Wearable-based activity recognition algorithms
Atopic Dermatitis, Eczema, and Psoriasis cause intense scratching, often during sleep. New drugs, especially biologics, greatly improve symptoms and QOL.
Circadic ran a pilot study to collect patient data (video and hand actigraphy), and designed and developed an algorithm to quantify nocturnal hand-scratching intensity using hand actigraphy.
Circadic has built a ground-truth data acquisition system comprised of a) a custom infrared camera, b) a smartphone (iPhone), and c) one or more smart watches (Apple Watch). All modules are connected and can acquire synchronized signals (internet time). Circadic designed and ran a pilot study to collect home-based patient data (video and hand motion). The video dataset was annotated by professional reviewers using ANYA annotation software: each motion event was characterized as scratching or not.
The data (ground truth annotations and motion IMU data) was then used to train an ML algorithm that can measure nocturnal (during sleep) hand-scratching intensity and frequency with AUC=0.94.
.png)
