采用Kappa统计量评价腰椎间盘突出的MRI诊断

Application of the Kappa statistic in MRI diagnosis of lumbar disk herniation

目的应用Kappa统计量评价MRI诊断腰椎间盘突出的一致性。方法检查100例腰腿痛患者的300个椎体。采用3．0 TMR扫描仪，脊柱表面线圈，应用快速自旋回波序列行腰椎矢状面T1WI及T2WI、横轴面T2WI。由2名放射科医师分别在有及无临床资料的情况下对其中50例患者的L3～4、L4～5、L5～S1椎间盘的膨出、突出进行2次评价；再由此2名医师在无临床资料的情况下共同对其中52例患者的156个腰椎间盘进行评价。应用Kappa统计量评价2名医师前后2次的诊断结果及其对相同病例诊断结果的一致性。结果甲乙2名放射科医师前后2次诊断结果相符合的椎间盘数量分别为114和109个，不相符合的椎间盘数量分别为36和41个，诊断结果的一致性为中度，Kappa值分别为0．60±0．06和0．57±0．06。在有临床资料的情况下，腰椎间盘膨出的诊断较前明显增加，分别增加了10个和31个。在没有临床资料的情况下，2名放射科医师之间的诊断结果相符合的椎间盘数量为77个，不相符合者为79个，诊断结果的一致性为弱（Kappa=0．24±0．06）。在有无临床资料的情况下，最大的诊断差异出现在对正常和膨出的腰椎间盘的鉴别上，2名医师2次诊断膨出的差异分别达20和30个，分别占各自总不符合率的55．6％（20／36）和73．2％（30／41）；2名医师对156个椎间盘诊断膨出与正常的差异达56个，为总不符合率的70．9％（56／79）。结论对腰椎间盘膨出判断的不一致是造成同一医师2次诊断结果和2名医师之间诊断结果Kappa值低的最主要原因。

Objective To assess the concordance of MRI diagnosis for patients suspected of lumbar disk herniation by using Kappa statistic. Methods One hundred patients （48 males and 52 females） with lumbosacral radicular pain, aged from 17 to 86 （average 61 ）. All patients underwent fast spin-echo T1 and T2 weighted imaging on a 3.0 T MR scanner and spine surface coil. Two radiologists （ doctor A and doctor B） evaluated the lumbar disks from L3-4, L4-5, and L5-S1 in 50 out of the 100 patients independently. The presence of a bulging disk or a herniation was reported. Images were interpreted twice： once before and once after disclosure of clinical information. And disks of 52 patients out of the 100 samples were interpreted by the two radiologists independently without clinical information as well. The Kappa statistics was employed to assess the concordance of each radiologist＇s diagnoses as well as the observer variation of the two radiologists. Results Diagnoses before and after disclosure to clinical information were concordant in 114 disks for doctor A and in 109 for doctor B, respectively. Diagnoses before and after disclosure to clinical information were not concordant in 36 disks for doctor A and in 41 disks for doctor B, respectively. The Kappa values were 0. 60 ± 0. 06 and 0.57 ± 0. 06 for doctor A and doctor B, respectively. The concordance was moderate. After disclosure to clinical information, the numbers of reported bulging disks increased significantly, by 10 and 31 for doctor A and doctor B, respectively. Without clinical information, the diagnoses of the two radiologists were concordant in 77 disks, while not concordant in 79 disks. The interobserver agreement was poor （ Kappa =0. 24±0. 06）. The difference on diagenoses made between with and without clinical information mainly happened on the differential diagnosis of normal disks and bulging disks. The different ,diagnoses made between with and without clinical information were on 20 disks and on 30 disks for doctor A and doctor B, respectively ; that accounted for 55.6% （20/36） and 73. 2% （ 30/41 ） of total variation respectively. The diagnostic difference between the 2 doctors happened mainly on differentiation of bulging disks and normal disks, which happened in 56 disks, accountiong for 70. 9% （56/79） of total variation. Conclusion Variation on diagnoses of the same radiologist or between two radiologists was mainly caused by disagreement on bulging disks.