Fig. 2: Performance of the automated analysis pipeline.
From: Assessing motor development with wearables in low-resource settings: feasibility in rural Malawi
The example scatter plots (a) and Bland-Altman plots (b) show the correspondence between results obtained after manual (human observer) vs. automated (AI algorithm) playtime detections. The findings are shown for the holistic motor measures (BIMS, scale 0-100) and the time spent in prone or standing postures. Note their strong correlation (scatter plots), lack of significant bias as a function of motor measure (b), as well as the highly comparable distributions (KS-test) of data from Malawi (orange) and Finland (gray).
