The ART device comprises an adjustable headband with two bone conduction transducers positioned circumferentially around the head with a central alignment over the nasal bridge (Fig. 1). This device was previously investigated for the treatment of nasal congestion by Luong et al. [9] It features Bluetooth connectivity, which allows the accompanying smartphone app to administer the therapy. The app uses the front-facing camera of a smartphone to capture and annotate multiple facial anatomical landmarks. With more than 1,000 data points collected, a virtual mesh of the participant’s face is constructed, and a machine learning predictive model is used to estimate the sinus volumes of the participant. The lid-cheek junction and certain bony landmarks, such as the orbital rims, nasal bridge, zygomatic arch, and mandibular angle, are among the craniofacial features least affected by fat changes and aging. The lid-cheek junction remains relatively stable over time, as shown in serial photographs, in contrast to the more dynamic facial regions. While bony landmarks may undergo some resorption with age, they are generally less affected by fat redistribution and volume loss than other facial areas. The unique resonant frequency and associated harmonics for each participant are calculated using a proprietary algorithm with all resonant frequencies within the audible spectrum of 100 Hz to 1 kHz (Fig. 2).

Participants received 45-minute audio sessions delivered to the ART device via Bluetooth. They were instructed to don the device and activate the therapy while attempting to sleep. The headband remained in place once sleep was initiated. The participants who were unable to fall asleep within the first 45 minutes were instructed to reuse the device for an additional session.

Eligible participants were adults aged 18 years or older who had experienced at least three months of persistent difficulties with sleep initiation, maintenance, or early morning awakenings, with associated daytime impairment. An ISI score greater than 15 was indicative of moderate-to-severe insomnia, establishing a severity benchmark [10].

Exclusion criteria included the use of sleep aids, both prescription and over-the-counter, as well as sedating, alerting, or antidepressant medications within three months prior to screening. Individuals with a prior diagnosis of any autoimmune or medical disorder likely to affect sleep were excluded. Pregnant females were not permitted to participate in this study.

The primary outcome measure was a statistically significant reduction in ISI scores from baseline after four weeks of daily therapy. The secondary outcome was the response rate, defined as a 7-point or greater reduction in the ISI from baseline.

Safety endpoints were the absence of serious adverse events such as death, life-threatening conditions, permanent impairments, hospital admissions, or medical or surgical interventions to prevent permanent impairment.

Sample size calculations were based on 80% power and an alpha of 0.05, anticipating a 7-point decrease in the ISI, which aligns with the effect size seen in CBT-I [11]. Twenty participants were required to achieve this power.

Data were presented using graphs and descriptive statistics. Categorical variables were reported as counts and percentages, while continuous variables were summarized as means, standard deviations, standard errors, and 95% confidence intervals. ISI score changes from baseline to the four-week mark were analyzed using paired t-tests and linear mixed models with repeated measures. These models accounted for demographic variables and yielded an adjusted treatment effect estimate.

The clinical significance threshold was set at a 7-point or greater reduction in ISI from baseline, matching the accepted impact of CBT-I to maintain high sensitivity and specificity for identifying moderate-to-marked improvements in participants [11].