SoftSpot3D: prospective study of mobile 3D reconstruction for infants' cranial conditions.

Abstract

Purpose Deformational plagiocephaly and craniosynostosis require early detection for effective and less invasive interventions. Although 3D imaging and automated analysis allow for image-based assessment by nonspecialists, current 3D imaging systems are often costly and immobile. We present SoftSpot3D, a mobile three-dimensional head reconstruction framework designed as an accessible, low-cost screening solution for infant cranial assessment. Approach The system is independent of marker placement, leveraging a patterned cap strictly for feature-based tracking rather than geometric registration. Developed for iPad using consumer-grade RGB-D sensors, our pipeline mitigates infant motion through RGB-D odometry and generalized iterative closest point refinement to generate accurate cranial models. Clinical metrics are then automatically extracted by analyzing cross-sectional contours at the standardized measurement plane. Results Validated against the 3dMD (3dMD LLC, Atlanta, Georgia, United States) medical-grade stereophotogrammetry system on 143 infants (mean age: 5.73 ± 2.99 months; 98 males, 45 females), the system achieved an average cranial Hausdorff distance of 2.94 ± 1.32    mm compared with 3dMD reference meshes. Key diagnostic shape indices demonstrated high clinical agreement, with mean absolute errors of 2.18 ± 1.63 for cranial index and 2.16 ± 1.90 for cranial vault asymmetry index (both unitless ratios). Conclusion These results establish the clinical accuracy of the system as an accessible, low-cost screening solution, providing the evidence base for its integration into routine pediatric care to enable more efficient early detection and referrals. Furthermore, the framework automatically generates a structured clinical report incorporating quantitative metrics and multiview visualizations, facilitating immediate point-of-care decision-making.