Cermin Cerdas Multimodal Untuk Skrining Kesehatan Otomatis Berbasis Deep Learning

Suryanto, Nehemy Davis (2026) Cermin Cerdas Multimodal Untuk Skrining Kesehatan Otomatis Berbasis Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023221004-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (12MB) | Request a copy

Abstract

Pelayanan kesehatan di Indonesia masih menghadapi berbagai tantangan, seperti keterbatasan tenaga kesehatan, waktu tunggu pasien yang panjang, serta proses pengukuran tanda vital yang umumnya masih dilakukan secara terpisah menggunakan beberapa alat. Penelitian ini bertujuan mengembangkan cermin cerdas multimodal untuk skrining kesehatan otomatis yang mampu mengukur dan mengestimasi delapan parameter kesehatan, yaitu tinggi badan, berat badan, suhu tubuh, denyut jantung, laju pernapasan, saturasi oksigen, tekanan darah, dan kadar gula darah. Sistem dikembangkan dengan mengintegrasikan berbagai sensor biomedis, algoritma pengolahan sinyal ECG dan PPG, metode Signal Quality Assessment (SQA), model deep learning, serta rekam medis elektronik berbasis Internet of Things (IoT). Hasil evaluasi menunjukkan bahwa sistem mampu melakukan pengukuran dengan performa yang baik, ditunjukkan oleh rata-rata error pengukuran tinggi badan sebesar 0,884 cm, Mean Absolute Error (MAE) suhu tubuh sebesar 0,29°C, MAE denyut jantung sebesar 0,93 bpm, MAE laju pernapasan sebesar 1,95 brpm, serta MAE saturasi oksigen sebesar 0,84%. Selain itu, model estimasi kadar gula darah berbasis Hybrid CNN-BiLSTM-Attention menghasilkan MAE sebesar 11,07 mg/dL, Root Mean Squared Error (RMSE) sebesar 14,96 mg/dL, Mean Absolute Relative Difference (MARD) sebesar 10,94%, dan seluruh hasil estimasi berada pada Zona A dan B dalam Clarke Error Grid Analysis. Adapun model estimasi tekanan darah berbasis CNN dengan sinyal ECG dan PPG menghasilkan MAE (7,57/5,25) mmHg (SBP/DBP), RMSE (9,79/6,47) mmHg (SBP/DBP), dan mencapai grade C/B (SBP/DBP) pada standar akurasi British Hypertension Society (BHS). Evaluasi usability menggunakan System Usability Scale (SUS) menghasilkan skor 72,5 yang menunjukkan tingkat kemudahan penggunaan yang baik. Secara keseluruhan, penelitian ini menunjukkan bahwa cermin cerdas yang dikembangkan memiliki potensi sebagai solusi skrining kesehatan awal yang terintegrasi, otomatis, dan praktis untuk mendukung layanan kesehatan preventif.
========================================================================================================================================
Healthcare services in Indonesia continue to face various challenges, including a shortage of healthcare personnel, long patient waiting times, and vital sign measurements that are generally performed separately using multiple devices. This study aims to develop a multimodal smart mirror for automated health screening capable of measuring and estimating eight health parameters: height, weight, body temperature, heart rate, respiratory rate, oxygen saturation, blood pressure, and blood glucose level. The system was developed by integrating multiple biomedical sensors, ECG and PPG signal processing algorithms, a Signal Quality Assessment (SQA) method, deep learning models, and an Internet of Things (IoT)-based electronic medical record system. Evaluation results demonstrated that the system achieved satisfactory measurement performance, with a mean height measurement error of 0.884 cm, a body temperature MAE of 0,29°C, a heart rate MAE of 0,93 bpm, a respiratory rate MAE of 1,95 brpm, and an oxygen saturation MAE of 0,84%. Furthermore, the proposed Hybrid CNN-BiLSTM-Attention model for blood glucose estimation achieved an MAE of 11,07 mg/dL, an RMSE of 14,96 mg/dL, a MARD of 10,94%, and all predictions fell within Zones A and B of the Clarke Error Grid Analysis. In addition, the CNN-based blood pressure estimation model utilizing ECG and PPG signals achieved an MAE of (7,57/5,25) mmHg (SBP/DBP), an RMSE of (9,79/6,47) mmHg (SBP/DBP) and attained Grade C/B (SBP/DBP) according to the British Hypertension Society (BHS) accuracy standard. Usability evaluation using the System Usability Scale (SUS) yielded a score of 72,5, indicating good ease of use and user acceptance. Overall, the results demonstrate that the developed smart mirror has strong potential as an integrated, automated, and practical solution for early health screening to support preventive healthcare services.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Health Screening, Hybrid Deep Learning, Multisensor, Multi-Vital Signs, Smart Mirror, Cermin Cerdas, Hybrid Deep Learning, Multisensor, Multitanda Vital, Skrining Kesehatan
Subjects:	R Medicine > R Medicine (General) > R856.2 Medical instruments and apparatus. R Medicine > R Medicine (General) > R858 Deep Learning R Medicine > RC Internal medicine > RC683.5.E5 Electrocardiography T Technology > T Technology (General) > T58.62 Decision support systems T Technology > T Technology (General) > T59.7 Human-machine systems. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5102.9 Signal processing.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Nehemy Davis Suryanto
Date Deposited:	31 Jul 2026 00:18
Last Modified:	31 Jul 2026 00:18
URI:	http://repository.its.ac.id/id/eprint/140224

Actions (login required)

View Item