Sistem Deteksi Anemia melalui Citra RAW Digital Konjungtiva dengan Metode Color Correction berbasis Smartphone

Qalbi, Ahmad Dzakwan Dhiya' (2026) Sistem Deteksi Anemia melalui Citra RAW Digital Konjungtiva dengan Metode Color Correction berbasis Smartphone. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023221039-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Golden standard untuk melakukan diagnosis anemia masih bergantung pada pemeriksaan darah yang bersifat invasif, mahal, dan memerlukan fasilitas laboratorium. Sistem deteksi non-invasif melalui citra konjungtiva berbasis smartphone telah dikembangkan sebagai alternatif, namun performanya masih sangat terbatas. Tantangan utama dari sistem ini adalah ketergantungan pada kondisi pencahayaan terkontrol dan variasi karakteristik tiap-tiap perangkat kamera yang menyebabkan inkonsistensi warna citra, sehingga menurunkan akurasi prediksi kadar Hemoglobin (Hb). Penelitian ini bertujuan untuk mengembangkan sistem deteksi anemia non-invasif berbasis citra konjungtiva smartphone yang akurat terhadap variasi pencahayaan dan perangkat. Untuk menjaga konsistensi data, sistem menggunakan citra dalam format RAW (.dng) dan mengimplementasikan algoritma color correction menggunakan color checker card. Proses segmentasi konjungtiva sebagai ROI dilakukan secara otomatis menggunakan arsitektur U-Net Based Conjunctiva Segmentation Model (UNBCSM) dengan pendekatan Transfer Learning. Hasil penelitian menunjukkan bahwa tahap prapemrosesan berhasil menekan rata-rata eror warna (∆E) secara drastis dari 16,30 menjadi 3,27. Pada tahap segmentasi, model UNBCSM berhasil menghasilkan Mean Intersection over Union (IoU) sebesar 74,9% pada data Testing. Untuk prediksi kadar Hemoglobin, dari evaluasi komparasi disimpulkan bahwa algoritma Random Forest mencatatkan performa paling optimal, menghasilkan tingkat eror RMSE sebesar 1,31 g/dL, MAE sebesar 1,0 g/dL, dan nilai determinasi R2 sebesar 0,482.
================================================================================================================================
The gold standard for diagnosing anemia still relies on blood tests, which are invasive, expensive, and require laboratory facilities. Non-invasive detection systems using smartphone-based conjunctival images have been developed as an alternative, but their performance remains highly limited. The main challenge of these systems is their reliance on controlled lighting conditions and the variations in camera device characteristics, which cause image color inconsistencies, thereby reducing the accuracy of Hemoglobin (Hb) level prediction. This study aims to develop a non-invasive anemia detection system based on smartphone conjunctival images that is accurate across variations in lighting and devices. To maintain data consistency, the system utilizes images in RAW format (.dng) and implements a color correction algorithm using a color checker card. The segmentation process of the conjunctiva as the Region of Interest (ROI) is performed automatically using the U-Net Based Conjunctiva Segmentation Model (UNBCSM) architecture with a Transfer Learning approach. The results indicate that the preprocessing stage successfully reduced the average color error (ΔE) significantly from 16.30 to 3.27. In the segmentation stage, the UNBCSM model successfully achieved a Mean Intersection over Union (IoU) of 74.9% on the testing data. For hemoglobin level prediction, the comparative evaluation concluded that the Random Forest algorithm recorded the most optimal performance after optimization, yielding an RMSE error rate of 1.31 g/dL, a MAE of 1.0 g/dL, and an R2 determination coefficient of 0.482.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Anemia, Deteksi Non-Invasif, Konjungtiva, Smartphone, Color Correction; Anemia, Non-Invasive Detection, Conjunctiva, Smartphone, Color Correction.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. R Medicine > R Medicine (General) > R858 Deep Learning T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Ahmad Dzakwan Dhiya' Qalbi
Date Deposited:	01 Aug 2026 03:05
Last Modified:	01 Aug 2026 03:05
URI:	http://repository.its.ac.id/id/eprint/141556

Actions (login required)

View Item