Objective: To explore the feasibility of using deep learning to differentiate individuals with parkinsonism from those without parkinsonism and without cognitive impairment through facial image analysis.

Background: Masked face is a well-recognized clinical feature of parkinsonism, but its utility in automated diagnosis remains underexplored. While masked face is useful for diagnosis in clinical settings, AI-based facial analysis could enable early detection at home, prompting individuals to seek medical evaluation sooner.

Method: Facial images of 80 individuals with parkinsonism and 1,184 individuals without parkinsonism and without cognitive impairment were analyzed. A deep learning model was trained using a 10-fold cross-validation approach to classify the two groups based on facial features.

Results: The deep learning model achieved an accuracy of approximately 90% in differentiating individuals with parkinsonism from those without parkinsonism and without cognitive impairment.

Conclusion:

This preliminary study suggests that deep learning-based facial analysis has potential for detecting parkinsonism. Further validation with larger datasets and clinical applications is needed to assess its reliability and practical utility.

References: Umeda-Kameyama Y, Kameyama M, Tanaka T, Son BK, Kojima T, Fukasawa M, Iizuka T, Ogawa S, Iijima K, Akishita M. Screening of Alzheimer’s disease by facial complexion using artificial intelligence. Aging (Albany NY). 2021 Jan 25;13(2):1765-1772. doi: 10.18632/aging.202545. Epub 2021 Jan 25. PMID: 33495415; PMCID: PMC7880359.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/exploring-deep-learning-based-facial-analysis-for-detecting-parkinsonism/