An Artificial Intelligence-Assisted Method for Dementia Detection Using Images from the Clock Drawing Test.

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• Early dementia detection is crucial for clinical trials and interventions.
• The Clock Drawing Test (CDT) shows potential for diagnosing dementia-related disorders.
• Developing a computer-aided screening tool using CDT images is feasible.

Purpose of the Study:

• To evaluate a machine learning model's ability to predict mild cognitive impairment or dementia using Clock Drawing Test (CDT) images.
• Assessing the predictive power of AI for cognitive decline.

Main Methods:

• Utilized images from 3,263 cognitively intact and 160 impaired subjects' CDT evaluations.
• Employed a deep learning algorithm to process CDT images, age, and education level for dementia status prediction.
• Framingham Heart Study data was used for in-person dementia evaluations.

Main Results:

• The deep learning model achieved an 81.3% AUC using only CDT images.
• A composite model incorporating age, education, and CDT predictions reached a 91.9% AUC and 94.6% F1 score.
• Demonstrated significant predictive accuracy for dementia status.

Conclusions:

• Established a proof-of-principle for using deep learning on CDT images to predict dementia.
• The approach offers a potentially cost-effective and easily deployable method for cognitive impairment detection.
• Further validation is needed for clinical implementation.