Distortion correction of echo‐planar diffusion‐weighted images of uterine cervix

PURPOSE: To investigate the clinical utility of the reverse gradient algorithm in correcting distortions in diffusion‐weighted images of the cervix and for increasing diagnostic performance. MATERIALS AND METHODS: Forty‐one patients ages 25–72 years (mean 40 ± 11 years) with suspected or early stage cervical cancer were imaged at 3T using an endovaginal coil. T (2)‐weighted (W) and diffusion‐weighted images with right and left phase‐encode gradient directions were obtained coronal to the cervix (b = 0, 100, 300, 500, 800 s mm(−2)). Differences in angle of the endocervical canal to the x‐axis between T (2)W and right‐gradient, left‐gradient, and corrected images were measured. Uncorrected and corrected images were assessed for diagnostic performance when viewed together with T (2)W images by two independent observers against subsequent histology. RESULTS: The angles of the endocervical canal relative to the x‐axis were significantly different between the T (2)W images and the right‐gradient images (P = 0.007), approached significance for left‐gradient images (P = 0.055), and were not significantly different after correction (P = 0.95). Corrected images enabled a definitive diagnosis in 34% (n = 14) of patients classified as equivocal on uncorrected images. Tumor volume in this subset was 0.18 ± 0.44 cm(3) (mean ± SD; sensitivity of detection 100% [8/8], specificity 50% [3/6] for an experienced observer). Correction did not improve diagnostic performance for the less‐experienced observer. CONCLUSION: Distortion‐corrected diffusion‐weighted images improved correspondence with T (2)W images and diagnostic performance in a third of cases. J. Magn. Reson. Imaging 2016;43:1218–1223.