Winardi, Ratna Prastiwi Putri (2025) Monitoring Kadar Gula Darah Untuk Deteksi Diabetes Mellitus Menggunakan Sinyal Photoplethysmography. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5023211016-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5023211016-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy |
Abstract
Diabetes melitus gestasional (DMG) merupakan gangguan intoleransi karbohidrat yang terjadi selama kehamilan dan menyebabkan peningkatan kadar glukosa darah. Kondisi ini dapat menimbulkan dampak serius, baik bagi ibu maupun bayi, seperti komplikasi persalinan, makrosomia, serta peningkatan risiko diabetes tipe 2 di kemudian hari. Deteksi dan intervensi dini sebelum minggu ke-15 kehamilan menjadi kunci penting dalam mencegah komplikasi jangka panjang. Untuk mendukung hal tersebut, penelitian ini merancang sistem pemantauan kadar gula darah secara non-invasif menggunakan sensor photoplethysmography (PPG) dua LED yang terdiri dari LED infrared dan LED hijau, yang terhubung dengan modul ESP32. Sinyal PPG yang diperoleh dikirim secara real-time ke komputer server untuk dilakukan pemrosesan. Ekstraksi fitur dilakukan dengan mengambil nilai statistik seperti rata-rata dan simpangan baku amplitudo pada titik-titik penting gelombang PPG (systolic peak, dicrotic notch, dan diastolic peak). Selanjutnya, proses seleksi fitur dilakukan menggunakan beberapa metode yaitu Pearson Correlation, Recursive Feature Elimination (RFE), Lasso Regression, dan SelectKBest. Hasil seleksi fitur digunakan sebagai input pada model machine learning seperti Linear Regression, Random Forest, XGBoost, dan ElasticNet. Model ElasticNet dengan seleksi fitur RFE menunjukkan performa terbaik dengan nilai MAE 14.29 mg/dL. Validasi prediksi juga menunjukkan bahwa model lebih akurat memprediksi kadar glukosa pada rentang normal, namun masih memiliki tantangan dalam memprediksi data ekstrem akibat ketidakseimbangan distribusi data. Hasil prediksi kadar gula darah ditampilkan melalui aplikasi smartphone yang dibangun dengan MIT App Inventor dan dilengkapi fitur penyimpanan riwayat data.
=======================================================================================================================================
Gestational diabetes mellitus (GDM) is a carbohydrate intolerance disorder that occurs during pregnancy, leading to elevated blood glucose levels. This condition can have serious consequences for both the mother and the baby, including complications during childbirth, macrosomia, and an increased risk of developing type 2 diabetes later in life. Early detection and intervention before the 15th week of pregnancy are crucial to preventing long-term complications. To support this goal, this study developed a non-invasive blood glucose monitoring system using a dual-LED photoplethysmography (PPG) sensor, consisting of an infrared LED and a green LED, connected to an ESP32 module. The acquired PPG signals are transmitted in real time to a server computer for processing. Feature extraction is performed by calculating statistical values such as the mean and standard deviation of the amplitude at key points of the PPG waveform (systolic peak, dicrotic notch, and diastolic peak). Feature selection is carried out using several methods, including Pearson Correlation, Recursive Feature Elimination (RFE), Lasso Regression, and SelectKBest. The selected features are then used as input for machine learning models such as Linear Regression, Random Forest, XGBoost, and ElasticNet. Among these, the ElasticNet model combined with RFE feature selection achieved the best performance, with a mean absolute error (MAE) of 14.29 mg/dL. Validation results also indicate that the model predicts blood glucose levels more accurately within the normal range but still faces challenges in predicting extreme values due to imbalanced data distribution. The predicted blood glucose results are displayed through a smartphone application built using MIT App Inventor, which also includes a data history storage feature.
Actions (login required)
 |
View Item |