To develop a scalable solution for dementia testing.

Over 57 million cases of dementia exist worldwide, with projections rising to 153 million by 2050. The healthcare system is currently overwhelmed, taking an average of three years to diagnose symptoms. Moreover, up to 50% of dementia patients die without a diagnosis. In the age of dementia therapies, timely and accurate testing is urgently needed.

We created dementia tests that allow patients to speak, draw, and write as they would with in-person cognitive testing. A suite of algorithms was developed to address various cognitive aspects traditionally difficult to assess. These include custom neural networks for complex drawing evaluation, geolocation for orientation, natural language processing for language skills, memory formation tests, and touchscreen-based executive function assessments. Based on patient feedback from a pilot study, we refined the tests for accessibility. A randomized-controlled trial validated the tests against traditional paper-based evaluations. We also developed a one-minute screening test using touchscreen drawings. Cloud computing infrastructure was then developed to distribute these tests universally.

Our short screening test achieved an ROC AUC of 0.80 (95% CI 0.67-0.91) with a sensitivity of 1.0 (95% CI 1.0-1.0). The detailed diagnostic test showed an ROC AUC of 0.93 (95% CI 0.85-0.99) and a specificity of 1.0 (95% CI 1.0-1.0). An administration platform was created to deploy tests to caregivers or patients.

Our testing system serves two purposes: 1) it collects high-quality neurological data at an unprecedented scale, fostering new research and innovation; 2) it offers a scalable, expert-quality testing to anyone, anywhere. Our two-step platform first screens patients, then directs those with positive results to a more comprehensive diagnostic test. This facilitates fast-tracking for treatment, while those testing negative can be re-evaluated annually.