Interobserver agreement of non-parallel transmission technology zoomed-diffusion-weighted imaging (DWI) for ordinal biparametric Vesical Imaging-Reporting and Data System (VI-RADS)-based scoring in bladder cancer lesions ≥1 cm

Abstract

Background Preoperative assessment of muscle invasion in bladder cancer using the Vesical Imaging-Reporting and Data System (VI-RADS) relies heavily on diffusion-weighted imaging (DWI), but conventional single-shot echo-planar imaging diffusion-weighted imaging (ss-EPI-DWI) may be limited by image distortion and artifacts, potentially affecting interobserver agreement. Zoomed diffusion-weighted imaging without parallel transmission technology (non-pTx zoomed-DWI) may improve image quality. This study aimed to compare the interobserver agreement of non-pTx zoomed-DWI versus conventional ss-EPI-DWI for ordinal biparametric VI-RADS-based scoring in bladder cancer lesions ≥1 cm across different morphological types among radiologists with varying levels of experience. Methods This retrospective single-center study included patients with pathologically confirmed bladder cancer who underwent preoperative biparametric magnetic resonance imaging (MRI), including T2-weighted imaging (T2WI), conventional ss-EPI-DWI, and non-pTx zoomed-DWI, before transurethral resection of bladder tumor or radical cystectomy. Three radiologists independently assigned ordinal biparametric VI-RADS-based F-scores from 1 to 5 for each eligible lesion using ss-EPI-DWI and non-pTx zoomed-DWI datasets separately. T2WI was used for lesion localization and morphological classification, whereas the corresponding DWI sequence, including high-b-value DWI images and apparent diffusion coefficient (ADC) maps, was used to assign ordinal biparametric VI-RADS-based F-scores according to the standard VI-RADS scoring criteria. Subjective image quality was rated on a 4-point scale. Interobserver agreement was assessed using linear weighted kappa statistics with 95% confidence intervals. Results A total of 162 patients (138 males, 24 females; median age 67 years) with 214 eligible bladder cancer lesions were included. Non-pTx zoomed-DWI provided significantly higher subjective image quality scores than ss-EPI-DWI for both readers (P<0.001). Based on the original ordinal VI-RADS-based F-scores, overall pairwise interobserver agreement was good with ss-EPI-DWI (linear weighted κ=0.708–0.761) and excellent with non-pTx zoomed-DWI (linear weighted κ=0.851–0.870). In morphology-specific analyses, agreement also improved with non-pTx zoomed-DWI across lesion subgroups. The improvement was particularly evident in broad-based nodules, for which agreement increased from good with ss-EPI-DWI (κ=0.700–0.752) to excellent with non-pTx zoomed-DWI (κ=0.892–0.919), and in nodules with pedicles, for which agreement improved from moderate-to-good (κ=0.535–0.640) to good-to-excellent (κ=0.786–0.847). Conclusions Non-pTx zoomed-DWI improved subjective image quality and interobserver agreement for ordinal biparametric VI-RADS-based scoring in bladder cancer lesions ≥1 cm compared with conventional ss-EPI-DWI. These findings suggest that non-pTx zoomed-DWI may support more consistent MRI-based assessment of muscle invasion risk, particularly in morphologically complex lesions. Further prospective studies with lesion-level pathological correlation are needed to determine whether improved agreement translates into improved diagnostic accuracy.