Parametrial invasion in cervical cancer: fused T2-weighted imaging and high-b-value diffusion-weighted imaging with background body signal suppression at 3 T.

Abstract

To retrospectively investigate the value of fused T2-weighted and high-b-value diffusion-weighted imaging with background body signal suppression (DWIBS) at 3 T to evaluate parametrial invasion in cervical cancer. In this institutional review board-approved study, 152 consecutive patients with biopsy-proven cervical cancer who underwent radical hysterectomies also underwent pretreatment magnetic resonance imaging (T2-weighted imaging and DWIBS) at 3 T. Two radiologists independently evaluated the presence of parametrial invasion at T2-weighted imaging, fused T2-weighted imaging and high-b-value DWIBS (ie, fused T2-weighted DWIBS), and combined T2-weighted imaging and fused T2-weighted DWIBS, and the results were compared with histopathologic findings. Parametrial invasion was identified by pathologic analysis in 37 of 152 patients (24.3%). For association with parametrial invasion, the specificity and accuracy of fused T2-weighted DWIBS (97.4% and 90.1%, respectively, for reader 1; 95.7% and 89.5%, respectively, for reader 2) and combined T2-weighted imaging and fused T2-weighted DWIBS (99.1% and 93.4%, respectively, for reader 1; 96.5% and 92.8%, respectively, for reader 2) were significantly better than those of T2-weighted imaging alone (88.7% and 85.5%, respectively, for reader 1; 85.2% and 83.6%, respectively, for reader 2) (all P < .05). The respective sensitivity of T2-weighted imaging, fused T2-weighted DWIBS, and combined T2-weighted imaging and fused T2-weighted DWIBS was 75.7%, 67.6%, and 75.7% for reader 1 and 78.4%, 70.3%, and 81.1% for reader 2, and did not show significant differences (P value, ≤.375 to >.999). The respective area under the curve for association with parametrial invasion of T2-weighted imaging, fused T2-weighted DWIBS, and combined T2-weighted imaging and fused T2-weighted DWIBS was 0.912, 0.951, and 0.976 for reader 1 and 0.890, 0.932, and 0.968 for reader 2 (P < .05). Interreader agreements were excellent (κ = 0.89, 0.9, and 0.86 for T2-weighted imaging, fused T2-weighted DWIBS, and combined T2-weighted imaging and fused T2-weighted DWIBS, respectively). Fusion of high-b-value DWIBS with T2-weighted imaging can improve the diagnostic performance in association with parametrial invasion in cervical cancer compared with T2-weighted imaging alone.