DermDepth: Toward Monocular Metric Scale 3D Reconstruction Models for Dermatology

2026-07-14

AI Summary

This study addresses the lack of monocular image-based methods for real-scale 3D reconstruction in dermatology that do not require additional hardware, which has hindered precise quantification of skin lesion morphology. The authors propose DermDepth, the first model to achieve dense, real-scale 3D reconstruction and surface normal estimation from a single dermoscopic image. Key contributions include the creation of D-Synth, the first synthetic dermoscopy dataset with pixel-level 3D annotations, and a training strategy combining synthetic data pretraining with fine-tuning on a small number of real samples. Evaluated on clinical data, the method reduces scale error from over 16× to within 1.1×, significantly improving geometric accuracy and texture fidelity. Its strong generalization is further demonstrated across three diverse benchmarks encompassing varying scales, skin tones, and chronic wound types.

Abstract

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.

Problem

• monocular 3D reconstruction
• metric scale
• dermatology
• skin lesion measurement
• 3D surface normals

Innovation

• monocular 3D reconstruction
• metric scale estimation
• dermoscopic imaging
• synthetic dataset
• surface normal estimation