Research Publications

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 5 of 1121
  • Item
    Heartbeat Detection in Photoplethysmography Signals for the Monitoring of Cardiac Arrhythmias
    (2024-01-24) Jeanningros, Loïc; Meister, Théo A.; Soria, Rodrigo; Tanner, Hildegard; Vesin, Jean-Marc; Thiran, Jean-Philippe; Lemay, Mathieu; Braun, Fabian; Rexhaj, Emrush
    Cardiac Arrhythmias (CAs) are a critical health issue associated with serious complications, such as stroke or heart failure. Wearable devices based on photoplethysmography (PPG), which can be worn in everyday life, have the potential to detect CAs earlier than current methods relying on electrocardiography (ECG). They could therefore enable a more preventive approach to treatment. While most PPG-based heartbeat detection algorithms have been evaluated on normal sinus rhythm or atrial fibrillation in clinical settings, their performance in patients with other cardiac arrhythmias in ambulatory settings remains unexplored to date. Methods: The PPG-beats framework, developed by Charlton and colleagues, evaluates the performance of several open-source heartbeat detectors. We applied the PPG-beats framework on a newly acquired dataset including forty-four patients referred for an ambulatory Holter ECG at Inselspital in Bern. This dataset comprises not only atrial fibrillation, but also bigeminy and premature contractions. Results: The heartbeat detector named MSPTD, performed best on normal sinus rhythm (with a median F1 score of 94.6%) and the detector named QPPG was top-ranked both on atrial fibrillation (91.6%) and bigeminy (80.0%). Conclusions: The heartbeats detectors named QPPG and MSPTD consistently achieved higher performance than other detectors. However, the detection of heartbeats from PPG signals is compromised in presence of bigeminy. This study sets the stage for continuous monitoring of cardiac arrhythmias in everyday life.
  • Item
    Data-Driven IQC-Based Robust Control Design for Hybrid Micro-Disturbance Isolation Platform
    (2023-12-15) Klauser, Elias; Gupta, Vaibhav; Karimi, Alireza
    A novel approach for robust controller synthesis, which models uncertainty as an elliptical set, is proposed in the paper. Given a set of frequency response functions of linear time-invariant (LTI) multiple-input multiple-output (MIMO) systems, the approach determines the ‘best’ linear nominal model and the corresponding elliptical uncertainty set, which is consistent with the data. Using a novel split representation, the uncertainty set is represented as an equivalent integral quadratic constraint (IQC). Finally, this IQC is integrated into a data-driven frequency-domain controller synthesis method using convex optimisation. The proposed method is used to design a controller, which is robust against mechanical uncertainties for a hybrid micro-disturbance isolation platform for space applications. The experimental results show that the proposed method provides a less conservative uncertainty set and improves attenuation performance compared to classical methods that use disk uncertainty.
  • Item
    PrivaTree: Collaborative Privacy-Preserving Training of Decision Trees on Biomedical Data
    (2023-06-15) Zein, Yamane El; Lemay, Mathieu; Huguenin, Kévin
    Biomedical data generation and collection have become faster and more ubiquitous. Consequently, datasets are increasingly spread across hospitals, research institutions, or other entities. Exploiting such distributed datasets simultaneously can be beneficial; in particular, classification using machine learning models such as decision trees is becoming increasingly common and important. However, given that biomedical data is highly sensitive, sharing data records across entities or centralizing them in one location are often prohibited due to privacy concerns or regulations. We design PrivaTree, an efficient and privacy-preserving protocol for collaborative training of decision tree models on distributed, horizontally partitioned, biomedical datasets. Although decision tree models may not always be as accurate as neural networks, they have better interpretability and are helpful in decision-making processes, which are crucial for biomedical applications. PrivaTree follows a federated learning approach, where raw data is not shared, and where every data provider computes updates to a global decision tree model being trained, on their private dataset. This is followed by privacy-preserving aggregation of these updates using additive secret-sharing, in order to collaboratively update the model. We implement PrivaTree, and evaluate its computational and communication efficiency on three different biomedical datasets, as well as the accuracy of the resulting models. Compared to the model centrally trained on all data records, the obtained collaborative model presents a modest loss of accuracy, while consistently outperforming the accuracy of the local models, trained separately by each data provider. Moreover, PrivaTree is more efficient than existing solutions, which makes it usable for training decision trees with numerous nodes, on large complex datasets, with both continuous and categorical attributes, as often found in the biomedical field.