Robust Assessment of Photoplethysmogram Signal Quality in the Presence of Atrial Fibrillation

Abstract

A great deal of algorithms currently available to assess the quality of photoplethysmogram (PPG) signals is based on the similarity between pulses to derive signal quality indices. This approach has limitations when pulse morphology become variable due to the presence of some arrhythmia as in the case of atrial fibrillation (AFib). AFib is a heart arrhythmia characterized in the electrocardiogram mainly by an irregular irregularity. This arrhythmicity is reflected on PPG pulses by the presence of non-uniform pulses and poses challenges in the evaluation of the signal quality. In this work, we first test the performance of few algorithms from the body of methods reported in literature using a dataset of PPG records with AFib, and demonstrate their limitation. Second, we present a novel SVM-based classifier for PPG quality assessment in 30s-long segments of PPG records extracted from pulse oximetry data of 13 stroke patients admitted to the UCSF medical center neuro ICU. 40 time-domain, frequency domain and non-linear features were extracted from all segments. Using an independent test set, the classifier reached a 0.94 accuracy, 0.95 sensitivity and 0.91 specificity. These results demonstrate the robustness of the proposed method in properly evaluating PPG signal quality in the presence of atrial fibrillation.

A Survey of Photoplethysmography and Imaging Photoplethysmography Quality Assessment Methods

“…Manual annotation [51][52][53][54] Guided By rules [55,56] By reference signal [35,[57][58][59] Perturbations artificially created during recording PR = Pulse Rate, HR = Heart Rate, ECG = Electrocardiogram, cPPG = Contact Photoplethysmography.…”

“…They also compared their method using SVM with two other algorithms [43,77] on the Capnobase dataset [83]. Pereira et al [52] also found the best performance with their SVM classifier against seven representative methods [42,43,68,70,77,89,90]. Elgendi used the SVM algorithm to find the most discriminant feature to classify the quality of a PPG pulse [51].…”

Deep learning approaches for plethysmography signal quality assessment in the presence of atrial fibrillation

“…A body of studies has proposed approaches for pulsatile signal quality assessment that are applicable to PPG signals. Early approaches were based on measures of similarity between consecutive pulses (Karlen et al 2012, Pereira et al 2018, nevertheless, they fail to correctly assess the signals quality in the case of AF because consecutive PPG pulses become morphology different (Asgari et al 2010). Machine learning (ML) techniques can overcome this problem since they are trained with AF cases (Pereira et al 2018).…”

“…Early approaches were based on measures of similarity between consecutive pulses (Karlen et al 2012, Pereira et al 2018, nevertheless, they fail to correctly assess the signals quality in the case of AF because consecutive PPG pulses become morphology different (Asgari et al 2010). Machine learning (ML) techniques can overcome this problem since they are trained with AF cases (Pereira et al 2018). In our previous work, we proposed a two-class SVM-based quality assessment model with a 0.95 accuracy that showed to be more accurate than the approaches reported in the literature that were based on either traditional statistical (Asgari et al 2010, Sukor et al 2011, Li and Clifford 2012, Karlen et al 2012, Orphanidou et al 2015, Liu et al 2016 or other ML methods (Chong et al 2014).…”