评估一种新型自动扫查超声机器人在肾脏检查中的应用价值。

连续纳入2025年3月至2025年5月在复旦大学附属中山医院厦门医院拟行泌尿系统超声检查的患者。所有患者依次接受自动超声机器人扫查和超声医师人工检查。肾长径测量值采用配对t检验分析；检查耗时及影像质量评分不符合正态分布或为等级资料，采用Wilcoxon符号秩检验分析；采用Kappa检验和组内相关系数（ICC）分别评估两种方法在疾病分类诊断及肾长径测量方面的一致性，使用Spearman相关分析评估影像质量评分之间的相关性；以P＜0.05为差异有统计学意义。

研究共纳入52例患者（103个肾脏）。所有患者均顺利完成两种检查。问卷调查显示患者对机器人检查的体验和接受度良好。机器人测量的肾长径[（103.78±8.13） mm]略短于人工测量结果[（108.07±8.58） mm]，平均差值为-4.3 mm（95%CI: -5.56~-3.02，t=-6.69，P＜0.01）。机器人扫查耗时显著长于人工检查差异有统计学意义[（101.84±41.42）s vs（92.84±85.65）s，Z=-3.26，P＜0.01]。机器人获取的静态图像质量评分低于人工检查[中位数：4（4，5） vs 5（4，5），Z=-5.892，P＜0.001]，而二者动态视频影像评分比较无显著统计学意义[中位数：4（4，5） vs 4（4，5），Z=-0.45，P=0.653]。在影像诊断方面，二者在多数肾脏疾病诊断中具有高度一致性（0.62≤κ≤1.00），95%CI下限均大于0。亚组分析显示，机器人检查病灶检出效能存在解剖位置差异，在肾上、下极的敏感度（81.25%、77.27%）低于中部（100.00%），而人工检查在各部位均保持较高水平（93.75%~100.00%）。

该自动超声机器人系统具有良好的患者接受度，在肾长径测量、检查耗时、影像质量及病变检出方面均接近人工扫查水平，在推动肾脏超声检查自动化、标准化及远程化方面展现出潜力，有望成为超声医师的可靠辅助工具。

To evaluate the feasibility and effectiveness of a novel automated scanning ultrasound robot in renal examination.

Consecutive patients scheduled for urinary system ultrasound examinations at Zhongshan Hospital (Xiamen), Fudan University, from March to May 2025 were enrolled. All participants underwent both automated scanning with the ultrasound robot and conventional manual examination by a sonographer. The paired t-test was used to analyze renal length measurements, which conformed to a normal distribution; the Wilcoxon signed-rank test was used to analyze examination time and image quality scores, which were not normally distributed or were ordinal data. Kappa analysis and the intraclass correlation coefficient (ICC) were employed to assess the agreement between the two methods in disease classification and renal length measurements, respectively. Spearman correlation analysis was used to evaluate the correlation between image quality scores. A P-value＜0.05 was considered statistically significant.

A total of 52 patients (103 kidneys) were enrolled, and all patients successfully completed both examinations. Questionnaire survey demonstrated favorable patient acceptance and experience with the robotic examination. The renal length measured by the robot (103.78±8.13) mm was significantly shorter than that measured manually (108.07±8.58) mm, with a mean difference of −4.3 mm (95%CI: −5.56~−3.02, t=−6.69, P＜0.01). The robotic examination required significantly longer scanning time compared to manual examination [(101.84±41.42) s vs (92.84±85.65) s, Z=−3.26, P＜0.01]. Static image quality scores obtained by the robot were lower than those of manual examination [median: 4 (4, 5) vs 5 (4, 5), Z=−5.892, P＜0.001], whereas no significant difference was observed in dynamic video image quality scores between the two methods [median: 4 (4, 5) vs 4 (4, 5), Z=−0.45, P=0.653]. For imaging diagnosis, the two methods showed high agreement in the diagnosis of most renal diseases (0.62≤κ≤1.00), with the lower limits of the 95% CIs all exceeding 0. Subgroup analysis demonstrated that the lesion detection efficacy of the robotic examination exhibited anatomic location-dependent differences, with lower sensitivity at the upper and lower poles (81.25% and 77.27%) compared to the mid-kidney (100.00%), whereas manual examination maintained consistently high sensitivity across all locations (93.75%~100.00%).

The automated ultrasound robot system demonstrates favorable patient acceptability and performs comparably to manual examination in terms of renal length measurement, examination time, image quality, and lesion detection. It shows promise in promoting automation, standardization, and tele-ultrasonography of renal ultrasound examinations, and may serve as a reliable adjunctive tool for ultrasonographers.