基于混合现实与人工智能算法在股前外穿支血管定位中的前瞻性研究

Prospective study on the localization of anterolateral thigh perforator vessel based on mixed reality and artificial algorithm

目的构建混合现实技术与人工智能算法的系统,评估其在股前外穿支皮瓣手术中定位血管的效果,为临床实践提供新的思路。方法选取20例行股前外穿支皮瓣修复的患者,在下肢黏贴定位装置后行CT血管造影(CTA)扫描,将所得二维数据制成定位装置与血管等的三维模型,并通过混合现实技术实现穿支血管三维可视化。在HoloLens 2中开发人工算法,术中使用人工智能算法自动匹配定位装置与其三维模型以实现穿支血管与其三维模型的重叠。记录制取皮瓣区域内定位的穿支血管数量及术中探查的实际数量,以此计算基于CTA数据重建的血管识别正确率;测量系统定位的穿支血管穿出点与实际血管穿出点之间的距离,计算误差值;记录系统辅助下制取股前外穿支皮瓣的手术时间,并与常规方法制取皮瓣的手术时间进行对比分析。探讨该系统的临床实用性。结果CTA数据重建的穿支血管30个,手术中探查发现实际存在32个,识别正确率为93.75%;系统定位的穿支血管穿出点与实际穿出点的平均距离为(1.65±0.52)mm,系统定位下制取皮瓣的平均手术时间为(43.45±4.6)min,常规方法制取皮瓣的手术时间为(57.6±7.9)min。所有穿支皮瓣均成活,1例皮瓣术后7 d出现感染,1例皮瓣局部坏死,经对症治疗后延迟愈合。结论本研究搭建的系统可以通过混合现实技术实现穿支血管三维可视化,利用人工智能算法提高穿支血管定位精度,在穿支皮瓣制取手术中具有应用潜力。

Objective This paper aims to construct a system integrating mixed reality technology with artificial algorithm and to evaluate its effectiveness in vascular localization during anterolateral thigh perforator flap surgery to provide new insights for clinical practice.Methods Twenty patients undergoing anterolateral thigh perforator flap repair were selected.After attaching positioning devices on the lower limb,CT angiography(CTA)scans were performed.The 2D data obtained were converted into a 3D model of the positioning device and vessels.Mixed reality technology was utilized to achieve 3D visualization of perforator vessels.An artificial algorithm was developed in HoloLens 2 to match the positioning device automatically with its 3D model intraoperatively to overlap the perforator vessels with their 3D models.The number of perforator vessels identified within the flap harvesting area and the actual number detected during surgery were recorded to calculate the accuracy rate of vessel identification based on CTA data reconstruction.The distance between the perforator vessel exit points located by the system and the actual exit points was measured,and the error values were calculated.The surgical time required for the system to harvest the anterolateral thigh perforator flap was documented and compared with the surgical time required by conventional methods.The clinical applicability of the system was discussed.Results The CTA data reconstruction identified 30 perforator vessels,while the actual number found during surgery was 32,resulting in an identification accuracy rate of 93.75%.The average distance between the perforator vessel exit points located by the system and the actual exit points was(1.65±0.52)mm.The average surgical time for flap harvesting with the assistance of the system was(43.45±4.6)min compared with(57.6±7.9)min required by conventional methods.All perforator flaps survived the procedure.One case of flap infection occurred seven days postoperatively,and one case of partial flap necrosis was treated with symptomatic therapy,resulting in delayed healing.Conclusion The system constructed in this paper can achieve 3D visualization of perforator vessels through mixed reality technology and improve the accuracy of perforator vessel localization using artificial algorithms,hence demonstrating potential application in anterolateral thigh perforator flap harvesting surgeries.