Deteksi Dini Penyakit Parkinson Melalui Gambar Tangan Dengan Menggunakan Metode EfficientDet

Abstract

Parkinson’s disease is a progressive neurodegenerative disorder that affects the central nervous system and impairs fine motor skills such as writing or drawing. Early detection is crucial to slow the progression of symptoms and improve patients’ quality of life. This study aims to detect early signs of Parkinson’s disease by analyzing hand-drawn spiral and wave patterns using the EfficientDet method. The dataset used consists of 3,264 annotated images obtained from the Kaggle platform, which were preprocessed and augmented using Roboflow. The EfficientDet-D0 model was trained with varying numbers of epochs to evaluate its classification performance. The model used in this study was obtained at epoch 86, which achieved the best performance. Experimental results show that the system can effectively distinguish between drawings from Parkinson’s patients and healthy individuals, achieving an accuracy of 90.8% a precision of 97.6%, a recall of 92.8%, and an F1-score of 95.1%. Furthermore, the model was deployed as a Flask-based web application to classify input images into two categories: Parkinson and Healthy. These findings indicate that EfficientDet-D0 has strong potential as an efficient, accurate, and non-invasive tool for early diagnosis of Parkinson’s disease.