基于灰度双阈值自动测量MRI所示膝骨关节炎积液体积

Automatic measurement of effusion volume of knee osteoarthritis on MRI based on gray dual-threshold

目的基于灰度双阈值开发自动测量MRI中膝关节腔积液体积的方法,并验证其可重复性。方法基于Matlab平台对151例膝骨关节炎(KOA)患者的MRI(测试集)进行滤波、双阈值分割等操作,自动提取ROI后计算积液面积,经内插处理后得到积液体积;同时以人工勾画方法测量积液体积并进行等级评价。采用组内相关系数(ICC)评价观察者间测量积液体积及等级评价结果的一致性;分别以Pearson相关性分析和Spearman相关性分析评价自动、手动测量测试集与外部验证集(n=50)积液等级的相关性。结果自动测量测试集平均耗时(4.38±0.98)s,手动测量平均耗时(12.54±5.37)min。观察者间测量积液体积结果的一致性很好(ICC:0.93~0.99,P<0.05)。测试集和验证集中,自动测量与手动测量评价积液等级结果均呈高度正相关(r=0.78、0.87,P均<0.01)。结论基于灰度双阈值自动测量MRI所示膝关节积液方法与手动测量结果的一致性较好,且用时短、人工干预少,可重复性高。

Objective To develop an automatic method for measuring knee joint effusion volume on MRI based on gray dual-threshold,and to verify its repeatability.Methods Based on Matlab platform,filter and dual-threshold segmentation were performed on MRI of 151 cases of knee osteoarthritis(KOA)in the testing set.ROI was automatically extracted to calculate the area of effusion,and the volume rendering was then performed to calculate effusion volume after interpolation.At the same time,the volume of effusion was measured and evaluated manually.Intra-class correlation coefficient(ICC)was used to evaluate the intra-observer consistency of measuring effusion volume and grade evaluation.Pearson correlation analysis and Spearman correlation analysis were used to observe the correlation between automatically and manually evaluated effusion levels in the testing set and external validation set(n=50),respectively.Results The average time of automatic and manual measurements of the testing set was(4.38±0.98)s and(12.54±5.37)min,respectively.The intra-observer consistency of effusion volume was good(ICC:0.93-0.99,P<0.05).In both testing set and validation set,automatic measured effusion volumes were highly positively associated with manual measured volumes(r=0.78,0.87,both P<0.01).Conclusion The automatic measurement method of knee joint effusion volume developed based on gray dual-threshold had good consistency with the manual measurements,with less time-consuming,less manual intervention and high repeatability.