| 摘要: |
| 目的? 利用深度学习算法构建肩袖损伤程度肌骨超声智能鉴别系统，并验证其应用效果。方法? 回顾性分析2022年8月~2023年8月医院收治的1064例肩袖损伤患者临床资料，作为训练集，收集其肌骨超声检查图像与参数，利用深度学习算法构建肩袖损伤程度肌骨超声智能鉴别系统；另回顾性分析2023年9月~2024年9月医院收治的108例肩袖损伤患者临床资料作为验证集，以肩关节镜检查结果为金标准，分析肩袖损伤程度肌骨超声智能鉴别系统对肩袖损伤程度的诊断效能。结果? 肩袖损伤程度肌骨超声智能鉴别系统对肩袖损伤Ⅰ级、Ⅱ级、Ⅲ级的诊断灵敏度（96.55%、95.56%、97.06%）、特异度（100.00%、96.83%、97.30%）和准确度（99.07%、96.30%、97.22%）均高于临床医师（灵敏度:89.66%、93.33%、94.12%，特异度:98.73%、93.65%、95.95%，准确度:97.22%、93.52%、95.37%），肩袖损伤程度肌骨超声智能鉴别系统诊断肩袖损伤分级与关节镜具有高度一致性（Kappa=0.976、0.924、0.936，P＜0.001）。结论? 利用深度学习算法构建的肩袖损伤程度肌骨超声智能鉴别系统对肩袖损伤分级的诊断效能佳，具有良好的临床应用价值。 |
| 关键词: |
| DOI： |
| 分类号: |
| 基金项目: |
|
|
|
Shao Yunqing1,2,3
|
|
1.Rugao People'2.'3.s Hospital
|
| Abstract: |
| Objective??? To construct an intelligent musculoskeletal ultrasonic identification system for rotator cuff injury degree by deep learning algorithm, and to verify its application effect . Methods??? The clinical data of 1064 patients with rotator cuff injury admitted to the hospital from August 2022 to August 2023 were analyzed retrospectively and they were be used as a training set, and the images and parameters of musculoskeletal ultrasound examination were collected, and an musculoskeletal ultrasonic intelligent identification system for rotator cuff injury degree was constructed by using deep learning algorithm. In addition, the clinical data of 108 patients with rotator cuff injury admitted to the hospital from September 2023 to September 2024 were retrospectively analyzed and they were be used as the verification set, and the results of shoulder arthroscopy were used as the gold standard, and the diagnostic efficiency of musculoskeletal ultrasonic intelligent identification system for rotator cuff injury degree was analyzed. Results: The diagnostic sensitivity (96.55%, 95.56%, 97.06%), specificity (100.00%, 96.83%, 97.30%) and accuracy (99.07%, 96.30%, 97.22%) of the musculoskeletal ultrasonic intelligent identification system for rotator cuff injury grade were higher than those of musculoskeletal ultrasound (sensitivity: 89.66%, 93.33%, 94.12%, specificity: 98.73%, 93.65%, 95.95%, accuracy: 97.22%, 93.52%, 95.37%), and the classification of rotator cuff injury diagnosed by musculoskeletal ultrasonic intelligent identification system was highly consistent with arthroscopy (Kappa=0.976, 0.924, 0.936, P < 0.001). Conclusion: The musculoskeletal ultrasonic intelligent identification system based on deep learning algorithm has good diagnostic efficiency for rotator cuff injury classification and has good clinical application value. |
| Key words: |
