Key result
An artificial neural network using photoplethysmogram morphology features accurately predicted psychiatric disorders, achieving 96.43% accuracy, 92.86% sensitivity, and 100% specificity.
Why the study?
Prompt diagnosis and intervention for psychiatric disorders are important, but symptoms are often not easily recognized, leading to complications later in life.
Does a machine learning approach using photoplethysmogram morphology accurately predict psychiatric disorders?
Case-Control (n=92)
Does a machine learning approach using photoplethysmogram morphology accurately predict psychiatric disorders?
Effect estimate: 96.43% accuracy
Machine learning models using photoplethysmogram morphology can accurately distinguish patients with psychiatric disorders from healthy controls.
PPG-based neural networks may aid psychiatric screening research; leaves open prospective validation before clinical consideration.
Psychiatric disorders (PDs) interfere with one's functioning and greatly affect a person's quality of life. Prompt diagnosis and intervention at the early stages of these illnesses are important. However, most people are oblivious or unaware of their mental health status as the symptoms may not be easily recognizable. Consequently, complications occur later in life. In this study, a machine learning (ML) approach that distinguishes between case (PD-diagnosed patients) and control (healthy) groups was developed using photoplethysmogram (PPG) morphology. 92 subjects with gender and age-matched PPG data were collected during two phases; baseline and stimulus state of a 10-min experiment. 60 features from PPG morphology were extracted from each phase, and another 30 were obtained from differences between the two phases. A total of 27 out of 90 features exhibited a significant difference. Twelve features extracted by heatmap based on the correlation analysis were fed to five types of ML algorithms: discrimination analysis, k-nearest neighbor, decision tree, support vector machine, and artificial neural network (ANN). The results showed the best performance of 92.86%, 100.00%, and 96.43% for sensitivity, specificity, and accuracy by ANN. Thus, a PD prediction model was developed using machine learning techniques from PPG morphology extraction.
No takes yet. Share an insight, caveat, or question.
Awang et al. (2022) conducted a case-control in Psychiatric disorders (n=92). Machine learning using photoplethysmogram morphology vs. Healthy controls was evaluated on Prediction of psychiatric disorders (96.43% accuracy). An artificial neural network using photoplethysmogram morphology features accurately predicted psychiatric disorders, achieving 96.43% accuracy, 92.86% sensitivity, and 100% specificity.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: