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

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 lumbosaeral 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 elinical 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 diffcrence 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,aceountiong for 70.9%(56/79)of total variation.Conclusion Variation on diagnoses of the same radiologist or between tworadiologists was mainly caused by disagreement on bulging disks. Key words: Intervertebral disk displacement; Lumbago-leg pain; Factor analysis,statistical; Magnetic resonance imaging