人工智能混合现实技术在特殊类型复杂性上尿路结石经皮肾镜手术规划中的应用

Application of AI-MR in the planning of PCNL for special types of complex upper urinary stones

目的 评价人工智能混合现实技术(AI-MR)在规划特殊类型的复杂性上尿路结石行超声引导下经皮肾镜碎石术(PCNL)中的应用价值。方法 该项目为前瞻性单中心单臂临床研究,纳入2022年8月—2023年5月于清华大学附属北京清华长庚医院收治的行超声引导下PCNL的15例特殊类型复杂性上尿路结石患者,其中男性9例、女性6例,包括盆腔异位肾结石3例、马蹄肾结石5例、脊柱畸形合并肾结石3例、移植肾结石4例。基于患者术前计算机断层扫描尿路造影(CTU)数据进行数字三维重建,并使用AI-MR在真实空间中投射手术相关的三维影像,获得术区的“透视化”信息,从而进行术前设计与规划,内容主要包括目标肾盏、通道数目及辅助措施。并对目标肾盏及通道数的符合度、一期净石率、总手术时间、主皮肾通道建立时间、血红蛋白下降量、手术并发症、术后住院时间等情况进行分析。结果 15例患者均使用AI-MR进行术前规划,顺利完成一期超声引导下PCNL。基于术前规划,我们采用了单一标准通道经皮肾镜碎石术(S-PCNL)以及S-PCNL联合针状肾镜(Needle-perc)或顺/逆行输尿管软镜(FURS)的手术方式,其中4例患者行单通道S-PCNL,3例行多通道S-PCNL,8例行S-PCNL联合Needle-perc或FURS。目标肾盏及通道数术前规划和术中实施的符合率为86.7%,一期净石率80.0%,平均主皮肾通道建立时间(2.3±0.3)min、平均总手术时间(61.5±12.2)min、血红蛋白下降值(9.6±1.2)g/L,平均术后住院时间(4.6±0.5)d。无输血、脏器损伤和尿源性脓毒症等Clavien-Dindo≥Ⅱ级并发症发生。结论 AI-MR可量化分析特殊类型复杂性上尿路结石患者的术前影像数据,实现术区三维透视化效果,利于制定手术方案,优化穿刺路径,有效规避周围脏器损伤风险,达到减少并发症、缩短治疗周期以及提高一期净石率的效果。

Objective To evaluate the application value of artificial intelligence mixed reality(AI-MR)technology in the planning of ultrasound-guided percutaneous nephrolithotomy(PCNL)for special types of complex upper urinary stones.Methods The prospective single-center,single-arm clinical study involved 15 patients with complex upper urinary stones undergoing ultrasound-guided PCNL during Aug.2022 and May 2023,including 9 male and 6 female,3 cases of pelvic ectopic kidney stones,5 cases of horseshoe kidney stones,3 cases of renal stones combined with spinal deformity,and 4 cases of transplant kidney stones.Based on preoperative computed tomography urography(CTU)data,digital three-dimensional reconstruction was performed,and Al-MR was used to project surgery-related three-dimensional images in real space to obtain"perspective"information of the surgical area.This facilitated preoperative design and planning,including target calyx,number of channels,and auxiliary measures.The compliance of target calyx and number of channels,stone clearance rate,total operation time,time required to establish the percutaneous renal channel,decrease in hemoglobin level,surgical complications,and postoperative hospital stay were analyzed.Results All 15 patients underwent preoperative planning using AI-MR and successfully completed one-stage ultrasound-guided PCNL.Based on the preoperative planning,we utilized S-PCNL alone or combined with Needle-perc or antegrade/retrograde FURS/RIRS.Among all patients,4 underwent single-channel S-PCNL,3 multi-channel S-PCNL,and 8 S-PCNL combined with Needle-perc or FURS.The compliance of target calyx and number of channels was 86.7%,the one-stage stone clearance rate was 80.0%,the average time for establishing the channel was(2.3±0.3)minutes,the average total operation time was(61.5±12.2)minutes,the mean decrease in hemoglobin level was(9.6±1.2)g/L,and the average postoperative hospital stay was(4.6±0.5)days.There were no Clavien-Dindo grade≥Ⅱ complications,such as blood transfusion,organ injury,or urosepsis.Conclusion Before surgery,AI-MR can be used to quantitatively analyze imaging data for patients with special types of complex upper urinary stones,which can achieve three-dimensional fluoroscopy effects,formulate surgical plans,optimize puncture paths,effectively avoid the risk of damage to surrounding organs,reduce complications,shorten treatment cycle and improve the first-stage stone clearance rate.