| 骆姝君,蒋晨琳,宋哲明,李梦娇,樊宇航,顾静军,詹强.多级网络体系结构人体关键点识别技术在脊柱侧弯筛查中的验证研究[J].浙江中西医结合杂志,2025,35(12): |
| 多级网络体系结构人体关键点识别技术在脊柱侧弯筛查中的验证研究 |
| Validation of RSN Human Key Point Recognition Technology in Scoliosis Screening |
| 投稿时间：2025-07-08 修订日期：2025-11-05 |
| DOI： |
| 中文关键词: 儿童 青少年 脊柱侧凸 体表形态 机器学习 人工智能 脊柱侧弯 |
| 英文关键词:Children Adolescents Scoliosis Trunk surface Machine Learning Artificial Intelligence |
| 基金项目: |
|
| 摘要点击次数: 91 |
| 全文下载次数: 0 |
| 中文摘要: |
| 目的：本研究旨在验证多级网络体系结构(Residual Steps Network,RSN)人体关键点识别技术辅助诊断脊柱侧弯的准确性,为脊柱侧弯筛查提供一种标准化、无辐射,可便携的手段；方法：选取2022年4月至2023年12月期间在杭州市儿童医院就诊的6至18岁儿童青少年,其中非脊柱侧弯 138例,脊柱侧弯 237例,通过将RSN人体关键点识别技术识别的背部图像结果与全脊柱正位X线诊断金标准科布氏角(Cobb’s Angle, Cobb角)进行对比,建立受试者操作特征曲线(Receiver Operating Characteristic,ROC),判断RSN关键点识别技术对筛查脊柱侧弯的应用价值。结果：RSN人体关键点识别技术识别结果与X线诊断金标准cobb角对比ROC分析显示曲线下面积为0.632(P值<0.05),95%置信区间从0.572-0.689,敏感性为74.47%,特异性为71.80%。在对软件识别结果、cobb角诊断结局的差异性分析显示二者均在年龄的差异具有统计学意义(P值<0.05)。结论：RSN关键点识别技术有较高的诊断效能,尤其对于大年龄段儿童,可以为脊柱侧弯筛查提供标准化、无辐射,可便携的手段。 |
| 英文摘要: |
| Objective: In this study, we used the multi-level network structure ( Residual Steps Network, RSN ) human key point recognition technology to analyze human back photos to assist in the diagnosis of idiopathic scoliosis ( Idiopathic Scoliosis, IS ), and evaluate the accuracy and effectiveness of the technology. The aim is to provide a standardized, non-radiative, portable means for its screening ; Methods: Children and adolescents aged 6 to 18 years old who were treated in Hangzhou Children "s Hospital and the physical examination center of the hospital from April 2022 to December 2023 were selected, including 138 cases of non-IS and 237 cases of IS. By comparing the back image results identified by RSN key point recognition technology with the gold standard Cobb "s angle ( Cobb "s Angle ) of whole spine anteroposterior X-ray diagnosis, the receiver operating characteristic ROC curve was established to judge the application value of RSN key point recognition technology in screening IS. Results: The ROC analysis showed that the area under the curve was 0.632 ( P < 0.05 ), the 95 % confidence interval was from 0.572 to 0.689, the sensitivity was 74.47 %, and the specificity was 71.80 %. The difference analysis of software identification results and cobb angle diagnosis results showed that the difference between the two was statistically significant in age ( P < 0.05 ). Conclusion: RSN key point identification technology has high diagnostic efficiency, especially for older children, which can provide standardized, non-radiative and portable means for IS screening. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
