Search for collections on Undip Repository

IDENTIFIKASI OSTEOARTRITIS PINGGUL MENGGUNAKAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK

MUTTAQIN, Faisal and Purwanto, Purwanto and Farikhin, Farikhin (2025) IDENTIFIKASI OSTEOARTRITIS PINGGUL MENGGUNAKAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK. Doctoral thesis, UNIVERSITAS DIPONEGORO.

[thumbnail of 1. Cover.pdf] Text
1. Cover.pdf

Download (132kB)
[thumbnail of Cover lengkap.pdf] Text
Cover lengkap.pdf
Restricted to Repository staff only

Download (3MB)
[thumbnail of 15. BAB 1.pdf] Text
15. BAB 1.pdf

Download (406kB)
[thumbnail of 16. BAB 2.pdf] Text
16. BAB 2.pdf

Download (1MB)
[thumbnail of 17. BAB 3.pdf] Text
17. BAB 3.pdf
Restricted to Repository staff only

Download (599kB)
[thumbnail of 18. BAB 4.pdf] Text
18. BAB 4.pdf
Restricted to Repository staff only

Download (1MB)
[thumbnail of 19. BAB 5.pdf] Text
19. BAB 5.pdf
Restricted to Repository staff only

Download (246kB)
[thumbnail of 20. Daftar Pustaka.pdf] Text
20. Daftar Pustaka.pdf

Download (278kB)

Abstract

Osteoartritis (OA) merupakan penyakit sendi ortopedi yang paling umum terjadi pada orang tua, dan dapat mempengaruhi jutaan orang di seluruh dunia. Secara klinis OA menyebabkan penyempitan ruang sendi, pembentukan osteofit dan sklerosis pada sendi yang terkena. Osteoartritis pinggul merupakan jenis OA kedua yang paling sering terjadi setelah lutut. Tren prevalensi osteoartritis dari Global Burden of Disease Study 2019 menunjukkan bahwa selama periode sepuluh tahun, prevalensi kasus OA meningkat sebesar 113,25%, dari 247,51 juta pada tahun 1990 menjadi 527,81 juta pada tahun 2019. Tujuan penelitian ini yaitu Mengembangkan model untuk identifikasi OA pinggul menggunakan ROI (Region of Interest), CLAHE (Contrast Limited Adaptive Histogram Equalization), Bayesian search dan Convolutional neural network dengan arsitektur DenseNet-169 Metode yang digunakan yaitu CNN dengan arsitektur DenseNet-169, ROI, CLAHE dan Bayesian search. Bahasa pemrograman yang digunakan yaitu bahasa pemrograman Python. Pelabelan data dilakukan oleh dua (2) orang dokter spesialis yang berpengalaman lebih dari 10 tahun. Pertama, merupakan dokter spesialis orthopaedi dan traumatologi yang memiliki sub-keahlian Hip & Knee Adult Reconstruction. Kedua, merupakan dokter spesialis radiologi konsultan muskuloskeletal. Jumlah data yang digunakan dalam penelitian ini yaitu 750 dataset dibagi menjadi 3 kelas yaitu OA Normal = 250, OA Ringan = 250 dan OA Berat = 250. Penelitian ini dimulai dengan melakukan proses ROI untuk melakukan segmentasi citra yang bertujuan untuk mengambil bagian tertentu dari citra pinggul yaitu acetabulum (socket), femoral head, dan femur, kemudian dilanjutkan dengan peningkatan kontras citra menggunakan CLAHE. Pelatihan model dilakukan dengan Bayesian search untuk memperoleh kombinasi hyperparameter terbaik. Terakhir dilakukan identifikasi menggunakan DenseNet-169. Hasil evaluasi dari penelitian yang telah dilakukan didapatkan nilai untuk accuracy = 95,33%, precision = 95,91%, recall = 95,33, dan F1-score = 95,39%. Kesimpulannya bahwa pendekatan integrasi CNN dengan arsitektur DenseNet-169, prapengolahan citra, dan optimasi hyperparameter secara efektif dapat mengidentifikasi osteoartritis pinggul.
Kata kunci: Osteoartritis, Osteoartritis pinggul, ROI, CLAHE, Bayesian search, DenseNet169

Osteoarthritis (OA) is the most common orthopedic joint disease affecting the elderly and impacts millions of people worldwide. Clinically, OA is characterized by joint space narrowing, osteophyte formation, and sclerosis in the affected joints. Among all types of OA, hip osteoarthritis ranks as the second most prevalent after knee osteoarthritis. According to the Global Burden of Disease Study 2019, the prevalence of OA cases increased significantly by 113.25% over a span of 30 years, from 247.51 million in 1990 to 527.81 million in 2019. This study aimed to develop an identification model for hip osteoarthritis by applying Region of Interest (ROI) extraction, Contrast Limited Adaptive Histogram Equalization (CLAHE), Bayesian hyperparameter optimization, and a Convolutional Neural Network (CNN) with the DenseNet-169 architecture. The methodology employed a CNN framework using DenseNet-169, integrated with ROI extraction, CLAHE preprocessing, and Bayesian optimization. All programming and model training were conducted using the Python language. Data labeling was performed by two experienced medical specialists with over ten years of clinical expertise: one is an orthopedic and traumatology specialist with sub-specialization in Hip & Knee Adult Reconstruction, and the other is a musculoskeletal consultant radiologist. The dataset consisted of 750 hip X-ray images, divided evenly into three classes: Normal OA (250 images), Mild OA (250 images), and Severe OA (250 images). The process began with ROI segmentation to isolate specific anatomical regions of the hip—namely, the acetabulum (socket), femoral head, and femur—followed by contrast enhancement using CLAHE. The model was trained using Bayesian optimization to determine the optimal set of hyperparameters, and classification was conducted using the DenseNet-169 architecture. The evaluation results demonstrated strong performance, with accuracy = 95.33%, precision = 95.91%, recall = 95.33%, and F1-score = 95.39%. In conclusion, the integration of CNN with DenseNet-169, combined with advanced image preprocessing and hyperparameter tuning, proved effective in identifying hip osteoarthritis.
Keyword: Osteoartritis, Osteoartritis pinggul, ROI, CLAHE, Bayesian search, DenseNet169.

Item Type: Thesis (Doctoral)
Uncontrolled Keywords: Osteoartritis, Osteoartritis pinggul, ROI, CLAHE, Bayesian search, DenseNet169
Subjects: Medicine
Sciences and Mathemathic
Divisions: Postgraduate Program > Doctor Program in Information System
Depositing User: ekana listianawati
Date Deposited: 04 Sep 2025 03:55
Last Modified: 04 Sep 2025 03:55
URI: https://eprints2.undip.ac.id/id/eprint/37952

Actions (login required)

View Item View Item