An Artificial Intelligence Olfactory-Based Diagnostic Model for Parkinson’s Disease Using Volatile Organic Compounds from Ear Canal Secretions

Abstract

Parkinson’s Disease (PD), a frequently diagnosed neurodegenerative condition, poses a major global challenge. Early diagnosis and intervention are crucial for PD treatment. This study proposes a diagnostic model for PD that analyzes volatile organic compounds (VOCs) from ear canal secretions (ECS). Using gas chromatography-mass spectrometry (GC-MS) to examine ECS samples from patients, four VOC components (ethylbenzene, 4-ethyltoluene, pentanal, and 2-pentadecyl-1,3-dioxolane) were identified as biomarkers with statistically significant differences between PD and non-PD patients. Diagnostic models based on these VOC components demonstrate strong capability in identifying and classifying PD patients. To enhance the accuracy and efficiency of the PD diagnostic model, this study introduces a protocol for extracting features from chromatographic data. By integrating gas chromatography-surface acoustic wave sensors (GC-SAW) with a convolutional neural network (CNN) model, the system achieves an accuracy of up to 94.4%. Further enhancements to the diagnostic model could pave the way for a promising new PD diagnostic solution and the clinical use of a bedside PD diagnostic device.