Early diagnosis of Parkinson disorder based on spiral drawing image modalities using hybrid transfer learning architecture

Article type: Research Article

Authors: Anisha, C.D.^{; *} | Arulanand, N.

Affiliations: Department of Computer Science and Engineering, PSG College of Technology, Peelamedu, Coimbatore, India

Correspondence: [*] Corresponding author. C.D. Anisha, Department of Computer Science and Engineering, PSG College of Technology, Peelamedu, Coimbatore, India. E-mail: [email protected].

Abstract: The Spiral Drawing Test (SDT) has become a prominent clinical marker for the early diagnosis of Parkinson’s Disorder (PD) by capturing tremor symptoms. The integration of AI algorithms into a PD diagnosis system has proven to be a breakthrough objective assessment that aids professionals in decision-making. However, there is a need for improvisation of the workflow architectures of AI models to optimize the diagnosis system by reducing the misdiagnosis rate. The proposed system presents PD prediction using a Spiral Drawing Test (SDT) image modality integrated with an Artificial Intelligence (AI) algorithm. The proposed study presents three hybrid workflow architectures formed by integrating three core layers: a data augmentation layer, Transfer Layer (TL)-based feature extraction layer, and Deep Learning (DL)-based classification layer. The results were analyzed by conducting 18 experiments based on the hyperparameter values and workflow architectures. The highest accuracy obtained by the proposed study is 98% for Hybrid Workflow Architecture II.

Keywords: Parkinson disorder, transfer learning

DOI: 10.3233/JIFS-231202

Journal: Journal of Intelligent & Fuzzy Systems, vol. 46, no. 1, pp. 769-787, 2024