What a Clock Drawing Can Tell Us If We Can Read the Whole Drawing

How the Digital Pen Feature Extraction DREAM Challenge is mobilizing researchers to unlock data for early cognitive screening, by Gaia Andreoletti, Solly Sieberts, and Milan Vu

For decades, the clock drawing test has been one of the simplest and most widely used tools in cognitive screening: ask someone to draw a clock face, place all twelve numbers, and set the hands to a specified time. It requires no special equipment, takes only minutes, and draws on visuospatial planning, executive function, working memory, and semantic knowledge all at once.

But the traditional paper version has a real limitation: it only captures the finished drawing, not the process behind it. Two people can produce similarly scored clocks by very different routes — one drawing confidently and directly, another pausing, hesitating, or self-correcting along the way. That process-level detail is invisible to a paper-based score, even though it may carry real clinical signals.

Digitally administered versions of the test — the Digital Clock Drawing Test (dCDT) and related tasks — solve that by recording the drawing itself: timestamped pen position, stroke sequence, and pressure throughout. The result is a rich time series, not a static image.

Published research suggests that features derived from digital clock drawing test process data can relate to cognitive status in ways a visual score alone cannot, potentially enabling earlier detection of subtle change.

What’s the challenge?

The barrier has been practical, not scientific: converting raw pen coordinates and pressure readings into interpretable features is a nontrivial signal-processing task, and the commercial software capable of it is generally priced for clinical or industry use, leaving it out of reach for most academic groups. As a result, digital pen data already collected across several large cohort studies remains underused.

The Digital Pen Feature Extraction DREAM Challenge, organized by Sage Bionetworks with Tufts Medical Center and Boston University, is designed to close that gap. Using deidentified digital pen data contributed by the Rush Alzheimer’s Disease Center, the Long Life Family Study, the Integrative Longevity Omics project, and the Framingham Heart Study, participating teams will develop open-source methods to extract clinically relevant features, including those related to cognitive status and physical function.

This work is supported by the National Institute on Aging (NIA) and the ELITE Portal, 5U24AG078753 and 2U24AG061340. The ELITE Portal is an expanding hub for data, tools, publications and resources related to life span and health span research. Learn more at http://eliteportal.synapse.org .