Pembuatan Data Sintetik Pada Data Histori Pasien Menggunakan Metode Berbasis GAN Untuk Deteksi Risiko Komplikasi Terapi CAPD

Baihaqi, Muhammad Naufal (2025) Pembuatan Data Sintetik Pada Data Histori Pasien Menggunakan Metode Berbasis GAN Untuk Deteksi Risiko Komplikasi Terapi CAPD. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025211103-Muhammad Naufal Baihaqi-BukuTA.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Continuous Ambulatory Peritoneal Dialysis (CAPD) adalah alternatif terapi pengganti ginjal yang fleksibel untuk pasien penyakit ginjal tahap akhir, namun memiliki risiko tinggi terhadap komplikasi infeksi. Penelitian ini mengembangkan model klasifikasi multimodal untuk deteksi risiko komplikasi pada pasien CAPD. Pendekatan ini mengatasi masalah ketidakseimbangan dataset dan keterbatasan data melalui synthetic data generation berbasis Generative Adversarial Network (GAN). Model multimodal ini menggabungkan data citra effluent dialysate yang diekstraksi menggunakan model pre-trained (Swin Transformer, CaiT, CoAtNet, YOLOv9) dengan data klinis pasien menggunakan metode early fusion. Untuk mengatasi ketidakseimbangan data, empat model synthetic data generation berbasis GAN, yaitu CTGAN, CTABGAN, CTABGAN+, dan CopulaGAN, diterapkan untuk menghasilkan data sintetik. Sembilan algoritma machine learning digunakan untuk proses klasifikasi, yaitu logistic regression, k-nearest neighbors (KNN), support vector machine (SVM), decision tree, random forest, AdaBoost, eXtreme Gradient Boosting (XGBoost), naïve Bayes (NB), dan Extra Trees. Hasil terbaik terdapat pada model machine learning Extra Trees. Model Extra Trees ini mencapai recall 0,9355, precision 1,0000, dan F1-score 0,9667 pada skenario penggunaan PCA 85% variance, CaiT sebagai feature extractor, standard scaler, dan CTABGAN sebagai model pembuat data sintetik. Pengembangan model deteksi ini berfokus pada meminimalkan false negative dan mengoptimalkan metrik F1-score serta precision, mengingat pentingnya deteksi kasus abnormal secara akurat dalam konteks medis. Hasil penelitian ini akan diintegrasikan ke dalam aplikasi SahabatCAPD sebagai sistem pemantauan mandiri pasien selama menjalani terapi CAPD untuk mendukung keputusan medis yang lebih cepat dan tepat dalam perawatan pasien CAPD.
===================================================================================================================================
Continuous Ambulatory Peritoneal Dialysis (CAPD) is a flexible alternative renal replacement therapy for patients with end-stage renal disease, but has a high risk of infectious complications. This study develops a multimodal classification model for complication risk detection in CAPD patients. This approach overcomes the problem of dataset imbalance and data limitation through synthetic data generation based on Generative Adversarial Network (GAN). The multimodal model combines effluent dialysate image data extracted using pretrained models (Swin Transformer, CaiT, CoAtNet, YOLOv9) with patient clinical data using early fusion method. To overcome data imbalance, four GAN-based synthetic data generation models, namely CTGAN, CTABGAN, CTABGAN+, and CopulaGAN, were applied to generate synthetic data. Nine machine learning algorithms were used for the classification process, namely logistic regression, k-nearest neighbors (KNN), support vector machine (SVM), decision tree, random forest, AdaBoost, eXtreme Gradient Boosting (XGBoost), naïve Bayes (NB), and Extra Trees. The best results are found in the Extra Trees machine learning model. This Extra Trees model achieved recall 0.9355, precision 1.0000, and F1-score 0.9667 in the scenario of using PCA 85% variance, CaiT as feature extractor, standard scaler, and CTABGAN as synthetic data generation model. The development of this detection model focuses on minimizing false negatives and optimizing F1-score and precision metrics, given the importance of accurate detection of abnormal cases in a medical context. The results of this study will be integrated into the SahabatCAPD application as a patient self-monitoring system during CAPD therapy to support faster and more informed medical decisions in CAPD patient care.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	CAPD, Klassifikasi Multimodal, Data Sintetik, GAN, Effluent Dialysate, CAPD, Multimodal Classification, Synthetic Data, GAN, Effluent Dialysate
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models. T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Naufal Baihaqi
Date Deposited:	21 Jul 2025 03:19
Last Modified:	21 Jul 2025 03:19
URI:	http://repository.its.ac.id/id/eprint/120225

Actions (login required)

View Item