Klasifikasi Osteoartritis Lutut Dengan Temporal Convolutional Network Berbasis Markerless Vision Pada Video Gait

Salsabila, Aghnia Tias (2026) Klasifikasi Osteoartritis Lutut Dengan Temporal Convolutional Network Berbasis Markerless Vision Pada Video Gait. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002221107-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (14MB) | Request a copy

Abstract

Osteoartritis lutut merupakan penyakit degeneratif sendi yang diderita oleh 528 juta penduduk di dunia. Penanganan medis bagi penderita penyakit ini didasari oleh derajat keparahan dan kondisi gaya berjalan (gait) sehingga diterapkan analisis gait untuk mendukung hasil radiografi. Analisis gait dapat menentukan derajat keparahan secara objektif berdasarkan informasi kinematika pasien dan dinamika temporal selama siklus berjalan. Temporal Convolutional Network (TCN) merupakan arsitektur deep learning berbasis 1D fully-convolutional network yang dirancang untuk memproses informasi temporal dari data sekuensial. Penelitian ini menerapkan dan menganalisis kinerja model TCN berbasis markerless vision dalam mengklasifikasikan derajat keparahan osteoartritis lutut ke dalam tiga kelas, yaitu Early (EL), Moderate (MD), dan Severe (SV). Sebanyak 93 video gait diolah melalui estimasi pose 2D menggunakan MediaPipe, ekstraksi empat fitur kinematika (sudut lutut, sudut panggul, serta kecepatan dan percepatan sudut lutut), normalisasi Z-score, serta segmentasi sliding window sebelum diproses oleh arsitektur TCN dengan dilated causal convolutions. Eksperimen dilakukan melalui dua skenario, yaitu konfigurasi hyperparameter dan evaluasi kombinasi fitur. Model terbaik memperoleh akurasi 73,64%, precision 67%, recall 65%, dan F1-score 63% (macro average). Kelas SV diklasifikasikan paling baik (F1-score 97%, recall 96%), sedangkan kelas EL paling sulit diklasifikasikan karena informasi kinematikanya tumpang tindih dengan kelas MD. Hasil ini menunjukkan bahwa TCN berbasis markerless vision merupakan sistem yang dapat diandalkan untuk analisis gait osteoartritis lutut secara noninvasif dan terjangkau.
====================================================================================================================================
Knee osteoarthritis is a degenerative joint disease affecting 528 million people worldwide. Medical management for patients with this condition is based on severity and gait characteristics, making gait analysis a crucial complement to radiographic findings. Gait analysis objectively determines severity levels by utilizing the patient's kinematic information and temporal dynamics during the gait cycle. Temporal Convolutional Network (TCN) is a deep learning architecture based on a 1D fully-convolutional network designed to process temporal information from sequential data. This study implements and evaluates the performance of a markerless vision-based TCN model to classify knee osteoarthritis severity into three classes: Early (EL), Moderate (MD), and Severe (SV). A total of 93 gait videos were processed through 2D pose estimation via MediaPipe, extraction of four kinematic features (knee angle, hip angle, knee angular velocity, and knee angular acceleration), Z-score normalization, and sliding window segmentation prior to processing by the TCN architecture with dilated causal convolutions. Experiments were conducted under two scenarios: hyperparameter configuration and feature combination evaluation. The best-performing model achieved an accuracy of 73.64%, precision of 67%, recall of 65%, and a macro-average F1-score of 63%. The SV class was best identified (F1-score of 97%, Recall of 96%), whereas the EL class proved to be the most challenging to classify due to overlapping kinematic information with the MD class. These findings demonstrate that the markerless vision-based TCN is a reliable, non-invasive, and cost-effective system for knee osteoarthritis gait analysis.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Osteoartritis Lutut, Analisis Gait, Temporal Convolutional Network (TCN), Markerless Vision, Knee Osteoarthritis, Gait Analysis
Subjects:	Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) 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 Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Aghnia Tias Salsabila
Date Deposited:	28 Jul 2026 04:53
Last Modified:	28 Jul 2026 04:53
URI:	http://repository.its.ac.id/id/eprint/138509

Actions (login required)

View Item