Faradisa, Irmalia Suryani (2023) Klasifikasi Sinyal Fetal Phonocardioram Untuk Kesehatan Jantung Janin Menggunakan Optimized Neural Network. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
Text
0711156000011-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 April 2025. Download (7MB) | Request a copy |
Abstract
Pemantauan kesehataan janin merupakan hal yang penting dilakukan pada masa kehamilan, hal ini berguna untuk bisa melihat perkembangan janin dari waktu ke waktu. Kondisi kesehatan ibu berisiko tinggi sangat memprihatinkan, terutama di negara berkembang. Pemantauan intensif wajib dilakukan untuk mencegah hal tersebut. Namun, metode invasif jangka panjang pada wanita hamil membahayakan bayi dan ibunya. Auskultasi masih merupakan salah satu alat analitis paling dasar yang digunakan untuk menentukan keadaan fungsional jantung janin serta ukuran kesehatan janin. Hal Ini disebut dengan fonokardiografi janin (fPCG) dalam bentuk modernnya. Teknik fPCG bersifat pasif dan dapat digunakan untuk melacak jangka panjang.
Dalam penelitian ini, kami mengusulkan klasifikasi detak jantung janin non- invasif yang hemat biaya berdasarkan fonokardiograf dengan perakitan fitur. Karena tingginya jumlah fitur dan mahal secara komputasi, data balancing dilakukan dengan menggunakan synthetic minority oversampling technique (SMOTE). Sedangkan untuk mengurangi ukuran data set, dilakukan pemotongan ukurannya menjadi setengahnya dengan memanfaatkan Principal Component Analysis (PCA). Untuk meningkatkan kinerja klasifikasi maka digunakan metode neural network yang dioptimalkan menggunakan metode random search optimazition. Berdasarkan hasil pengujian dengan metode yang digunakan tersebut maka diperoleh posisi teratas di semua penyeimbangan data dibandingkan dengan algoritma pembelajaran mesin lainnya, dengan 91,7% untuk akurasi dan Area Under Curve dengan skor 91,6%
=================================================================================================================================
Monitoring the well-being of the fetus is an important thing to do during pregnancy, this is useful to be able to see the development of the fetus from time to time. The health conditions of high-risk mothers are very concerning, especially in developing countries. Intensive monitoring must be carried out to prevent this. However, long-term invasive methods in pregnant women harm both the baby and the mother. Auscultation is still one of the most basic analytical tools used to determine the functional state of the fetal heart as well as a measure of fetal well being. This is called fetal phonocardiography (fPCG) in its modern form. The fPCG technique is passive and can be used for long-term tracking. In this study, we propose a cost-effective non-invasive classification of fetal heartbeats based on phonocardiographs with assembly features. Due to the high number of features and the high cost of processing, data balancing was performed using a synthetic minority oversampling (SMOTE) technique. Meanwhile, to reduce the size of the data set, the size is cut in half by using Principal Component Analysis (PCA). To improve classification performance, the neural network method is used which is optimized using the random search optimization method. Based on the test results with the method used, the top position is obtained in all data balancing compared to other machine learning algorithms, with 91.7% for accuracy and Area Under Curve with a score of 91.6%.
Item Type: | Thesis (Doctoral) |
---|---|
Uncontrolled Keywords: | Aritmia, Fitur Ensemble, Fetal Phonokardiogram, Neural Network, Pengurangan Dimensi,Arrhythmia, Dimensional Reduction, Ensemble Features, Fetal Phonocardiogram, Neural Network |
Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5102.9 Signal processing. |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20001-(S3) PhD Thesis |
Depositing User: | Irmalia Suryani Faradisa |
Date Deposited: | 15 Feb 2023 05:12 |
Last Modified: | 15 Feb 2023 05:12 |
URI: | http://repository.its.ac.id/id/eprint/97230 |
Actions (login required)
View Item |