New magnetic resonance (MR) molecular imaging techniques offer the potential for non-invasive, simultaneous quantification of metabolic and perfusion parameters in tumors. This study applied a 3D dynamic dual-agent hyperpolarized 13C magnetic resonance spectroscopic imaging (MRSI) approach with 13C-pyruvate and 13C-urea to investigate differences in perfusion and metabolism between low and high grade tumors in the TRAMP transgenic mouse model of prostate cancer. Dynamic MR data were corrected for T1 relaxation and RF excitation and modeled to provide quantitative measures of pyruvate to lactate flux (kPL) and urea perfusion (urea AUC) that correlated with TRAMP tumor histologic grade. kPL values were relatively higher for high-grade TRAMP tumors. The increase in kPL flux correlated significantly with higher lactate dehydrogenase activity and mRNA expression of Ldha, Mct1 and Mct4 as well as with more proliferative disease. There was a significant reduction in perfusion in high-grade tumors that associated with increased hypoxia and mRNA expression of Hif1α and Vegf and increased ktrans, attributed to increased blood vessel permeability. In 90% of the high-grade TRAMP tumors, a mismatch in perfusion and metabolism measurements was observed, with low perfusion being associated with increased kPL. This perfusion-metabolism mismatch was also associated with metastasis. The molecular imaging approach we developed could be translated to investigate these imaging biomarkers for their diagnostic and prognostic power in future prostate cancer clinical trials.

ObjectiveTo identify distinct cognitive phenotypes in temporal lobe epilepsy (TLE) and evaluate patterns of white matter (WM) network alterations associated with each phenotype.MethodsSeventy patients with TLE were characterized into 4 distinct cognitive phenotypes based on patterns of impairment in language and verbal memory measures (language and memory impaired, memory impaired only, language impaired only, no impairment). Diffusion tensor imaging was obtained in all patients and in 46 healthy controls (HC). Fractional anisotropy (FA) and mean diffusivity (MD) of the WM directly beneath neocortex (i.e., superficial WM [SWM]) and of deep WM tracts associated with memory and language were calculated for each phenotype. Regional and network-based SWM analyses were performed across phenotypes.ResultsThe language and memory impaired group and the memory impaired group showed distinct patterns of microstructural abnormalities in SWM relative to HC. In addition, the language and memory impaired group showed widespread alterations in WM tracts and altered global SWM network topology. Patients with isolated language impairment exhibited poor network structure within perisylvian cortex, despite relatively intact global SWM network structure, whereas patients with no impairment appeared similar to HC across all measures.ConclusionsThese findings demonstrate a differential pattern of WM microstructural abnormalities across distinct cognitive phenotypes in TLE that can be appreciated at both the regional and network levels. These findings not only help to unravel the underlying neurobiology associated with cognitive impairment in TLE, but they could also aid in establishing cognitive taxonomies or in the prediction of cognitive course in TLE.

No abstract

To assess feasibility of training a convolutional neural network (CNN) to automate liver segmentation across different imaging modalities and techniques used in clinical practice and to apply this technique to enable automation of liver biometry. Materials and Methods: A two-dimensional U-Net CNN was trained for liver segmentation in two stages by using 330 abdominal MRI and CT examinations. First, the neural network was trained with unenhanced multiecho spoiled gradient-echo images from 300 MRI examinations to yield multiple signal weightings. Then, transfer learning was used to generalize the CNN with additional images from 30 contrast material-enhanced MRI and CT examinations. Performance of the CNN was assessed by using a distinct multiinstitutional dataset curated from multiple sources (498 subjects). Segmentation accuracy was evaluated by computing Dice scores. These segmentations were used to compute liver volume from CT and T1-weighted MRI examinations and to estimate hepatic proton density fat fraction (PDFF) from multiecho T2*-weighted MRI examinations. Quantitative volumetry and PDFF estimates were compared between automated and manual segmentation by using Pearson correlation and Bland-Altman statistics. Results: Dice scores were 0.94 ± 0.06 for CT (n = 230), 0.95 ± 0.03 (n = 100) for T1-weighted MRI, and 0.92 ± 0.05 for T2*weighted MRI (n = 168). Liver volume measured with manual and automated segmentation agreed closely for CT (95% limits of agreement: −298 mL, 180 mL) and T1-weighted MRI (95% limits of agreement: −358 mL, 180 mL). Hepatic PDFF measured by the two segmentations also agreed closely (95% limits of agreement: −0.62%, 0.80%). Conclusion: By using a transfer-learning strategy, this study has demonstrated the feasibility of a CNN to be generalized to perform liver segmentation across different imaging techniques and modalities. With further refinement and validation, CNNs may have broad applicability for multimodal liver volumetry and hepatic tissue characterization.

No abstract

No abstract

BACKGROUND The 2016 World Health Organization Classification of Tumors of the Central Nervous System incorporates the use of molecular information into the classification of brain tumors, including grade II and III gliomas, providing new prognostic information that cannot be delineated based on histopathology alone. We hypothesized that these genomic subgroups may also have distinct imaging features. METHODS A retrospective single institution study was performed on 40 patients with pathologically proven infiltrating WHO grade II/III gliomas with a pre-treatment MRI and molecular data on IDH, chromosomes 1p/19q and ATRX status. Two blinded Neuroradiologists qualitatively assessed MR features. The relationship between each parameter and molecular subgroup (IDH-wildtype; IDH-mutant-1p/19q codeleted-ATRX intact; IDH-mutant-1p/19q intact-ATRX loss) was evaluated with Fisher’s exact test. Progression free survival (PFS) was also analyzed. RESULTS A border that could not be defined on FLAIR was most characteristic of IDH-wildtype tumors, whereas IDH-mutant tumors demonstrated either well-defined or slightly ill-defined borders (p = 0.019). Degree of contrast enhancement and presence of restricted diffusion did not distinguish molecular subgroups. Frontal lobe predominance was associated with IDH-mutant tumors (p = 0.006). The IDH- wildtype subgroup had significantly shorter PFS than the IDH-mutant groups (p < 0.001). No differences in PFS were present when separating by tumor grade. CONCLUSIONS FLAIR border patterns and tumor location were associated with distinct molecular subgroups of grade II/III gliomas. These imaging features may provide fundamental prognostic and predictive information at time of initial diagnostic imaging.

To develop and evaluate a system to prescribe imaging planes for cardiac MRI based on deep learning (DL)−based localization of key anatomic landmarks. Materials and Methods: Annotated landmarks on 892 long-axis (LAX) and 493 short-axis (SAX) cine steady-state free precession series from cardiac MR images were retrospectively collected between February 2012 and June 2017. U-Net−based heatmap regression was used for localization of cardiac landmarks, which were used to compute cardiac MRI planes. Performance was evaluated by comparing localization distances and plane angle differences between DL predictions and ground truth. The plane angulations from DL were compared with those prescribed by the technologist at the original time of acquisition. Data were split into 80% for training and 20% for testing, and results confirmed with fivefold cross-validation. Results: On LAX images, DL localized the apex within mean 12.56 mm ± 19.11 (standard deviation) and the mitral valve (MV) within 7.68 mm ± 6.91. On SAX images, DL localized the aortic valve within 5.78 mm ± 5.68, MV within 5.90 mm ± 5.24, pulmonary valve within 6.55 mm ± 6.39, and tricuspid valve within 6.39 mm ± 5.89. On the basis of these localizations, average angle bias and mean error of DL-predicted imaging planes relative to ground truth annotations were as follows: