基于神经网络辐射场的人工智能技术在下肢软组织缺损修复中的应用  

Application of artificial intelligence based on neural network radiation field in repair of soft tissue defects at lower limbs

作　　者：吴飞[1] 刘丰[1] 孙志博 肖文霞[1] 罗文娜[1] 马勘 杨越[1] Wu Fei;Liu Feng;Sun Zhibo;Xiao Wenxia;Luo Wenna;Ma Kan;Yang Yue(Department of Orthopaedic Surgery,Renmin Hospital of Wuhan University,Wuhan 430060,China)

机构地区：[1]武汉大学人民医院骨外科,武汉430060

出　　处：《中华创伤骨科杂志》2023年第3期213-218,共6页Chinese Journal of Orthopaedic Trauma

基　　金：武汉大学自然科学自主科研项目(413000640)。

摘　　要：目的探讨基于神经网络辐射场的人工智能技术在下肢软组织缺损修复中的应用效果。方法回顾性分析武汉大学人民医院骨外科2020年6月至2022年5月期间收治的23例下肢软组织缺损患者,男14例,女9例;年龄为(38.6±6.7)岁。软组织缺损原因:交通伤9例,软组织良性或恶性原发肿瘤6例,机械伤4例,压砸伤2例,慢性溃疡2例;缺损部位:大腿3例,小腿7例,足踝部以远13例;软组织缺损面积为6.0 cm×3.8 cm~14.7 cm×12.8 cm。在基于神经网络辐射场的人工智能技术(在辐射场下通过体渲染快速进行三维模型图像构建并进行处理的前沿人工智能算法)辅助下完成股前外侧游离皮瓣(7例)、带蒂皮瓣(16例)移植重建缺损。记录患者的皮瓣成活率、治疗前后美观满意度、皮瓣切取移植时间、下肢功能恢复及并发症发生情况等。结果23例患者术后获32(28,36)周随访。所有患者的皮瓣均切取顺利并成活,皮瓣切取移植时间为65.8(50.0,76.0)min。患者治疗前美观满意度视觉模拟评分为(2.3±0.7)分,治疗后4周为(8.4±1.6)分,差异有统计学意义(P<0.05)。除1例患者因静脉危象而发生皮瓣远端浅表坏死(痂下愈合)外,其余患者皮瓣均愈合良好,未见血肿、感染等并发症发生。治疗后12周下肢功能恢复根据Enneking评价系统评定:评分为23.7(22.0,25.0)分,平均恢复肢体功能79%(23.7/30.0)。结论基于神经网络辐射场的人工智能技术在下肢软组织缺损皮瓣移植修复中的应用效果理想,具有手术快速精准、供区损伤小、学习曲线短等优点。Objective To investigate the application of artificial intelligence based on the neural network radiation field in repair of soft tissue defects at lower limbs.Methods A retrospective analysis was performed of the 23 patients who had been admitted to Department of Orthopedic Surgery,Renmin Hospital of Wuhan University from June 2020 to May 2022 for soft tissue defects at lower limbs.There were 14 males and 9 females,aged(38.6±6.7)years.Causes for soft tissue defects:traffic injury in 9 cases,benign or malignant primary soft tissue tumor in 6 cases,mechanical injury in 4 cases,crush injury in 2 cases,and chronic ulcer in 2 cases.Defect locations:the thigh in 3 cases,the lower leg in 7 cases,and the ankle and distal foot in 13 cases.The areas of soft tissue defect ranged from 6.0 cm×3.8 cm to 14.7 cm×12.8 cm.The defects were repaired and reconstructed by transplantation of an anterolateral femoral free flap in 7 cases and a pedicled flap in 16 cases with the assistance of artificial intelligence based on the neural network radiation field,a cutting-edge artificial intelligence algorithm that can quickly construct and process three-dimensional model images through volume rendering under the radiation field.The flap survival rate,aesthetic satisfaction before and after treatment,time for skin flap harvesting and transplantation,functional recovery of lower limbs and incidence of complications were recorded.Results All the 23 patients were followed up for 32(28,36)weeks.All the flaps were harvested smoothly and survived.The time for flap harvesting and transplantation was 65.8(50.0,76.0)min.The aesthetic satisfaction scored(2.3±0.7)points before treatment and(8.4±1.6)points 4 weeks after treatment,showing a statistically significant difference(P<0.05).The skin flaps healed well with no complications such as hematoma or infection in all but one patient who suffered from superficial necrosis at the distal skin flap due to venous crisis but healed with a scar.On average,the functional recovery of lower limbs scored

关 键 词：软组织损伤 下肢 外科皮瓣 组织移植 人工智能 神经网络辐射场 

分 类 号：R658.3[医药卫生—外科学]