[Paper Review] Image Based Artificial Intelligence in Wound Assessment: A Systematic Review

DM Anisuzzaman, Chuanbo Wang+4 more|arXiv (Cornell University)|Sep 15, 2020

Wound Healing and Treatments109 references114 citations

TL;DR

This systematic review evaluates image-based artificial intelligence (AI) for wound assessment, focusing on deep learning techniques for wound segmentation, classification, and measurement. It synthesizes 115 studies on AI-driven wound systems, highlighting advances in mobile apps and multimodal imaging, and identifies key challenges in standardization, data quality, and clinical integration for improved wound care outcomes.

ABSTRACT

Efficient and effective assessment of acute and chronic wounds can help wound care teams in clinical practice to greatly improve wound diagnosis, optimize treatment plans, ease the workload and achieve health related quality of life to the patient population. While artificial intelligence (AI) has found wide applications in health-related sciences and technology, AI-based systems remain to be developed clinically and computationally for high-quality wound care. To this end, we have carried out a systematic review of intelligent image-based data analysis and system developments for wound assessment. Specifically, we provide an extensive review of research methods on wound measurement (segmentation) and wound diagnosis (classification). We also reviewed recent work on wound assessment systems (including hardware, software, and mobile apps). More than 250 articles were retrieved from various publication databases and online resources, and 115 of them were carefully selected to cover the breadth and depth of most recent and relevant work to convey the current review to its fulfillment.

Motivation & Objective

• To systematically review computational and AI-based methods for image analysis in wound assessment.
• To evaluate the state-of-the-art in AI-driven wound measurement (segmentation) and diagnosis (classification).
• To analyze the design, functionality, and clinical applicability of existing wound assessment systems, including mobile apps and hardware.
• To identify gaps in data standardization, multimodal integration, and real-world clinical deployment of AI wound systems.

Proposed method

• Conducted a systematic literature search across Scopus, Web of Science, PubMed, IEEE Xplore, ScienceDirect, and ERIC from 1994 to 2020.
• Screened over 250 articles and selected 115 for in-depth review based on relevance to image-based AI in wound assessment.
• Categorized studies by methodology: rule-based, traditional machine learning (e.g., SVM, random forest), and deep learning (e.g., CNNs, U-Net variants).
• Reviewed wound measurement techniques using 2D RGB images, 3D surface models, and thermal imaging.
• Evaluated mobile applications for wound capture, segmentation, size estimation, and tissue color classification.
• Assessed multimodal systems integrating 2D/3D images, thermal, hyperspectral, and physiological sensors for comprehensive wound analysis.

Experimental results

Research questions

• RQ1 What are the current trends and limitations in AI-based wound segmentation and classification using medical images?
• RQ2 How do existing mobile and software-based wound assessment systems perform in clinical and real-world settings?
• RQ3 What role do multimodal data (e.g., RGB, 3D, thermal, EHR) play in improving wound diagnosis and prognosis?
• RQ4 What are the key technical and clinical challenges in deploying AI-powered wound care systems?
• RQ5 How can public, well-labeled, multimodal wound datasets support future AI development in wound care?

Key findings

• Deep learning models, particularly U-Net and its variants, outperformed traditional machine learning and rule-based methods in wound segmentation accuracy.
• Mobile applications such as MOWA and FootSnap improved standardization in wound image capture and enabled basic measurements like size and tissue color.
• Multimodal systems combining RGB, 3D, thermal, and physiological data showed promise for comprehensive wound assessment but remain underdeveloped clinically.
• Despite progress, significant challenges remain in lighting variability, background noise, and image quality affecting AI performance.
• No existing system fully integrates all modalities (images, 3D, text, EHR), and public, large-scale, well-labeled datasets are still lacking.
• The review identifies a critical need for standardized, publicly available datasets and clinical validation to advance AI in wound care.