Ramadhani, Andi Lisnaini (2026) Sistem Deteksi Gangguan Kesehatan Mental Berbasis EEG dan PPG dengan Visualisasi Topographic Brain Mapping. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5023221022-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5023221022-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (9MB) | Request a copy |
Abstract
Gangguan kesehatan mental seperti stres, kecemasan, dan depresi masih menjadi tantangan karena keterbatasan akses layanan profesional serta rendahnya kesadaran individu terhadap gejala. Deteksi dini menjadi penting karena perubahan kondisi mental dapat berkaitan dengan perubahan regulasi neurofisiologis dan kardiovaskular sebelum berkembang lebih berat dan menurunkan kualitas hidup. Sinyal fisiologis seperti Electroencephalography (EEG) dan Photoplethysmography (PPG) berpotensi mendukung skrining awal melalui pemantauan aktivitas otak dan dinamika denyut jantung. Namun, sistem EEG umumnya membutuhkan banyak elektroda, sedangkan multimodal berbasis ECG menambah kompleksitas pemasangan sensor. Untuk menjawab kebutuhan tersebut, penelitian ini mengembangkan sistem skrining gangguan kesehatan mental berbasis 4 channel EEG dan PPG dengan visualisasi topographic brain mapping setara 8 channel melalui Deep Learning EEG Super-Resolution. Sistem menggunakan OpenBCI Cyton, Raspberry Pi 5, elektroda EEG kering, sensor PPG earlobe, dan display. Sinyal EEG dan PPG diakuisisi secara sinkron, kemudian diproses melalui MAICA-FastICA, EEG Super-Resolution, ekstraksi dan seleksi fitur, serta klasifikasi menggunakan Support Vector Machine (SVM). Label ditentukan berdasarkan DASS-42 dan diklasifikasikan secara biner menjadi normal dan tidak normal pada dimensi stres, kecemasan, dan depresi. Hasil pengujian menunjukkan bahwa fitur PRV dari PPG memiliki kesesuaian yang baik terhadap HRV dengan rata-rata percentage error 3,22% dan korelasi rata-rata 0,9872. Model klasifikasi memperoleh akurasi 0,962 untuk stres, 0,962 untuk kecemasan, dan 0,886 untuk depresi. EEG Super-Resolution menghasilkan channel virtual yang lebih baik dibandingkan interpolasi spline, dengan penurunan klasifikasi sebesar 0,013 pada kecemasan dan 0,089 untuk depresi. Dengan demikian, sistem ini berhasil menjadi dasar pengembangan skrining awal berbasis EEG dan PPG yang lebih ringkas, minim elektroda, dan dilengkapi interpretasi visual melalui topographic brain mapping.
=======================================================================================================================================
Mental health disorders such as stress, anxiety, and depression remain major challenges due to limited access to professional mental health services and low individual awareness of symptoms. Early detection is important because changes in mental conditions may be associated with alterations in neurophysiological and cardiovascular regulation before progressing to more severe conditions and negatively affecting quality of life. Physiological signals such as Electroencephalography (EEG) and Photoplethysmography (PPG) have the potential to support early screening by monitoring brain activity and heart rate dynamics. However, EEG-based systems generally require multiple electrodes, while multimodal approaches using ECG increase sensor placement complexity. To address this need, this study developed a mental health screening system based on 4-channel EEG and PPG with 8-channel-equivalent topographic brain mapping visualization through a Deep Learning-based EEG Super-Resolution approach. The system was implemented using OpenBCI Cyton, Raspberry Pi 5, dry EEG electrodes, an earlobe PPG sensor, and a display interface. EEG and PPG signals were acquired synchronously and processed through MAICA-FastICA, EEG Super-Resolution, feature extraction and selection, and classification using Support Vector Machine (SVM). The labels were determined based on DASS-42 and classified into binary normal and abnormal classes for stress, anxiety, and depression. The results showed that PRV features derived from PPG had good agreement with HRV, with an average percentage error of 3.22% and an average correlation of 0.9872. The classification models achieved accuracies of 0.962 for stress, 0.962 for anxiety, and 0.886 for depression. EEG Super-Resolution produced virtual channels that outperformed spline interpolation, with classification performance decreases of 0.013 for anxiety and 0.089 for depression. Overall, the proposed system serves as a foundation for a compact early screening framework based on EEG and PPG, requiring fewer electrodes while providing visual interpretability through topographic brain mapping.
Actions (login required)
 |
View Item |