Deteksi Anemia Defisiensi Besi Pada Citra Sel Darah Merah Menggunakan Mikroskop Digital Berbasis Machine Learning

Prasetyo, Yohanes Eka Adi (2026) Deteksi Anemia Defisiensi Besi Pada Citra Sel Darah Merah Menggunakan Mikroskop Digital Berbasis Machine Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023221016-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Anemia Defisiensi Besi (IDA) merupakan masalah kesehatan global yang signifikan, khususnya pada kelompok rentan seperti ibu hamil dan anak-anak, yang erat kaitannya dengan risiko stunting dan kematian ibu. Metode diagnosis konvensional saat ini, yaitu pemeriksaan Complete Blood Count (CBC) dan analisis manual apusan darah tepi, memiliki keterbatasan utama berupa subjektivitas pengamatan, waktu pemrosesan yang lama, biaya tinggi, serta ketergantungan pada ketersediaan ahli klinis yang minim di fasilitas kesehatan dengan sumber daya terbatas. Solusi berbasis Deep Learning yang ada saat ini cenderung membutuhkan komputasi berat dan perangkat mahal yang sulit diakses. Untuk mengatasi kesenjangan tersebut, penelitian ini merancang dan mengimplementasikan sistem mikroskop digital portabel berbasis Raspberry Pi 5 dengan modul kamera Raspberry Pi v3 untuk akuisisi citra apusan darah tepi. Dari sisi hardware, mikroskop dirancang dengan desain multilens fixed tube 16 cm untuk mencapai target perbesaran hingga 1000x, mekanisme meja preparat naik turun berbasis stepper motor, serta sensor VL53L0X untuk estimasi perbesaran. Pada sisi software, alur mencakup preprocessing, eliminasi sel tidak relevan menggunakan K-means clustering pada ruang warna CIELAB, serta segmentasi dan pemisahan sel yang tumpang tindih menggunakan metode Bounded Opening-Fast Radial Symmetry (BO-FRS) dikombinasikan dengan Pixel Replication dan Gaussian Mixture Model (GMM). Sebanyak 40 fitur kuantitatif yang mencakup morfologi, central pallor, GLCM, dan momen warna diekstraksi dari setiap sel, kemudian diseleksi menggunakan Mutual Information dan dilatih pada model klasifikasi Support Vector Machine (SVM) dengan kernel Radial Basis Function (RBF) yang dioptimalkan melalui GridSearchCV dan memberikan performa dengan akurasi testing sebesar 88,23%, presisi 88,32%, recall 88,50%, spesifisitas 90,64%, dan F1-Score 88,22%. Hasil ini menunjukkan bahwa sistem yang dikembangkan mampu menjadi perangkat skrining awal yang efisien khususnya di daerah terpencil dan fasilitas kesehatan dengan sumber daya terbatas.
======================================================================================================================================
Iron Deficiency Anemia (IDA) is a significant global health issue, particularly among vulnerable groups such as pregnant women and children, and is closely linked to the risk of stunting and maternal mortality. Current conventional diagnostic methods namely, Complete Blood Count (CBC) tests and manual analysis of peripheral blood smears have major limitations, including subjectivity in observation, long processing times, high costs, and dependence on the limited availability of clinical experts in healthcare facilities with scarce resources. Existing Deep Learning based solutions tend to require heavy computing power and expensive devices that are difficult to access. To address these gaps, this study designed and implemented a portable, digital microscope system based on the Raspberry Pi 5 with a Raspberry Pi v3 camera module for acquiring peripheral blood smear images. On the hardware side, the microscope features a 16-cm fixed-tube multilens design to achieve a target magnification of up to 1000x, a stepper-motor-driven up-and-down stage mechanism, and a VL53L0X sensor for magnification estimation. On the software side, the workflow includes preprocessing, elimination of irrelevant cells using K-means clustering in CIELAB color space, and segmentation and separation of overlapping cells using the Bounded Opening-Fast Radial Symmetry (BO-FRS) method combined with the pixel replication and Gaussian Mixture Model (GMM). A total of 40 quantitative features including morphology, central pallor, GLCM, and color moments were extracted from each cell, then selected using Mutual Information and trained on a Support Vector Machine (SVM) classification model with a Radial Basis Function (RBF) kernel optimized via GridSearchCV, reached a testing accuracy of 88.23%, precision of 88.32%, recall of 88.50%, specificity of 90.64%, and an F1-score of 88.22%. These results indicate that the developed system has the potential to serve as an efficient particularly in remote areas and healthcare facilities with limited resources.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Anemia Defisiensi Besi, Mikroskop Digital, Raspberry Pi, Pengolahan Citra, Machine Learning, BO-FRS, Segmentasi Sel Darah, Iron Deficiency Anemia, Digital Microscope, Raspberry Pi, Machine Learning, Image Processing, BO-FRS, Blood Cell Segmentation
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. R Medicine > RB Pathology 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:	Yohanes Eka Adi Prasetyo
Date Deposited:	31 Jul 2026 02:18
Last Modified:	31 Jul 2026 02:18
URI:	http://repository.its.ac.id/id/eprint/140303

Actions (login required)

View Item