Estimasi Kadar Gula Darah Secara Non-invasif Menggunakan Sensor Microwave Berbasis Complementary Split-Ring Resonator (CSRR)

Zakariya, Aisyah Nur Rahmasari (2026) Estimasi Kadar Gula Darah Secara Non-invasif Menggunakan Sensor Microwave Berbasis Complementary Split-Ring Resonator (CSRR). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023221035_Undergraduate_Thesis.pdf Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Diabetes melitus memerlukan pemantauan kadar gula darah secara berkala, namun metode konvensional masih bersifat invasif sehingga menimbulkan rasa sakit dan ketidaknyamanan. Penelitian ini mengembangkan sistem estimasi kadar gula darah secara non-invasif menggunakan sensor microwave berbasis Dual-Parallel Complementary Split-Ring Resonator (DP-CSRR) yang diintegrasikan dengan ADALM-PLUTO Software Defined Radio (SDR) dan memanfaatkan machine learning untuk proses estimasi. Sensor hasil fabrikasi memiliki frekuensi resonansi sebesar 2,58 GHz pada kondisi tanpa beban berdasarkan hasil karakterisasi menggunakan VNA. Setelah dibebani jaringan biologis pada pengukuran subjek manusia, resonansi operasional bergeser ke 2,70 GHz akibat karakteristik dual-notch sensor. Pluto SDR dioperasikan menggunakan metode Stepped-Frequency Continuous-Wave (SFCW) pada rentang frekuensi 2,0-3,0 GHz. Sinyal I/Q hasil akuisisi diolah menjadi kurva respons transmisi (|S21|) melalui transformasi Fast Fourier Transform (FFT) dan incoherent averaging. Selanjutnya, dilakukan evaluasi empat model machine learning, yaitu LightGBM, XGBoost, LASSO, dan Random Forest, untuk mengestimasi kadar gula darah berdasarkan karakteristik resonansi sensor. Hasil penelitian menunjukkan bahwa model LightGBM memberikan performa terbaik dengan Mean Absolute Error (MAE) sebesar 15,67 mg/dL, Root Mean Square Error (RMSE) sebesar 20,37 mg/dL, dan Mean Absolute Relative Difference (MARD) sebesar 14,33%. Selain itu, seluruh hasil estimasi berada pada zona A dan B berdasarkan Clarke Error Grid Analysis (EGA), sehingga masih dapat diterima secara klinis pada rentang kadar gula darah subjek penelitian. Hasil tersebut menunjukkan bahwa sistem yang dikembangkan berhasil membangun pipeline estimasi kadar gula darah secara non-invasif berbasis sensor microwave dan SDR, meskipun masih memerlukan pengembangan lebih lanjut melalui penambahan jumlah subjek dan rentang kadar gula darah yang lebih luas untuk meningkatkan kemampuan generalisasi model.
======================================================================================================================================
Diabetes mellitus requires regular blood glucose monitoring. However, conventional methods remain invasive, causing pain and discomfort. This study develops a non-invasive blood glucose estimation system using a Dual-Parallel Complementary Split-Ring Resonator (DP-CSRR) microwave sensor integrated with an ADALM-PLUTO Software Defined Radio (SDR) and combined with machine learning for blood glucose estimation. The fabricated sensor exhibited a resonant frequency of 2.58 GHz under unloaded conditions during VNA characterization. When loaded with biological tissue during human subject measurements, the operational resonance shifted to approximately 2.70 GHz due to the sensor’s dual-notch characteristics. The Pluto SDR was operated using the Stepped-Frequency Continuous-Wave (SFCW) method over a frequency range of 2.0-3.0 GHz. The acquired I/Q signals were processed into transmission response (|S21|) curves through Fast Fourier Transform (FFT) and incoherent averaging. Subsequently, four machine learning models which LightGBM, XGBoost, LASSO, and Random Forest, were evaluated to estimate blood glucose levels based on the sensor’s resonant characteristics. The results showed that the LightGBM model achieved the best performance, with a Mean Absolute Error (MAE) of 15.67 mg/dL, a Root Mean Square Error (RMSE) of 20,37 mg/dL, and a Mean Absolute Relative Difference (MARD) of 14.33%. Furthermore, all predictions fell within Zones A and B of the Clarke Error Grid Analysis (EGA), indicating clinically acceptable performance within the blood glucose range of the study subjects. These results demonstrate that the proposed system successfully establishes a microwave sensor and SDR pipeline for non-invasive blood glucose estimation, although further studies involving more subjects and a wider blood glucose range are required to improve model generalization.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Gula Darah, Gelombang Mikro, Non-invasif, Complementary Split-Ring Resonator (CSRR), Software Defined Radio : Blood Glucose, Microwave, Non-invasive, Complementary Split-Ring Resonator (CSRR), Software Defined Radio
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.6 Antennas (Electronics) T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7872 Electromagnetic Devices
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Aisyah Nur Rahmasari Zakariya
Date Deposited:	31 Jul 2026 06:02
Last Modified:	31 Jul 2026 06:02
URI:	http://repository.its.ac.id/id/eprint/140526

Actions (login required)

View Item