Deteksi Sinyal EEG Penderita Epilepsi Menggunakan Continuous Wavelet Transform (CWT) dan Deep Learning

Zaef, Rieke Syochrani (2025) Deteksi Sinyal EEG Penderita Epilepsi Menggunakan Continuous Wavelet Transform (CWT) dan Deep Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022221041-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (3MB) | Request a copy

Abstract

Epilepsi, gangguan neorologis yang memengaruhi kualitas hidup penderitanya, menjadi penyebab morbiditas neurologis kedua tertinggi menurut World Health Organization (WHO). Kehadiran Interictal Epileptiform Discharge (IED) pada routine EEG (rEEG) pasien anak dengan kejang absans adalah tantangan diagnostik dalam penilaian risiko epilepsi di masa mendatang.
Penelitian ini menerapkan Deep Learning dengan Transfer Learning menggunakan arsitektur ResNet50 untuk mengatasi tantangan diagnostik epilepsi berdasarkan data rEEG. Fitur scalogram diekstraksi dari sinyal Electroencephalography (EEG) melalui Continuous Wavelet Transform (CWT), sebagai input ke dalam Deep Learning. Model pempelajari pola spasial aktivitas listrik epilepsi menggunakan data Ambulatory EEG (aEEG) dari 23 pasien epilepsi. Data diperoleh dari open source dataset Children Hospital Boston – Massachusset Institute of Technology (CHB-MIT), yang terdiri dari 9 hingga 42 sesi perekaman per pasien dan mencakup beberapa kejang dengan status epilepticus. Akurasi diagnostik model dievaluasi menggunakan data satu kali perekaman rEEG sejumlah 20 pasien dengan status absans di RSUP. Dr. M. Djamil Padang.
Model mencapai kinerja terbaik dengan konfigurasi learning rate 0.1, batch size 8, epoch 30, dan optimizer SGD. Pengujian pada dataset CHB-MIT, diperoleh akurasi pengujian sebesar 97%. Matrik konfusi menunjukkan false positive rate (FPR) 4,41%, false negative rate (FNR) 1,24%, precision 95,7%, recall 98,5%, dan F1-score 97,5%. Pengujian pada dataset RSUP Dr. M. Djamil, diperoleh akurasi pengujian sebesar 95.35%. Matrik konfusi menunjukkan FPR 5,63%, FNR 3,68%, precision 94,5%, recall 96,3% dan F1-score 95,5%.

====================================================================================================================================

Epilepsy, a neurological disorder that affects the quality of life of sufferers, is the second leading cause of neurological morbidity according to the World Health Organization (WHO). The presence of Interictal Epileptiform Discharges (IEDs) on routine EEG (rEEG) in pediatric patients with absence seizures poses a diagnostic challenge in assessing future epilepsy risk.
This research applies Deep Learning with Transfer Learning using the ResNet50 architecture to address the diagnostic challenges of epilepsy based on rEEG data. Scalogram features are extracted from Electroencephalography (EEG) signals via Continuous Wavelet Transform (CWT), serving as input into the Deep Learning model. The model learns spatial patterns of epileptic electrical activity using ambulatory EEG (aEEG) data from 23 epilepsy patients. The data was obtained from the open-source Children's Hospital Boston – Massachusetts Institute of Technology (CHB-MIT) dataset, which consists of 9 to 42 recording sessions per patient and includes multiple seizures with status epilepticus. The diagnostic accuracy of the model was evaluated using a single rEEG recording from 20 patients with a status of FUS at RSUP. Dr. M. Djamil Padang.
The model achieved the best performance with a learning rate configuration of 0.1, a batch size of 8, 30 epochs, and an SGD optimizer. Testing on the CHB-MIT dataset yielded a testing accuracy of 97%. The confusion matrix showed a false positive rate (FPR) of 4.41%, a false negative rate (FNR) of 1.24%, a precision of 95.7%, a recall of 98.5%, and an F1-score of 97.5%. Testing on the RSUP Dr. M. Djamil data resulted in a testing accuracy of 95.35%. The confusion matrix showed an FPR of 5.63%, an FNR of 3.68%, a precision of 94.5%, a recall of 96.3%, and an F1-score of 95.5%.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	ilepsi, routine EEG (rEEG), Kejang absans, Continuous Wavelet Transform (CWT), Deep Learning. Epilepsy, routine EEG (rEEG), Absance Seizure, Continuous Wavelet Transform (CWT), Deep Learning.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Rieke Syochrani Zaef
Date Deposited:	03 Feb 2025 01:28
Last Modified:	03 Feb 2025 01:28
URI:	http://repository.its.ac.id/id/eprint/117686

Actions (login required)

View Item