Department of Radiology, Mayo Clinic, 200 First Street SW, Rochester, MN, 55905, USA.
Department of Laboratory Medicine and Pathology, Division of Nephrology and Hypertension, Mayo Clinic, Rochester, MN, 55905, USA.
Emerg Radiol. 2021 Aug;28(4):781-788. doi: 10.1007/s10140-021-01915-4. Epub 2021 Mar 1.
To evaluate the ability of a semi-automated radiomic analysis software in predicting the likelihood of spontaneous passage of urinary stones compared with manual measurements.
Symptomatic patients visiting the emergency department with suspected stones in either kidney or ureters who underwent a CT scan were included. Patients were followed for up to 6 months for the outcome of a trial of passage. Maximum stone diameters in axial and coronal images were measured manually. Stone length, width, height, max diameter, volume, the mean and standard deviation of the Hounsfield units, and morphologic features were also measured using automated radiomic analysis software. Multivariate models were developed using these data to predict subsequent spontaneous stone passage, with results expressed as the area under a receiver operating curve (AUC).
One hundred eighty-four patients (69 females) with a median age of 56 years were included. Spontaneous stone passage occurred in 114 patients (62%). Univariate analysis demonstrated an AUC of 0.83 and 0.82 for the maximum stone diameter determined manually in the axial and coronal planes, respectively. Multivariate models demonstrated an AUC of 0.82 for a model including manual measurement of maximum stone diameter in axial and coronal planes. The same AUC was found for a model including automatic measurement of maximum height and diameter of the stone. Further addition of morphological parameters measured automatically did not increase AUC beyond 0.83.
The semi-automated radiomic analysis of urinary stones shows similar accuracy compared with manual measurements for predicting urinary stone passage. Further studies are needed to predict clinical impacts of reporting the likelihood of urinary stone passage and improving inter-observer variation using automatic radiomic analysis software.
评估半自动化放射组学分析软件预测尿石自然排出可能性的能力，并与手动测量进行比较。
纳入因疑似肾结石或输尿管结石而就诊于急诊科的症状性患者，所有患者均接受 CT 扫描。对患者进行最长达 6 个月的随访，以确定是否尝试自然排出。手动测量轴向和冠状图像上的最大结石直径。还使用自动化放射组学分析软件测量结石的长度、宽度、高度、最大直径、体积、平均和标准偏差的 Hounsfield 单位以及形态特征。使用这些数据开发多变量模型，以预测随后的自发性结石排出，结果以接受者操作曲线（AUC）下面积表示。
共纳入 184 例（69 例女性）年龄中位数为 56 岁的患者。114 例患者（62%）发生自发性结石排出。单变量分析显示，手动测量轴向和冠状平面最大结石直径的 AUC 分别为 0.83 和 0.82。多变量模型显示，包括轴向和冠状平面最大结石直径手动测量的模型 AUC 为 0.82。包括自动测量结石最大高度和直径的模型也具有相同的 AUC。进一步增加自动测量的形态参数并不能使 AUC 超过 0.83。
与手动测量相比，半自动化放射组学分析在预测尿石排出方面具有相似的准确性。需要进一步的研究来预测使用自动放射组学分析软件报告尿石排出可能性和改善观察者间差异的临床影响。
