JMIR Publications

ABSTRACT

Background:

Positional cranial deformities are common in infancy. Smartphone-based applications are increasingly used to assess infant head shape, but evidence comparing these tools with three-dimensional (3D) scanners across age groups in community-based settings remains limited.

Objective:

This study evaluated the measurement agreement and screening performance of a mobile application for estimating cranial vault asymmetry index (CVAI) and cephalic index (CI), compared with a 3D scanner, when used by a trained examiner during routine municipal health checkups in Japan.

Methods:

We conducted a cross-sectional study in Goto City, Japan, between May 2023 and March 2025. Independent, age-specific cohorts of children attending municipal health checkups at 4, 18, and 42 months were included. Paired cranial measurements were obtained during the same visit by a trained examiner using a VECTRA H2 3D scanner and the mobile application according to a standardized protocol. Agreement was assessed using mean differences, 95% limits of agreement (LoA), and intraclass correlation coefficients (ICCs). Screening performance was evaluated using sensitivity and specificity based on prespecified clinical thresholds. Receiver operating characteristic (ROC) analyses were secondary exploratory analyses. The primary outcome was app-based CVAI for detecting scanner-defined moderate-or-worse plagiocephaly (CVAI ≥ 7%). Secondary outcomes were app-based CI for detecting brachycephaly (CI > 94%) and dolichocephaly (CI < 79%).

Results:

The final analytic samples comprised 262 children at 4 months, 171 at 18 months, and 216 at 42 months. Agreement for continuous CVAI was highest at 4 months (mean difference 0.14; 95% LoA −4.18 to 4.46; ICC 0.76, 95% CI 0.71–0.81) and lower at 18 months (ICC 0.42, 95% CI 0.29–0.54) and 42 months (ICC 0.49, 95% CI 0.38–0.59). For CI, the app consistently overestimated values relative to the scanner, with mean differences of 3.83, 3.24, and 2.61 points at 4, 18, and 42 months, respectively. For detecting scanner-defined moderate-or-worse plagiocephaly, app sensitivity/specificity were 84.8%/88.0% at 4 months, 50.0%/88.4% at 18 months, and 53.3%/89.6% at 42 months. Exploratory ROC analyses for app-based CVAI yielded AUCs of 0.93, 0.77, and 0.75 at 4, 18, and 42 months, respectively. Estimates at later ages and CI-based exploratory cutoff estimates should be interpreted cautiously because scanner-positive cases were limited.

Conclusions:

Under municipal health checkup conditions with a trained examiner, app-based CVAI showed the strongest performance at 4 months and may be useful for early screening. However, performance was weaker at 18 and 42 months. App-based CI systematically exceeded scanner-based values, indicating clinically meaningful overestimation that may increase false-positive classifications for brachycephaly when identical thresholds are applied. The generalizability of these findings to real-world use by untrained, non-medical users (e.g., caregivers) remains uncertain and warrants further investigation. Clinical Trial: UMIN Clinical Trials Registry (UMIN000052342)