To develop algorithms using convolutional neural networks (CNNs) for automatic segmentation of acute ischemic lesions on diffusion-weighted imaging (DWI) and compare them with conventional algorithms, including a thresholding-based segmentation. Between September 2005 and August 2015, 429 patients presenting with acute cerebral ischemia (training:validation:test set = 246:89:94) were retrospectively enrolled in this study, which was performed under Institutional Review Board approval. Ground truth segmentations for acute ischemic lesions on DWI were manually drawn under the consensus of two expert radiologists. CNN algorithms were developed using two-dimensional U-Net with squeeze-and-excitation blocks (U-Net) and a DenseNet with squeeze-and-excitation blocks (DenseNet) with squeeze-and-excitation operations for automatic segmentation of acute ischemic lesions on DWI. The CNN algorithms were compared with conventional algorithms based on DWI and the apparent diffusion coefficient (ADC) signal intensity. The performances of the algorithms were assessed using the Dice index with 5-fold cross-validation. The Dice indices were analyzed according to infarct volumes (< 10 mL, ≥ 10 mL), number of infarcts (≤ 5, 6-10, ≥ 11), and b-value of 1000 (b1000) signal intensities (< 50, 50-100, > 100), time intervals to DWI, and DWI protocols. The CNN algorithms were significantly superior to conventional algorithms (p < 0.001). Dice indices for the CNN algorithms were 0.85 for U-Net and DenseNet and 0.86 for an ensemble of U-Net and DenseNet, while the indices were 0.58 for ADC-b1000 and b1000-ADC and 0.52 for the commercial ADC algorithm. The Dice indices for small and large lesions, respectively, were 0.81 and 0.88 with U-Net, 0.80 and 0.88 with DenseNet, and 0.82 and 0.89 with the ensemble of U-Net and DenseNet. The CNN algorithms showed significant differences in Dice indices according to infarct volumes (p < 0.001). The CNN algorithm for automatic segmentation of acute ischemic lesions on DWI achieved Dice indices greater than or equal to 0.85 and showed superior performance to conventional algorithms.
In patients with isocitrate dehydrogenase wild-type glioblastoma, necrosis of more than 20% was associated with a greater rate of false-positive measurement of 2-hydroxyglutarate with MR spectroscopy. BackgroundIsocitrate dehydrogenase (IDH) mutation has become one of the most important prognostic biomarkers in glioma management. Measurement of 2-hydroxyglutarate (2HG) with MR spectroscopy has shown high pooled sensitivity, although false-positive results with MR spectroscopy have been reported.PurposeTo investigate factors associated with false-positive 2HG measurements at MR spectroscopy in patients with IDH wild-type glioblastoma.Materials and MethodsThis retrospective study was approved by the institutional review board, and informed consent was waived. Consecutive patients with histopathologically confirmed pre- and posttreatment glioblastoma were evaluated between December 2017 and August 2018. Spectroscopy parameters, including 2HG measurements, were obtained with single-voxel point-resolved spectroscopy, and apparent diffusion coefficient (ADC) values were calculated. Necrosis was graded according to the proportion of necrosis within a volume of interest. Poisson regression analyses were perfor...
BACKGROUND AND PURPOSE: Contrast-enhanced T1-weighted sampling perfection with application-optimized contrasts by using different flip angle evolution (SPACE) with the delay alternating with nutation for tailored excitation (DANTE) pulse could suppress the blood flow signal and provide a higher contrast-to-noise ratio of enhancing lesion-to-brain parenchyma than the MPRAGE sequence. The purpose of our study was to evaluate the usefulness of SPACE with DANTE compared with MPRAGE for detecting brain metastases. MATERIALS AND METHODS: Seventy-one patients who underwent contrast-enhanced SPACE with DANTE and MPRAGE sequences and who were suspected of having metastatic lesions were included. Two neuroradiologists determined the number of enhancing lesions, and diagnostic performance was evaluated using figure of merit, sensitivity, positive predictive value, interobserver agreement, and reading time. Contrast-to-noise ratiolesion/parenchyma and contrast-to-noise ratiowhite matter/gray matter were also assessed. RESULTS: SPACE with DANTE (observer one, 328; observer two, 324) revealed significantly more small (<5 mm) enhancing lesions than MPRAGE (observer one, 175; observer two, 150) (P < 0.001 for observer 1, P ≤ .0001 for observer 2). Furthermore, SPACE with DANTE showed significantly higher figure of merit and sensitivity and shorter reading time than MPRAGE for both observers. The mean contrast-to-noise ratiolesion/parenchyma of SPACE with DANTE (52.3 ± 43.1) was significantly higher than that of MPRAGE (17.5 ± 19.3) (P ≤ .0001), but the mean contrast-to-noise ratiowhite matter/gray matter of SPACE with DANTE (−0.65 ± 1.39) was significantly lower than that of MPRAGE (3.08 ± 1.39) (P ≤ .0001). CONCLUSIONS: Compared with MPRAGE, SPACE with DANTE significantly improves the detection of brain metastases.
20 August 2020 in American Journal of Neuroradiology
BACKGROUND AND PURPOSE: Carotid-cavernous fistulas are abnormal vascular shunts that can cause various neurologic or orbital symptoms. The purpose of this retrospective study was to evaluate the diagnostic performance of thin-section MR imaging for carotid cavernous fistula in patients with clinically suspected carotid cavernous fistula, and to identify possible imaging predictors of carotid cavernous fistula. MATERIALS AND METHODS: A total of 98 patients who were clinically suspected of having carotid cavernous fistula (according to their symptoms and physical examinations) between January 2006 and September 2018 were included in this study. The patients underwent pretreatment thin-section MR imaging and DSA. Thin-section MR imaging consisted of 2D coronal T1- and T2WI with 3-mm thickness and 3D contrast-enhanced T1WI with 0.6 mm thickness. The diagnostic performance of thin-section MR imaging for carotid cavernous fistula was evaluated with the reference standard of DSA. Univariate logistic regression analysis was performed to determine possible imaging predictors of carotid cavernous fistula. RESULTS: Among the 98 patients, DSA confirmed 38 as having carotid cavernous fistula. The overall accuracy, sensitivity, and specificity of thin-section MR imaging were 88.8%, 97.4%, and 83.3%, respectively. Possible imaging predictors on thin-section MR imaging included abnormal contour of the cavernous sinus (OR: 21.7), internal signal void of the cavernous sinus (OR: 15.3), prominent venous drainage flow (OR: 54.0), and orbital/periorbital soft tissue swelling (OR: 40.4). CONCLUSIONS: Thin-section MR imaging provides high diagnostic performance and possible imaging predictors of carotid cavernous fistula in patients with clinically suspected carotid cavernous fistula. Thin-section MR imaging protocols could help decide appropriate management plans for patients with clinically suspected carotid cavernous fistula.
1 January 2023 in Journal of the Korean Society of Radiology