learning tasks for PPG data, even beyond the use cases presented in this work.

cs.LG↗

arXiv2021

Moving the epidemic tipping point through topologically targeted social distancing

The epidemic threshold of a social system is the ratio of infection and recovery rate above which a disease spreading in it becomes an epidemic. In the absence of pharmaceutical interventions (i.e. vaccines), the only way to control a given disease is to move this threshold by non-pharmaceutical interventions like social distancing, past the epidemic threshold corresponding to the disease, thereby tipping the system from epidemic into a non-epidemic regime. Modeling the disease as a spreading process on a social graph, social distancing can be modeled by removing some of the graphs links. It has been conjectured that the largest eigenvalue of the adjacency matrix of the resulting graph corresponds to the systems epidemic threshold. Here we use a Markov chain Monte Carlo (MCMC) method to study those link removals that do well at reducing the largest eigenvalue of the adjacency matrix. The MCMC method generates samples from the relative canonical network ensemble with a defined expectation value of $λ_{max}$. We call this the "well-controlling network ensemble" (WCNE) and compare its structure to randomly thinned networks with the same link density. We observe that networks in the WCNE tend to be more homogeneous in the degree distribution and use this insight to define two ad-hoc removal strategies, which also substantially reduce the largest eigenvalue. A targeted removal of 80\% of links can be as effective as a random removal of 90\%, leaving individuals with twice as many contacts.

physics.soc-ph↗

arXiv2021

Relative Canonical Network Ensembles -- (Mis)characterizing Small-World Networks

What do generic networks that have certain properties look like? We define Relative Canonical Network ensembles as the ensembles that realize a property R while being as indistinguishable as possible from a generic network ensemble. This allows us to study the most generic features of the networks giving rise to the property under investigation. To test the approach we apply it first to the network measure "small-world-ness", thought to characterize small-world networks. We find several phase transitions as we go to less and less generic networks in which cliques and hubs emerge. Such features are not shared by typical small-world networks, showing that high "small-world-ness" does not characterize small-world networks as they are commonly understood. On the other hand we see that for embedded networks, the average shortest path length and total Euclidean link length are better at characterizing small-world networks, with hubs that emerge as a defining feature at low genericity. We expect the overall approach to have wide applicability for understanding network properties of real world interest.

physics.soc-ph↗