Wound3DAssist: A Practical Framework for 3D Wound Assessment

Remi Chierchia, Rodrigo Santa Cruz, Léo Lebrat, Yulia Arzhaeva, Mohammad Ali Armin, Jeremy Oorloff · 2025

Read Wound3DAssist: A Practical Framework for 3D Wound Assessment by Remi Chierchia, Rodrigo Santa Cruz, Léo Lebrat, Yulia Arzhaeva, Mohammad Ali Armin, Jeremy Oorloff (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-17635.txt.

Introduction

Remi Chierchia and co-authors build a tool that turns phone video into 3D models of chronic wounds, in a 2025 paper. Wound care still leans on hand notes and ruler checks, which vary by nurse and eat time. Flat photo tools warp sizes and miss depth, above all on curved limbs. Their frame, Wound3DAssist, works from short handheld clips shot on plain consumer gear.

The pipeline rebuilds the wound surface in 3D, cuts out the wound bed, sorts its tissue types, and checks the rim of skin around it. Tests ran on digital models with known shapes, on silicone fakes, and on real patients. The builds were sharp, near millimeter accuracy in the clinic cases, with clear views of the wound bed. A full check took under twenty minutes. The team frames it as research gear, ready for trials in real care flow.

Catalog record

CREATORRemi Chierchia, Rodrigo Santa Cruz, Léo Lebrat, Yulia Arzhaeva, Mohammad Ali Armin, Jeremy Oorloff
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-17635

