Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces DermDepth, a novel monocular 3D reconstruction model tailored for dermatology, enabling accurate metric scale measurements and detailed surface normal texture estimates from single-view images. By leveraging the D-Synth dataset, which provides synthetic dermoscopic images with precise 3D information, the authors demonstrate a significant reduction in metric scale error from over 16x to under 1.1x when applied to real dermoscopic data. The model's ability to generalize across diverse skin tones and clinical cases underscores its potential for enhancing diagnostic accuracy in dermatological practice.
Achieving accurate 3D reconstructions from a single image could revolutionize how dermatologists assess and monitor skin lesions.
Dermatological practice routinely involves measuring and tracking lesion size, morphology and texture, as critical components of wound or skin cancer screening, monitoring and diagnosis. To accomplish this task, practitioners often image the skin surface with commonly available off-the-shelf camera sensors. This has led to an overwhelming research focus on 2D methods while these objectives naturally benefit from 3D information. In this paper, we demonstrate that dense monocular 3D reconstructions, metric scale measurements and rich surface normal texture estimates are achievable for both dermoscopic and macroscopic cases without the need for additional hardware or multiple captures. We present DermDepth, the first single-view metric scale 3D model for the dermatological domain and D-Synth, the first synthetic dermoscopic dataset with pixel-perfect 3D information. Our experiments show training DermDepth on D-Synth corrects metric scale error from over 16x to under 1.1x for real dermoscopic data, while preserving geometric quality and increasing texture richness. Fine-tuning on a small amount of real clinical samples generalizes our method across three real-world benchmarks spanning the few mm to hundred cm range, diverse skin-tones, chronic wound cases and produces measurements broadly consistent with disease size reported in medical literature. All code, data and models are available at https://github.com/hectorcarrion/dermdepth.