POST-COVID-19 PANDEMIC MENTAL DISORDERS PREDICTION USING PHOTOPLETHYSMOGRAM (Ramalan Gangguan Mental Selepas Pandemik COVID-19 Menggunakan Fotopletismogram)

Abstract

ABSTRACT

Mental disorders interfere with functioning and affect a person’s quality of life. The COVID-19 pandemic has caused an increase in the number of people who are suffering from mental disorders, including suicide. Many people are unaware of their mental health status because the signs might not be readily apparent. In this study, a machine learning (ML) approach that differentiates between case groups, i.e., patients diagnosed with mental disorders and control groups, was developed using photoplethysmogram (PPG) morphology. The subjects consisted of 92 volunteers, who were divided equally into case and control groups, matched in gender and age. PPG signals were collected individually in a 5-min experiment in a relaxation mode. Out of 13 morphological features of PPG, eight features were extracted using the pulse rate variability method and five features were extracted using the fiducial point detection method. Statistical and correlation analyses have verified eight features as inputs to five types of ML algorithms. The results showed that the kNN model achieved the best performance of 92.86%, 99.10%, and 96.43% for sensitivity, specificity, and accuracy. The value of error in prediction, or mean squared error for kNN, was the lowest at 0.036. In terms of model robustness, the area under the curve of the receiver operating characteristic curve for kNN (0.964) was higher than the other models. A mental disorder model has been developed using Machine Learning (ML) from PPG extraction. In conclusion, mental disorders can be detected by PPG based on the obtained results. This development also respects Islamic ethical values, emphasising the importance of upholding individuals’ privacy and dignity, especially when collecting and analysing sensitive health data like PPG signals.

ABSTRAK
Gangguan mental mengganggu fungsi dan mempengaruhi kualiti hidup seseorang. Pandemik COVID-19 telah menyebabkan peningkatan bilangan individu yang mengalami gangguan mental, termasuk peningkatan kes bunuh diri. Ramai yang tidak sedar dengan status kesihatan mental mereka kerana tanda-tanda mungkin tidak jelas. Dalam kajian ini, pendekatan pembelajaran mesin (ML) yang membezakan antara kumpulan kes, iaitu pesakit yang disahkan mengalami gangguan mental, dan kumpulan kawalan, telah dibangunkan dengan menggunakan morfologi fotopletismogram (PPG). Subjek terdiri daripada 92 peserta, dibahagikan kepada kumpulan kes dan kumpulan kawalan, dengan pertimbangan jantina dan umur yang seimbang. Isyarat PPG dikumpul secara individu dalam eksperimen 5 minit dalam keadaan rehat. Daripada 13 ciri morfologi PPG, lapan ciri diekstrak menggunakan kaedah variasi kadar nadi dan lima ciri diekstrak menggunakan kaedah penentuan titik fidusial. Analisis statistik dan korelasi telah mengesahkan lapan ciri sebagai input kepada lima jenis algoritma ML. Hasil menunjukkan bahawa model k-Nearest neighbour (kNN) mencapai prestasi terbaik pada 92.86%, 99.10%, dan 96.43% untuk kepekaan, kekhususan, dan ketepatan. Nilai ralat dalam ramalan, atau purata ralat kuasa dua bagi kNN, adalah yang terendah pada 0.036. Dari segi kebolehterimaan model, kawasan di bawah lengkung penerimaan ciri untuk kNN (0.964) lebih tinggi berbanding model-model lain. Model ramalan gangguan mental telah dibangunkan menggunakan Pembelajaran Mesin (ML) daripada ekstraksi PPG. Secara kesimpulannya, gangguan mental boleh dikesan dengan menggunakan PPG berdasarkan hasil yang diperoleh. Pembangunan model ramalan ini juga menghormati nilai-nilai etika Islam dengan menekankan kepentingan perlindungan privasi dan kerahsiaan individu, terutamanya semasa mengumpul dan menganalisis data kesihatan yang diperolehi melalui isyarat PPG.