基于 YOLO 神经网络构建压力性损伤自动检测和分期的人工智能模型
onstruction of an Artificial Intelligence-assisted System for Automatic Detection of Pressure Injuries Based on the YOLO Neural Network
背景 随着人口老龄化，压力性损伤（PI）的发病率逐渐增加，这不仅严重影响了患者的生存质量，还增加了医保支出。然而，PI的早期发现和准确分期极大地依赖于专业培训。目的 构建并测试一个用于PI自动检测和分期的人工智能模型，目的是提高PI诊断的实时性、准确性和客观性。方法 从常熟市第一人民医院压疮电子化管理系统中选取了2021年1月—2024年2月期间的693张PI图片，并按照2019NPUAP指南分为6期，包括：Ⅰ期154张、Ⅱ期188张、Ⅲ期160张、Ⅳ期82张、不可分期52张、深部组织损伤期57张。利用基于5种不同版本大小的YOLOv8神经网络和迁移学习，建立针对PI的深度学习目标检测模型。模型评价指标包括精确度、准确率、灵敏度、特异度及检测速度等。最后，通过UltralyticsHub平台将模型部署到手机应用程序（App）中，实现AI模型在临床工作中的应用。结果 在对包含142张PI图像的测试集进行评估时，YOLOv8l版本在确保高精确度（0.827）的同时，也展现了较快的推理速度（68.49fps），与其他YOLO版本相比，在精确度与速度之间取得了最佳的平衡。具体而言，其在所有类别上的整体准确率为93.18%，灵敏度为76.52%，特异度为96.29%，假阳性率为3.72%。在6个PI分期中，模型对“Ⅰ期”的准确率最高，达到95.97%；Ⅱ期、Ⅲ期、Ⅳ期、深部组织损伤期、不可分期，分别取得了91.28%、91.28%、91.95%、95.30%和93.29%的准确率。就处理速度而言，平均每秒可处理68.49张PI图像。结论 基于YOLOv8l网络的AI模型能够快速、准确地检测和分期PI。将该模型部署到手机App中，能够在临床实践中便携使用，具有很大的临床应用潜力。
Background With the aging populationthe incidence of pressure injuriesPIis gradually increasing. This not only severely impacts the quality of life for patients but also increases healthcare expenditures. Howeverthe early detection and accurate staging of PI heavily depend on specialized training. Objective To construct and validate an artificial intelligence model for the automatic detection and staging of pressure injuriesPIaimed at enhancing the real-time natureaccuracyand objectivity of PI diagnostics. Methods A total of 693 pressure injury images from the electronic management system of pressure ulcers at Changshu City First People's Hospital were selected from January 2021 to February 2024categorized into six stages according to guidelinesStage 154 imagesStage 188 imagesStage 160 imagesStage 82 imagesunstageable52 imagesand deep tissue injury57 images. A deep learning object detection model for PI was established using five different versions of the YOLOv8 neural network and transfer learning. The model evaluation metrics included accuracysensitivityspecificityfalse positive rateand detection speed. Finallythe model was deployed to a mobile application via the Ultralytics Hub platformfacilitating the application of the AI model in clinical practice.Results During the evaluation of a test set containing 142 PI imagesthe YOLOv8l version demonstrated high accuracy0.827 and fast inference speed68.49fpsachieving the best balance between precision and speed among the YOLO versions. Specificallyit achieved an overall accuracy of 93.18% across all categoriesa sensitivity of 76.52%a specificity of 96.29%and a false positive rate of 3.72%. Among the six stages of PIthe model achieved the highest accuracy for Stage at 95.97%.The accuracies for Stage Stage Stage deep tissue injuryand unstageable were 91.28%91.28%91.95%95.30%and 93.29%respectively. In terms of processing speedYOLOv8l took a total of 2.07 seconds to process 142 imagesaveraging 68.49 PI images per second. Conclusion The AI model based on the YOLOv8l network can quickly and accurately detect and stage PI. Deploying this model to a mobile app allows for portable use in clinical practicedemonstrating significant potential for clinical application.
顾丽华、徐晓丹、夏开建、须月萍、王珍妮
10.12114/j.issn.1007-9572.2024.0168
医药卫生理论医学研究方法自动化技术、自动化技术设备
压力性损伤人工智能深度学习YOLO目标检测神经网络模型pp
顾丽华,徐晓丹,夏开建,须月萍,王珍妮.基于 YOLO 神经网络构建压力性损伤自动检测和分期的人工智能模型[EB/OL].(2024-07-31)[2025-08-16].https://chinaxiv.org/abs/202407.00371.点此复制
评论