Deteksi Penyakit Parkinson Berbasis Tulisan Tangan Dan Ketukan Jari Menggunakan Metode Pembelajaran Mesin

Arraziqi, Dwi (2026) Deteksi Penyakit Parkinson Berbasis Tulisan Tangan Dan Ketukan Jari Menggunakan Metode Pembelajaran Mesin. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

Text 07111960010009-PHD_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Penyakit Parkinson merupakan gangguan neurodegeneratif progresif yang menurunkan fungsi motorik halus. Penelitian ini mengembangkan sistem deteksi Parkinson berbasis Internet of Things (IoT) menggunakan smartphone, stylus pen, cloud computing, dan perangkat lunak PIoT. Pada modul tulisan tangan spiral (off-line), data dianalisa menggunakan fitur feature descriptor (Histogram of Gradient (HOG), Oriented FAST and Rotated BRIEF, Speed-Up Robust Feature, Scale-Invariant Feature Transform, Color Gradient Histogram dan KAZE). Data dikurasi oleh Adriano de Oliveira Andrade dan Joao Paulo Folado dari NIATS Universitas Federal Uberlˆandia. Pengujian menggunakan Random Forest (RF) menghasilkan akurasi terbaik 81,67% pada HOG. Sementara pada modul tulisan tangan spiral (on-line), partisipan diminta menggambar spiral menggunakan stylus pen di smartphone. Data dianalisis dengan fitur kinematic serta tiga fitur baru: rotation consistency, 1D magnitude Dynamic Time Warping, dan micropause. Pengujian menggunakan RF dengan skema validasi Leave One Subject Out pada tiga dataset (in-house, ParkinsonHW, dan PaHaW) menghasilkan akurasi terbaik 99,23% pada data internal, 93,42% pada ParkinsonHW, dan 65,33% pada PaHaW. Pada modul ketukan jari, data dianalisis menggunakan jarak peak amplitude 1-10. Klasifikasi menggunakan K-Nearest Neighbors, Support Vector Machine, Convolutional Neural Network, Logistic Regression, Na¨ıve Bayes, dan Decision Tree dengan skema train:test (70:30, 80:20, 90:10) menunjukkan rata-rata akurasi 94%, dengan nilai AUC = 1 pada hampir semua model. Penelitian ini menunjukkan bahwa kedua pendekatan — tulisan tangan spiral dan ketukan jari berbasis IoT — mampu mendeteksi gangguan motorik akibat penyakit Parkinson secara akurat dan real-time. Ke depan, membuka peluang implementasi dalam telemedisin dan monitoring jarak jauh pasien.
=======================================================================================================================================
Parkinson’s disease is a progressive neurodegenerative disorder that impairs fine motor function. This study develops an Internet of Things (IoT)-based Parkinson’s detection system using smartphones, a stylus pen, cloud computing, and PIoT software. In the off-line spiral handwriting module, data were analyzed using feature descriptor techniques, including Histogram of Oriented Gradients (HOG), Oriented FAST and Rotated BRIEF (ORB), Speeded-Up Robust Features (SURF), Scale-Invariant Feature Transform (SIFT), Color Gradient Histogram, and KAZE. The data were curated by Adriano de Oliveira Andrade and Jo˜ao Paulo Folado from NIATS, Federal University of Uberlˆandia. Testing with the Random Forest (RF) classifier achieved the best accuracy of 81.67% using HOG features. In the on-line spiral handwriting module, participants were asked to draw a spiral using a stylus pen on a smartphone. The data were analyzed using kinematic features along with three newly proposed features: rotation consistency, 1D magnitude Dynamic Time Warping, and micropause. RF testing with a Leave-One-Subject-Out validation scheme on three datasets (in-house, ParkinsonHW, and PaHaW) achieved the best accuracies of 99.23% on the internal dataset, 93.42% on ParkinsonHW, and 65.33% on PaHaW. In the finger-tapping module, data were analyzed using peak amplitude distances 1–10. Classification using K-Nearest Neighbors, Support Vector Machine, Convolutional Neural Network, Logistic Regression, Na¨ıve Bayes, and Decision Tree models with train:test splits of 70:30, 80:20, and 90:10 demonstrated average accuracy of 94%, with an AUC value of 1 for almost all models. This study demonstrates that both approaches—IoT-based spiral handwriting and finger-tapping—are capable of accurately and real-time detecting motor impairments caused by Parkinson’s disease. In the future, this opens opportunities for implementation in telemedicine and remote patient monitoring.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Analisis tulisan tangan, analisis ketukan jari, IoT, penyakit Parkinson, pembelajaran mesin, telemedisin, Finger tapping analysis, handwriting analysis, IoT, machine learning, Parkinson’s disease, telemedicine
Subjects:	T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20001-(S3) PhD Thesis
Depositing User:	Dwi Arraziqi
Date Deposited:	28 Jan 2026 06:09
Last Modified:	28 Jan 2026 06:09
URI:	http://repository.its.ac.id/id/eprint/130181

Actions (login required)

View Item