Comparative Study on the Generalization Ability of Machine Learning Algorithms for PPG Quality Assessment

Photoplethysmogram (PPG) signals are often too noisy to accurately measure heart rate.Methods for PPG quality assessment based on machine learning (ML) concepts have gained considerable attention, but they have been commonly developed on datasets recorded under specific conditions and not tested on others.Hence, the aim of the present work is to compare their performance and generalization ability on datasets representative of different conditions.Three common classifiers, such as decision tree (DT), support vector machine (SVM), and K-Nearest Neighbor (KNN), were trained on a publicly available dataset and tested on two external datasets, also freely available.The obtained results reported a different generalization ability for each classifier.Whereas DT was unable to generalize for any external dataset, SVM and KNN were only able for an external dataset presenting PPG signals acquired with the same recording system and then with a morphology similar to those in the training one.Hence, the selection of datasets for training and validation seems a key aspect to be taken into account to know where a machine learning method can be reliably used for PPG quality assessment.