SCH:Smartphone Wound Image Parameter Analysis and Decision Support in Mobile Env

项目总结(见说明)： 在美国，慢性伤口影响着650万名患者，估计治疗成本为250亿美元。我们的团队建议进行研究，以推进我们现有的由NSF资助的智能手机伤口分析系统，该系统帮助患者监控他们的糖尿病足部溃疡，为他们提供关于愈合进度的即时反馈。我们的伤口系统分析患者伤口的智能手机图像，检测伤口区域和组织成分，并通过将当前图像与过去的图像进行比较来生成专有的愈合评分。我们设想的慢性伤口评估系统将支持护理团队在探访患者时做出的循证决策，并将伤口护理推向数字客观性。我们将数字客观性定义为从图像中自主提取的伤口评估指标的综合，以生成客观的可操作的反馈，使未受过伤口专家培训的临床医生能够提供“标准化的伤口护理”。数字客观性与目前基于医生经验对伤口进行主观、视觉检查的做法形成了鲜明对比。第一个目标是开发图像处理算法，以减少在一些临床或家庭环境中，以及从任意相机角度和距离拍摄伤口时，由于不理想的照明而导致的伤口分析误差。虽然我们之前的伤口系统在理想条件下工作良好，但不理想的光线会导致很大的误差，在极端情况下，健康皮肤被检测到作为伤口区域。第二个目标是扩展我们现有的伤口分析系统，该系统只针对糖尿病伤口，以处理动脉、静脉和压力溃疡，扩大了潜在用户。第三个目标将综合算法，这些算法自动生成可操作的伤口决策规则，这些规则是从实际伤口临床医生所做的决策中学习的。这项研究是伍斯特理工学院(WPI)(图像处理、机器学习和智能手机编程方面的技术专长)和马萨诸塞大学医学院(UMMS)(创伤临床专业知识和创伤患者招募)联合开展的，以验证我们的工作。

PROJECT SUMMARY (See instructions): Chronic wounds affect 6.5 million patients in the U.S., with an estimated treatment cost of $25 billion. Our team proposes research to advance our existing NSF-funded smartphone wound analysis system, which helps patients monitor their diabetic foot ulcers, providing them with instant feedback on healing progress. Our wound system analyzes a smartphone image of the patients' wound, detects the wound area and tissue composition, and generates a proprietary healing score by comparing the current image with a past image. Our envisioned chronic wound assessment system will support evidence-based decisions by the care team while visiting patients, and move wound care toward digital objectivity. We define digital objectivity as the synthesis of wound assessment metrics that are extracted autonomously from images in order to generate objective actionable feedback, enabling clinicians not trained as wound specialists to deliver "standardized wound care". Digital objectivity contrasts with the current practice of subjective, visual inspection of wounds based on physician experience. The first aim will develop image processing algorithms to mitigate wound analysis errors caused by non-ideal lighting in some clinical or home settings, and when the wound is photographed from arbitrary camera angles and distance. While our previous wound system worked well in ideal conditions, non-ideal lighting caused large errors and healthy skin was detected as the wound area in extreme cases. The second aim extends our existing wound analysis system that targets only diabetic wounds to handle arterial, venous and pressure ulcers, expanding the potential user. The third aim will synthesize algorithms that autonomously generate actionable wound decision rules that are learned from decisions taken by actual wound clinicians. This research is joint work of Worcester Polytechnic Institute (WPI) (technical expertise in image processing, machine learning and smartphone programming) and University of Massachusetts Medical School (UMMS) (clinical expertise on wounds, and wound patient recruitment to validate our work)