Détection des artefacts dans les signaux de photopléthysmographie

Abstract

The main objective of this project is to ensure accurate detection of motion artifacts in photoplethysmographic (PPG) signals, a crucial step in guaranteeing the reliability and accuracy of clinical and research analyses based on these data. Among the many types of artifacts, we were particularly interested in motion artifacts, as they have the potential to significantly disrupt PPG signals and compromise the interpretation of physiological parameters. To achieve this goal, we undertook a machine learning-based approach. Our dissertation details the entire process, starting with the initial pre-processing of the PPG signal, continuing with data annotation and the exploration of various data re-balancing methods, and finally culminating in a comparison of the performance of the supervised classifiers with that of the semi-supervised label propagation algorithm. The results obtained show that the semi-supervised label propagation machine learning algorithm achieved a precision of 91%, a recall of 90% and an F1 score of 90%. Although the supervised KNN (K-Nearest Neighbors) algorithm also achieved strong results, with a precision of 89%, a recall of 95% and an F1 score of 92%, the label propagation algorithm proved more effective at accurately detecting motion artifacts. In summary, this research contributes to the advancement of knowledge in the field of artifact detection in PPG signals, by highlighting the effectiveness of the semi-supervised label propagation algorithm in detecting motion artifacts.

Date	14 Dec 2023
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Rita Noumeir (Supervisor) & Philippe Jouvet (Co-supervisor)