Evaluating Neus-Facto Lesion Surface-Area Measurement Using a 3D-Printed Anatomical Phantom

Accurate measurement of the surface area of skin lesions and chronic wounds is a key element of monitoring their dynamics—both healing and progression—yet traditional manual methods are invasive, slow, and highly operator-dependent, while two-dimensional (2D) digital approaches based on smartphone imaging distort area geometry on curved anatomical surfaces. Three-dimensional (3D) surface reconstruction removes these limitations, and among current methods the family of neural implicit surfaces—in particular Neus-Facto—demonstrates the highest geometric accuracy in a recent empirical benchmark study. However, existing evaluations use a structured-light 3D scanner as the reference, whose own measurement error is confounded with the error of the method under test; the test phantoms are isolated from the anatomical context of the body, and repeatability is assessed on frame subsets of a single video. In this work, we propose an alternative path for evaluating Neus-Facto, where the reference geometry is defined not by measurement but by design—a digital CAD model realized as a 3D-printed anatomical foot phantom with an embedded lesion region. Capture is performed with a consumer-grade smartphone following a documented protocol; repeatability is assessed on genuinely separate physical sessions at different times, with variation in camera distance and viewpoint. This setup separates the error of the method from the error of the measurement reference and more closely approximates—though it does not yet reproduce—real clinical conditions; these results indicate potential for future clinical translation, while further validation under real clinical conditions is required.