Enhancing Early Detection of Parkinson's Disease Through Multimodal Analysis of Hand-Drawn Images and Voice Features

Parkinson’s disease (PD) is a neurological disorder that gradually impairs both movement and cognitive abilities, impacting millions of people worldwide. Timely identification of PD is essential, as it allows for interventions that can improve patient quality of life and slow disease progression. In this work, we introduce a multimodal approach that integrates analysis of hand-drawn spiral and meander patterns with voice recordings to facilitate earlier PD detection. We utilize a combination of established machine learning and deep learning techniques including Decision Tree, Random Forest, VGG16, and ResNet50 architectures to extract and classify distinctive features from these data sources. Unlike many previous investigations that concentrate on a single type of biomarker, our framework brings together handwriting and speech-based indicators to offer a more holistic assessment. The process involves careful data preparation, selection of relevant features, and systematic tuning of model parameters to achieve optimal results. Experimental results on publicly available datasets demonstrate that ResNet50 achieved up to 85% accuracy for spiral images, while Random Forest attained 96.61% accuracy on voice data. Comparative analysis with recent literature underscores the effectiveness of our multimodal strategy. By utilizing non-invasive, easily collectible digital biomarkers and computationally efficient machine learning techniques, our approach offers a scalable, costeffective, and environmentally sustainable solution for early PD diagnosis. This supports the broader goals of equitable healthcare and reduced resource consumption, aligning with the principles of sustainable development in medical technology. The findings highlight the potential of combining digital drawing and voice analysis as a non-invasive, cost-effective, and scalable solution for early PD diagnosis, offering valuable insights for clinical application and future research