Abstract Details

The digital Clock Drawing Task extracts cognitive- and motor-related features, enabling automatic CI detection in 3 minutes. Here, we applied the BPA to 4 other assessments to expand upon that capability.

To evaluate four novel tablet-based drawing assessments for detecting cognitive impairment (CI) using automated analysis of traditional and Boston Process Approach (BPA) features.

Novel process-based metrics captured during drawing/writing of Trails A/B, Symbol Digit Modalities Test (SDMT), Spiral Tracing, and Pathfinding were analyzed in two independent datasets: (1) 170 cognitively normal (CN) subjects to calculate test-retest reliability of features, and (2) 1000 CN and cognitively impaired (CI) subjects to quantify their group differences.

Average reliability was good across features; SDMT: r=0.54 (s.d.=0.2, range=-0.02–0.92), Trails B: r=0.42 (s.d.=0.25, range=-0.19–0.87), Spiral Tracing: r=0.5 (s.d.=0.19, range=-0.07–0.78), Pathfinding: r=0.47 (s.d.=0.2, range=0.07–0.78). Trails: CN drew faster and spent less time in each circle (|T|’s>2.8, p’s<0.001); SDMT: CI spent less time thinking/pausing (T=-4.01, p<0.0001) and more time writing (T=9.73, p<0.001); Pathfinding: CI drew faster with higher acceleration  (T=3.51, p<0.0001) and drew less (T=-6.48, p<0.0001); Spiral Tracing: CN deviated less (T=-2.88, p<0.0001), had lower velocity (T=-5.22, p<0.0001), and spent a larger portion of time drawing (T=11.57, p<0.0001) with less variability (T=-3.83, p<0.0001).

Our results demonstrate the value of process-based metrics during neuropsychological tasks in distinguishing between CN and CI individuals with fair-to-good test-retest reliability, thereby identifying specific processes that can lead to actionable recommendations.