Pengembangan Sistem Informasi untuk Klasifikasi Tingkat Depresi berdasarkan Sinyal EEG

Sitanggang, Richard Halomoan (2026) Pengembangan Sistem Informasi untuk Klasifikasi Tingkat Depresi berdasarkan Sinyal EEG. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221117-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy

Abstract

Depresi merupakan salah satu gangguan mental yang umum terjadi dan dapat berdampak besar pada kualitas hidup penderitanya. Oleh karena itu, diperlukan upaya deteksi dan pemantauan dini agar penderita depresi dapat memperoleh penanganan tepat waktu. Saat ini diagnosis tingkat depresi umumnya dilakukan melalui kuesioner atau wawancara klinis. Namun pendekatan tersebut bersifat subjektif karena bergantung pada persepsi pasien dan interpretasi tenaga medis. Untuk mendukung proses diagnosis yang lebih akurat dan objektif, analisis aktivitas listrik otak menggunakan Electroencephalography (EEG) dapat digunakan sebagai pendekatan alternatif. Tugas Akhir ini bertujuan mengembangkan model machine learning untuk klasifikasi tingkat depresi berbasis sinyal EEG serta sistem informasi berbasis web yang mengintegrasikan proses input data EEG, pemrosesan sinyal, dan penyajian hasil klasifikasi kepada tenaga medis. Data yang digunakan berupa rekaman EEG dari 24 subjek pada dua channel prefrontal (Fp1 dan Fp2) dengan tujuh stimulus. Klasifikasi dilakukan berdasarkan skor Beck Depression Inventory (BDI) menggunakan dua skema klasifikasi, yaitu tiga kelas (Normal, Ringan, dan Sedang) serta dua kelas (Normal dan Depresi). Fitur diekstraksi dari lima domain meliputi time domain (Hjorth dan statistik), frekuensi (relative band power), asimetri (frontal alpha asymmetry), nonlinier (Higuchi fractal dimension), dan konektivitas (coherence). Tujuh algoritma machine learning dievaluasi, yaitu Random Forest, LightGBM, CatBoost, SVM RBF, SVM Linear, Linear Discriminant Analysis (LDA), dan Logistic Regression, menggunakan Stratified Group K-Fold Cross-Validation (K=20 fold) untuk menguji sepuluh kombinasi fitur pada kedua skema klasifikasi. Hasil menunjukkan bahwa skema klasifikasi dua kelas dengan kombinasi fitur K-10 (49 fitur berbasis Hjorth) menghasilkan performa terbaik. Setelah optimasi hyperparameter menggunakan Grid Search, SVM RBF dipilih sebagai model final dengan rata-rata F1-Score 0,742, accuracy 0,740, precision 0,761, recall 0,740, dan AUC 0,759 lintas seluruh stimulus. Stimulus open eyes menghasilkan performa terbaik di seluruh metrik dengan F1-Score 0,865, accuracy 0,867, precision 0,865, recall 0,867, dan AUC 0,885. Sistem informasi dikembangkan menggunakan framework Laravel sebagai layanan web utama dan FastAPI sebagai layanan model yang dihubungkan melalui REST API, serta diimplementasikan pada lingkungan cloud. Hasil pengujian menggunakan metode black-box testing menunjukkan bahwa seluruh fungsi sistem berjalan sesuai dengan kebutuhan yang telah ditetapkan.
==================================================================================================================================================
Depression is a common mental disorder that can significantly affect an individual’s quality of life. Therefore, early detection and monitoring are essential to ensure that patients receive timely treatment. Currently, depression severity diagnosis is commonly conducted through questionnaires or clinical interviews. However, this approach is subjective as it relies on the patient's perception and the clinician's interpretation. To support a more accurate and objective diagnostic process, analysis of brain electrical activity using Electroencephalography (EEG) can be used as an alternative approach. This Final Project aims to develop a machine learning model for EEG-based depression classification and a web-based information system that integrates EEG data input, signal processing, and classification result visualization for healthcare professionals. The data used consists of EEG recordings from 24 subjects on two prefrontal channels (Fp1 and Fp2) with seven stimuli. Classification was performed based on Beck Depression Inventory (BDI) scores using two classification schemes, namely three classes (Normal, Mild, and Moderate) and two classes (Normal and Depression). Features were extracted from five domains including time domain (Hjorth and statistical), frequency (relative band power), asymmetry (frontal alpha asymmetry), nonlinear (Higuchi fractal dimension), and connectivity (coherence). Seven machine learning algorithms were evaluated, namely Random Forest, LightGBM, CatBoost, SVM RBF, SVM Linear, Linear Discriminant Analysis (LDA), and Logistic Regression, using Stratified Group K-Fold Cross-Validation (K=20 folds) to test ten feature combinations on both classification schemes. Results show that the two-class classification scheme with feature combination K-10 (49 Hjorth-based features) yielded the best performance. After hyperparameter optimization using Grid Search, SVM RBF was selected as the final model with an average F1-Score of 0.742, accuracy of 0.740, precision of 0.761, recall of 0.740, and AUC of 0.759 across all stimuli. The open eyes stimulus achieved the best performance across all metrics with an F1-Score of 0.865, accuracy of 0.867, precision of 0.865, recall of 0.867, and AUC of 0.885. The information system was developed using the Laravel framework as the main web service and FastAPI as the model service, connected through a REST API, and deployed in a cloud environment. Black-box testing results indicate that all system functions operate in accordance with the defined requirements.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Depresi, Electroencephalography, Machine learning, Sistem Informasi, Depression, Electroencephalography, Information System, Machine learning
Subjects:	R Medicine > RC Internal medicine > RC386.5 Electroencephalography. T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Richard Halomoan Sitanggang
Date Deposited:	25 Jul 2026 04:59
Last Modified:	25 Jul 2026 04:59
URI:	http://repository.its.ac.id/id/eprint/138330

Actions (login required)

View Item