Quantifying Itch and its Impact on Sleep Using Machine Learning and Radio Signals

TL;DR

This work presents a passive radio frequency sensing approach to objectively quantify nocturnal scratching and its impact on sleep in chronic pruritus. A Wi-Fi–like RF device paired with a three-component neural network detects scratching and assesses sleep from home, validated against infrared video ground truth with ROC AUC around $0.997$ and high per-participant accuracy. The study reveals robust associations between scratching and poorer sleep, including increased sleep latency and wakefulness, while showing only a weak link between objective scratching and subjective itch reports. The method offers privacy-preserving, long-term at-home monitoring that can support clinical care and pharmaceutical trials by providing objective, continuous itch and sleep measurements without wearable burden.

Abstract

Chronic itch affects 13% of the US population, is highly debilitating, and underlies many medical conditions. A major challenge in clinical care and new therapeutics development is the lack of an objective measure for quantifying itch, leading to reliance on subjective measures like patients' self-assessment of itch severity. In this paper, we show that a home radio device paired with artificial intelligence (AI) can concurrently capture scratching and evaluate its impact on sleep quality by analyzing radio signals bouncing in the environment. The device eliminates the need for wearable sensors or skin contact, enabling monitoring of chronic itch over extended periods at home without burdening patients or interfering with their skin condition. To validate the technology, we conducted an observational clinical study of chronic pruritus patients, monitored at home for one month using both the radio device and an infrared camera. Comparing the output of the device to ground truth data from the camera demonstrates its feasibility and accuracy (ROC AUC = 0.997, sensitivity = 0.825, specificity = 0.997). The results reveal a significant correlation between scratching and low sleep quality, manifested as a reduction in sleep efficiency (R = 0.6, p < 0.001) and an increase in sleep latency (R = 0.68, p < 0.001). Our study underscores the potential of passive, long-term, at-home monitoring of chronic scratching and its sleep implications, offering a valuable tool for both clinical care of chronic itch patients and pharmaceutical clinical trials.

Figures (5)

• Figure 1: Illustration of video-based and radio-based assessment of scratching and sleep. (a) Assessment of nocturnal scratching and sleep through human annotation of videos of patients in the bedrooms. (b) Assessment of nocturnal scratching and sleep by analyzing the radio frequency (RF) signals that bounce off patients during sleep using machine learning.
• Figure 2: Performance of radio-based scratching assessment. The left column presents results for scratching lasting longer than 3 seconds, while the right column presents results for all scratching occurrences. (a), (b) compare the scratching time per hour (STH) and scratching bouts per hour (SBH) estimated by our radio-based approach to manual annotations of scratching in videos, for scratching lasting more than 3 seconds. Each point in the figures corresponds to a full night of data for one patient (totaling 364 nights across n=20 participants). The figures show that the assessment of scratching using the radio device has a strong correlation with the camera-based ground truth (R = 0.95, p < 0.001 for STH, R = 0.83, p < 0.001 for SBH). (c) shows the ROC curve and the AUC of radio-based scratching assessment (ROC AUC = 0.997). (d) shows the Precision-Recall curve and the PR AUC of radio-based scratching assessment (PR AUC = 0.876). (e)-(h) plot the STH correlation, SBH correlation, ROC curve, and Precision-Recall curve, respectively, comparing the scratching predicted by the radio device against the ground truth from video recording, for all scratching occurrences regardless of their duration.
• Figure 3: Per-participant performance evaluation of radio-based scratching assessment. Each plot shows the Pearson’s correlation between the scratching time per hour inferred from radio signals and the ground truth video annotations for a particular participant. Each point in these plots corresponds to a full night of data for the participant. The participants are ordered with respect to their maximum scratching time per hour across all the nights, from highest to lowest. The scratching time measured by the radio device is significantly correlated with the ground truth scratching time per hour for all participants across different disease diagnoses (color coded with red, green, and blue for AD, PN, and CPUO, respectively) and varying disease severity.
• Figure 4: Relationship of scratching with sleep quality in individuals with chronic pruritus. (a) plots the relationship between sleep disturbance and scratching, where sleep disturbance is the percentage of time spent awake out of the total time in bed (for 341 nights from 20 participants). (b) plots the relationship between sleep latency and scratching, where sleep latency is the time until sleep onset. (c) plots the relationship between WASO and scratching, where WASO is defined as the time spent awake after sleep onset. (d) shows the average scratching per night per participant for each sleep stage. One-sample one-sided Wilcoxon rank sum tests were conducted to assess significance. In each box, the central line indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to 1.5 times the interquartile range. Three asterisks indicate p < 0.001.
• Figure 5: Relationship between scratching measurements and self-reported NRS. (a), (b) show scatterplots of the scratching time per hour (STH) and the self-reported NRS for each participant (342 nights from 20 participants), where scratching time is measured from video annotation in (a) and radio signals in (b). (c), (d), (e) focus on a participant that received a shot of Dupilumab 300mg during the study. In each subfigure we compare the STH, SBH, or NRS before and after one dose of treatment. Two-sample, one-sided Wilcoxon rank sum tests are performed to assess significance. Significant reduction is observed for all three cases (p < 0.001, p < 0.05 and p < 0.001, respectively). In each box plot, the central line indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to 1.5 times the interquartile range.