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Wound3DAssist: A Practical Framework for 3D Wound Assessment Thanks: This work was supported by the MRFF Rapid Applied Research Translation grant (RARUR000158), CSIRO AI4M Minimising Antimicrobial Resistance Mission, and Australian Government Training Research Program (AGRTP) Scholarship. Thanks: Remi Chierchia, David Ahmedt-Aristizabal, Rodrigo Santa Cruz, Yulia Arzhaeva, Mohammad Ali Armin, Jeremy Oorloff, and Chuong Nguyen are with the Imaging and Computer Vision Group, Data61, CSIRO (e-mail: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]). Thanks: Remi Chierchia, Léo Lebrat, Olivier Salvado, Clinton Fookes are with the School of Electrical Engineering & Robotics, Queensland University of Technology, Australia (e-mail: [email protected]; [email protected]; [email protected]; [email protected]). Remi Chierchia, Rodrigo Santa Cruz, Léo Lebrat, Yulia Arzhaeva, Mohammad Ali Armin, Jeremy Oorloff Abstract Managing chronic wounds remains a major healthcare challenge, with clinical assessment often relying on subjective and time-consuming manual documentation methods. Although 2D digital videometry frameworks aided the measurement process, these approaches struggle with perspective distortion, a limited field of view, and an inability to capture wound depth, especially in anatomically complex or curved regions. To overcome these limitations, we present Wound3DAssist, a practical framework for 3D wound assessment using monocular consumer-grade videos. Our framework generates accurate 3D models from short handheld smartphone video recordings, enabling non-contact, automatic measurements that are view-independent and robust to camera motion. We integrate 3D reconstruction, wound segmentation, tissue classification, and periwound analysis into a modular workflow. We evaluate Wound3DAssist across digital models with known geometry, silicone phantoms, and real patients. Results show that the framework supports high-quality wound bed visualization, millimeter-level accuracy, and reliable tissue composition analysis. Full assessments are completed in under 20 minutes, demonstrating feasibility for real-world clinical use. I Introduction Chronic wounds pose a significant clinical and economic burden, affecting millions of patients worldwide and costing healthcare systems billions annually [ 1 , 2 ] . Accurate and consistent wound assessment is essential for diagnosis, monitoring, and treatment planning [ 3 ] . Traditional wound documentation methods, such as manual measurements and 2D digital videometry, are invasive and present limitations including operator variability, perspective ambiguity, and lack of depth perception [ 4 ] . These challenges hinder longitudinal tracking and objective comparison of wound healing over time. Recent advances in computer vision and 3D reconstruction have opened the door to more precise and standardized wound analysis [ 5 , 6 ] . Three-dimensional (3D) data provide richer geometric context, enabling accurate wound area, perimeter, and depth measurements—critical metrics for evaluating wound progression [ 7 ] . Moreover, 3D models allow better visualization and communication among clinicians, improving documentation and decision-making. Several research efforts have explored the application of 3D imaging to wound care, employing technologies such as structured light [ 8 , 9 ] , multi-spectral [ 10 , 11 ] , and photogrammetry [ 12 , 13 ] . While promising, these works often remain limited in scope or usability. Many rely on proprietary hardware not publicly available or require extensive post-processing and technical expertise, making them impractical for routine clinical deployment. Critically, there is also a lack of standardized datasets for benchmarking, and few existing works provide an end-to-end pipeline that integrates 3D reconstruction, tissue segmentation, and automated wound measurement [ 13 , 11 , 14 , 5 , 15 ] into a single solution. In this paper, we introduce Wound3DAssist, a practical and extensible framework for 3D wound assessment. Our framework leverages photogrammetry techniques and consumer-grade video to create high-quality colored 3D models. It incorporates neural-network-based 2D segmentation, which is projected onto the reconstructed surface to enable automated tissue and wound region analysis. Wound3DAssist supports four core wound assessments: i) Wound bed visualization – Interactive 3D inspection of the wound site. ii) Wound bed segmentation and assessment - Automated delineation of the wound boundary and computation of clinical wound metrics such as area, perimeter, and depth. iii) Wound tissue assessment - Identification of tissue types (granulation, slough, necrotic, and epithelial) within the wound bed. iv) Periwound assessment - Characterization of the surrounding skin tissue. \begin{overpic}[report_figures/main_draft.png] \put(95.0,85.0){\small\parbox{85.35826pt}{\centering Frame Extraction\\ \& Selection\@add@centering}} \put(370.0,136.0){\small 3D Reconstruction} \put(345.0,300.0){\small{Wound Bed Visualization}} \put(305.0,-16.0){\small 2D Segmentation \& Post-processing} \put(625.0,59.0){\small\parbox{85.35826pt}{\centering 2D-to-3D Segmentation Mapping\@add@centering}} \put(615.0,233.0){\small{Wound Tissue assessment}} \put(851.0,68.0){\small\parbox{85.35826pt}{\centering 3D Wound\\ Documentation\@add@centering}} \put(850.0,240.0){\small\parbox{85.35826pt}{\centering{Wound Bed \& Periwound assessment}\@add@centering}} \end{overpic} Fig. 1 : Main components and capabilities of our proposed wound documentation framework Wound3DAssist : Frame Extraction & Selection (Sec. IV-A1 ); 3D Reconstruction (Sec. IV-A2 ) supporting interactive wound bed visualization; 2D Segmentation & Post-processing (Sec. IV-B1 ); 2D-to-3D Segmentation Mapping (Sec. IV-B2 ) enabling automated wound tissue assessment; and 3D Wound documentation (Sec. IV-C ) reporting comprehensive wound bed and periwound assessment. By combining off-the-shelf hardware with a robust computational pipeline, Wound3DAssist offers a deployable solution for real-world wound monitoring. In addition, we provide detailed evaluation results using a combination of clinical real-world recordings, silicone replicas (or phantoms), and digital datasets with ground-truth geometry. We also discuss the challenges of dataset availability, cross-modality consistency, and uncertainty introduced by imaging artifacts—offering insights and tools for the development of future wound analysis systems. II Experimental Design A critical step in validating 3D wound assessment frameworks is access to diverse, well-characterized datasets that reflect both controlled and real-world conditions. However, 3D wound datasets with ground truth annotations are scarce and often lack the realism or variability needed to test model robustness across clinical contexts. Collecting 3D data from actual wounds presents additional difficulties: it is time-consuming, may cause patient discomfort, and can disrupt clinical workflows—particularly when aiming to capture a broad range of wound types, skin tones, lighting conditions, and patient positions. Existing publicly available wound datasets are commonly tailored for 2D analysis of single-frame images [ 16 , 17 , 18 , 19 , 20 ] , with few allowing for 3D imaging [ 10 , 21 ] , synthetic modeling [ 22 ] , or medical phantoms such as Symour II 1 1 1 https://www.mentone-educational.com.au/health-education/health-education/seymour-ii-wound-care-model . To overcome these limitations, we design a multi-tiered evaluation strategy using: (i) a digital wounds dataset with known 3D geometry for quantitative evaluation, (ii) a high-resolution phantom wound dataset captured under controlled conditions, and (iii) a new clinical dataset featuring real-world wound recordings. Each dataset uniquely validates the framework’s components, gradually increasing scene complexity at development stages. We also define a set of metrics aligned with clinical wound documentation practices to evaluate the framework’s performance across the datasets. II-A Digital wounds dataset Synthetic datasets offer precise control over scene complexity, simulating various lighting conditions, skin tones, and camera trajectories. Direct access to ground truth geometry and segmentation enables systematic testing of reconstruction accuracy and robustness under idealised conditions. In this direction, we use a synthetic dataset combining 3D human body avatars, real wound textures, and parametrically defined wound geometries. For detailed methodology and results, we refer the reader to [ 23 ] . II-B Silicone wounds dataset Silicone phantoms are realistic replicas of clinical wounds, developed by companies such as TraumaSIM 2 2 2 https://traumasim.com.au/ to assist the training of healthcare professionals. We simulate a typical clinical scenario where an operator acquires handheld recordings in a controlled environment, introducing practical challenges such as multi-light reflections and motion blur. This allows us to evaluate the framework’s robustness to variations in camera quality and capture conditions, not present in the synthetic

Read the complete text