DETEKSI PENYAKIT MATA KATARAK MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK (CNN)

Tambunan, Ruth Octaviana Elizabeth and Aziza, Rosida Nur and Agtriadi, Herman Bedi (2025) DETEKSI PENYAKIT MATA KATARAK MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK (CNN). Diploma thesis, ITPLN.

[thumbnail of 201931213_RuthTambunan_Revisi_Skripsi 2_RUTH OCTAVIANA ELIZA.pdf] Text
201931213_RuthTambunan_Revisi_Skripsi 2_RUTH OCTAVIANA ELIZA.pdf
Restricted to Registered users only

Download (3MB)

Abstract

Katarak merupakan salah satu gangguan penglihatan yang umum terjadi pada Manusia dan menjadi penyebab utama kebutaan secara global. Seiring dengan perkembangan teknologi digital dan kecerdasan buatan, sistem deteksi dini berbasis citra digital menjadi solusi yang menjanjikan untuk meminimalkan keterlambatan diagnosis. Penelitian ini bertujuan membangun sistem deteksi katarak otomatis menggunakan metode Convolutional Neural Network (CNN) dengan arsitektur MobileNetV1, yang dikenal efisien untuk perangkat dengan keterbatasan sumber daya. Dataset penelitian berjumlah 1.200 citra mata yang terdiri dari dua kategori, yaitu katarak dan normal, diperoleh dari sumber terbuka (Kaggle). Tahapan penelitian meliputi preprocessing (resizing, normalisasi), augmentasi citra, pelatihan model dengan transfer learning, serta evaluasi performa menggunakan metrik akurasi, presisi, recall, dan F1-score. Hasil evaluasi menunjukkan model mencapai akurasi 99%, dengan presisi, recall, dan F1-score masing-masing sebesar 0,99. Pada kelas katarak diperoleh presisi 1,00 dan recall 0,98, sedangkan kelas normal memperoleh presisi 0,98 dan recall 1,00. Selama proses pelatihan, grafik akurasi menunjukkan peningkatan stabil hingga mendekati 100%, sementara nilai loss menurun signifikan tanpa indikasi overfitting. Model kemudian diintegrasikan ke dalam aplikasi berbasis Streamlit Cloud sehingga pengguna dapat mengunggah citra mata dan memperoleh hasil deteksi secara instan. Dengan hasil ini, sistem yang dikembangkan berpotensi menjadi alat bantu skrining awal yang cepat, ringan, dan mudah diakses masyarakat, khususnya pada wilayah dengan keterbatasan layanan kesehatan.

Cataract is one of the most common visual impairments in humans and is a leading cause of blindness worldwide. Along with the advancement of digital technology and artificial intelligence, early detection systems based on digital imaging have become a promising solution to minimize diagnostic delays. This study aims to develop an automatic cataract detection system using the Convolutional Neural Network (CNN) method with the MobileNetV1 architecture, which is known for its efficiency on resource-limited devices. The dataset used consists of 1,200 eye images categorized into cataract and normal, obtained from an open-source platform (Kaggle). The research stages include preprocessing (resizing, normalization), image augmentation, model training with transfer learning, and performance evaluation using accuracy, precision, recall, and F1-score metrics. The evaluation results indicate that the model achieved an accuracy of 99%, with precision, recall, and F1-score values of 0.99. For the cataract class, precision reached 1.00 and recall 0.98, while for the normal class, precision was 0.98 and recall 1.00. During training, the accuracy graph showed a stable increase approaching 100%, while the loss value decreased significantly without signs of overfitting. The final model was integrated into a Streamlit Cloud-based application that allows users to upload eye images and obtain instant detection results. These findings suggest that the proposed system has the potential to serve as an early screening tool that is fast, lightweight, and easily accessible, particularly in areas with limited healthcare services.

Item Type: Thesis (Diploma)
Uncontrolled Keywords: katarak, CNN, MobileNetV1, Citra Digital, Streamlit Cloud, Cataract, CNN, MobileNetV1, digital imaging, Streamlit Cloud
Subjects: Skripsi
Bidang Keilmuan > Teknik Informatika
Divisions: Fakultas Telematika Energi > S1 Teknik Informatika
Depositing User: Sudarman
Date Deposited: 14 Oct 2025 04:19
Last Modified: 14 Oct 2025 04:19
URI: https://repository.itpln.ac.id/id/eprint/2245

Actions (login required)

View Item
View Item