Rapid MRI of the breast in evaluating lesions discovered on screening.

BACKGROUND: In Canada, breast MRI has traditionally been reserved for evaluation of disease extent in patients with known breast malignancy. More recently, MRI has been emerging as an instrument for breast screening. However, its utilization is limited by increased relative cost and increased reader time. In this study, we evaluate a rapid MRI protocol for breast cancer screening within a breast screening population. METHODS: A series of 100 MRI studies performed in a high-risk breast cancer population were selected, ensuring a mix of malignant and benign pathology and normal cases. These were presented as full and abbreviated MRI protocols to 3 breast-trained radiologists. Each case was evaluated for Breast Imaging Reporting and Data Systems (BIRADS) category and the presence or absence of cancer. The time taken to complete and interpret each study was also recorded. RESULTS: Of the 100 cases, 17 were of histopathology-proven invasive carcinoma, 6 were ductal carcinoma in situ, 33 were benign, and 44 were normal cases. Sensitivity using the rapid protocol was 69.6% (CI: 47.1-86.8) vs 83% (CI: 61.2-95.1) using the full protocol. Specificity using the rapid protocol was 77.9% (CI: 67.0-86.6) vs 83% (CI: 61.2-95.1) using the full protocol. Intra-observer agreement of BIRADS category and cancer detection was very good (0.82-0.93 weighted Kappa and 0.81-0.9 weighted Kappa, respectively). Inter-observer variability of BIRADS category and cancer detection was moderate (0.54-0.59 and 0.53-0.58, respectively). CONCLUSION: Our study suggests that a rapid MRI protocol is comparable in performance to that of a standard MRI protocol. In addition, breast imagers are unlikely to change their BIRADS assessment of a study based on the additional sequences provided by the lengthier study. The use of a rapid MRI protocol can improve accessibility, thus making breast MRI a more utilized tool for breast cancer screening.

Quantitative non-contrast measurements improve diagnosing dural venous sinus thrombosis.

INTRODUCTION: The only direct sign of sinus thrombosis on non-contrast computerized tomography (NCCT) is the hyperdense sign. The purpose of our study was to assess quantitative parameters for diagnosis of superficial venous sinus thrombosis and to compare these quantitative criteria with the current standard of qualitative evaluation. METHODS: This retrospective case-control study included 18 patients with acute superficial sinus thrombosis and 18 matched controls. Three blinded readers independently evaluated the NCCT for the presence of hyperdense sign using axial slices only followed by axial slices with multiplanar reformats. Absolute attenuation values and ratios were calculated for thrombosed and non-thrombosed sinuses: Ratiotarget sinus/lowest attenuation sinus, Ratiotarget sinus/basilar artery, Ratiotarget sinus/internal carotid artery, Ratiotarget sinus/temporal lobe, and Ratiotarget sinus/frontal lobe. RESULTS: There was a significant difference in absolute attenuation values and ratios between thrombosed and non-thrombosed sinuses, with the absolute attenuation and the Ratiotarget sinus/lowest attenuation sinus being the most differentiating. The mean attenuation for thrombosed sinuses was 69 Hounsfield units (HU) (95 % CI 65-72 HU) vs. 52 HU (95 % CI 51-54) for non-thrombosed, P < 0.0001. The mean Ratiotarget/lowest attenuation was 1.5 (95 % CI 1.4-1.6) for thrombosed sinuses vs. 1.1 (95 % CI 1.0-1.1) for non-thrombosed, P < 0.0001. Optimal thresholds of 62 HU and 1.3 yielded sensitivities of 81 and 84 %, respectively. Hyperdense sign had a sensitivity of 63 % on axial images and 67 % with the addition of multiplanar reformats. CONCLUSION: Density measurements result in substantial improvement over visual inspection in the diagnosis of superficial venous sinus thrombosis on NCCT.

Inter-rater agreement in the characterization of cystic renal lesions on contrast-enhanced MRI.

PURPOSE: To evaluate inter-observer agreement of MRI features and classification of cystic renal masses among radiologist and radiology trainees. METHODS: Four readers (two radiologists and two radiology trainees) retrospectively reviewed 100 cystic renal lesions on gadolinium enhanced MRI and assigned each a Bosniak classification (1, 2, 2F, 3, and 4). Lesions were also assessed on their individual features including size, presence of nodules, septations, and enhancement. Readers ranked their level of confidence regarding Bosniak classifications. Inter-observer variability of lesion classification and features was evaluated between raters at both radiologist and radiology trainee levels as well as the level of agreement of all four readers using weighted Kappa and intraclass correlation coefficient (ICC). RESULTS: One hundred cystic renal lesions were evaluated. There was moderate and substantial classification agreement between trainees and radiologists (ICC 0.59 and 0.63, respectively). There was substantial classification agreement among all four readers (0.66) with the lowest level of agreement for Bosniak 2F lesions (ICC 0.14). There was moderate-substantial agreement for the presence of nodular component, septations, and enhancement. Staff demonstrated highest agreement when assessing for nodular components (0.73). Agreement for the presence of enhancement was lowest (0.37 and 0.42 for radiologists and trainees, respectively). Reported confidence was higher among radiologists compared with trainees. CONCLUSION: There is substantial overall inter-observer agreement in the MRI classification of cystic renal lesions. Confidence increases as rater experience increases.