MRI Posttreatment Surveillance for Head and Neck Squamous Cell Carcinoma: Proposed MR NI-RADS Criteria

Abstract

BACKGROUND AND PURPOSE:The current Neck Imaging Reporting and Data System (NI-RADS) criteria were designed for contrastenhanced CT with or without PET. Prior studies have revealed the capability of DWI and T2 signal intensity in distinguishing locoregional tumor residual and recurrence from posttreatment benign findings in head and neck cancers. We aimed to propose MR imaging NI-RADS criteria by adding diffusion criteria and T2 signal intensity to the American College of Radiology NI-RADS template. MATERIALS AND METHODS:This retrospective study included 69 patients with head and neck squamous cell carcinoma (HNSCC) who underwent posttreatment contrast-enhanced MRI imaging surveillance using a 1.5T scanner. The scans were interpreted by 2 neuroradiologists. Image analysis assessed the primary tumor site using the current American College of Radiology NI-RADS morphologic lexicon (mainly designed for contrast-enhanced CT with or without PET). NI-RADS rescoring was then performed based on our proposed criteria using T2 signal and diffusion features. The reference standard was a defined set of criteria, including clinical and imaging follow-up and pathologic assessment.RESULTS: Imaging assessment of treated HNSCC at the primary tumor site using T2 signal intensity and diffusion features as modifying rules to NI-RADS showed higher sensitivity, specificity, positive predictive value, negative predictive value, and accuracy (92.3%, 90.7%, 85.7%, 95.1%, and 91.3%, respectively) compared with the current NI-RADS lexicon alone (84.6%, 81.4%, 73.3%, 89.8%, and 82.6%, respectively). CONCLUSIONS:The addition of diffusion features and T2 signal to the American College of Radiology NI-RADS criteria for the primary tumor site enhances the specificity, sensitivity, positive predictive value, negative predictive value, and NI-RADS accuracy.

Glioblastoma Surgery Imaging–Reporting and Data System: Validation and Performance of the Automated Segmentation Task

“…The final architecture was as described in the original article [30] with five levels and [16,32,128,256,256] as the filter sizes, with the input size set to 128 × 128 × 144 voxels. All modules including the multiscale input, deep supervision, and singe attention modules were enabled.…”

“…The automatic generation of standardized reports is bound to the same principles and requirements as the segmentation task, where robustness and repeatability are key. Standard Reporting And Data Systems (RADSs) have been proposed for several solid tumors, such as in prostate cancer [ 14 ], breast cancer [ 15 ], head and neck cancer [ 16 ], and lung cancer [ 17 ]. Common rules for imaging techniques, reports’ terminology, and definitions of tumor features and treatment response were defined.…”

ADC for Differentiation between Posttreatment Changes and Recurrence in Head and Neck Cancer: A Systematic Review and Meta-analysis

“…The American College of Radiology has proposed that the NI-RADS accurately reports radiologic assessment of recurrence or residual of head and neck cancer, which is evaluated by [ 18 F] FDG-PET/CT or contrast-enhanced CT, 4,28,29 and it has been reported that diagnostic accuracy is improved when qualitative MR imaging findings such as DWI and T2-weighted images are incorporated. 30 Future studies are likely needed to evaluate the benefits of incorporating the quantitative ADC value into NI-RADS.…”

PET/MR Imaging in Evaluating Treatment Failure of Head and Neck Malignancies: A Neck Imaging Reporting and Data System–Based Study

“…Recent studies of NI-RADS scores applied to contrast-enhanced MR imaging report interreader agreement varying from low to substantial. 23,24 NI-RADS templates were not specifically designed for MR imaging, and recent work suggests that MR imaging features such as diffusion restriction and T2 signal be added to NI-RADS criteria, 25 possibly improving interreader agreement. Establishing MR imaging criteria may allow abbreviated MR imaging protocols and shorter image-acquisition times; this possibility could be significant, given the drawback of the time required for PET/MR imaging.…”