(WHS-P71) Review and Evaluation of AI-Based Algorithms for Wound Assessment and Decision Support
Friday, May 17, 2024
7:30 AM – 5:00 PM East Coast USA Time
Chronic wounds are a prevalent global health concern, annually affecting 6.5 million Americans with $25 billion burden on healthcare systems. Accurate and timely wound assessment is crucial for effective wound management, enabling clinicians to monitor healing progress, identify complications, and optimize treatment strategies. However, traditional wound assessment methods and manual measurements are often subjective, qualitative, time-consuming, and prone to errors, leading to suboptimal wound care outcomes and prolonged healing times. Artificial intelligence (AI) based wound assessment tools have emerged as promising solutions to address these challenges. These tools utilize machine learning algorithms like deep learning and convolutional neural networks to analyze wound images and provide objective, quantitative assessments of wound characteristics, such as wound area, tissue type, and healing status. A growing body of research has explored the development and application of AI-based wound assessment tools. The objective of this work is to review the current landscape of AI-based wound assessment tools and decision support systems, encompassing both smartphone apps and research papers, aiming at assessing their accuracy, identifying knowledge gaps, and informing future research directions in this field. We assessed performance metrics, input information, AI methods employed, accuracy & speed of assessment, level of user intervention, dataset size and quality, and clinical evaluation. We found that most of these studies focused on wound documentation and development of image-based wound area measurement algorithms like thresholding, while more recent studies focused on developing algorithms using convolutional neural network and region-based segmentation for tissue classification, with relatively small datasets ranging from 80 to 600 wound images. While most of the studies utilized wound pictures taken by smartphones, more recent publications focus on measuring the wound depths using RGB-D cameras and other medical imaging devices. A few works delved into developing decision support systems, aiming at providing clinicians with enhanced decision-making capabilities by leveraging AI algorithms to analyze complex wound data and offer personalized recommendations. Our findings also revealed that the current body of literature and existing applications suffer from the absence of reliable datasets, inaccurate measurements, and need for operator intervention. Additionally, there exists a knowledge gap in volumetric wound assessment and decision support systems. As the body of literature expands and technology progresses, there is a growing need for AI-based wound assessment tools to transition from research labs to clinical applications.