106 - Cranial Shape and Head Circumference Measurements using Digital Photography

Deformational plagiocephaly and brachycephaly (DPB) is a common condition observed in 20% of newborns and infants. DPB can be detected by measuring the cranial index (CI) and cranial vault asymmetry index (CVAI). However, pediatricians do not have access to tools that measure these indices, which results in delays in diagnosis and treatment. In addition, head circumference (HC) is currently measured with a tape that results in variable data.

Objective:

To evaluate an automated method for cranial measurements, i.e., HC and cranial shape indices, using digital photography data.

Design/Methods: Three-dimensional (3D) retrospective photograms of 445 infant heads (303 male, 142 female) were collected after IRB approval and parental consent as required by the devices shown in Table 1. The average age of subjects was 12 ± 9 months with a range of 3-89 months. Distribution of cranial conditions for the subjects is shown in Figure 1. Top views of each head model were rendered as digital photographs of the child’s head and used with image analysis algorithms to automatically calculate HC, CI, and CVAI from the head contour. A reference sticker of 1 in diameter was placed on the head. An XGBoost model was trained and tested in a leave-one-out-cross-validation manner using the age, sex, and sticker shape and location to improve HC measurements. Results were compared to the ground truth measured on the 3D model via Spearman correlation and Bland-Altman (BA) analysis.

Results:

The mean absolute errors (MAE), Spearman correlation coefficients, and BA analysis results are presented in Table 2. BA plots are presented in Figure. 1. Spearman coefficients > 0.90 were measured between ground truth and automated measurements from digital photographs. Based on previous publications, MAE of CI and CVAI < 2, and BA limits of agreement smaller than [-4.55%, 3.65%] for HC are acceptable, and the automated method met both criteria in this study.

Conclusion(s):

The photographic-based digital method enables automated precise measurements of HC, CI, and CVAI. The results of this retrospective study show promise for the integration with a smartphone-based tool to simultaneously measure both cranial shape and size. The digital tool can be used in pediatrician offices and at the point of care to help with monitoring of growth and early detection of cranial conditions.