Stage migration as a surrogate of survival in hepatocellular carcinoma treated with transarterial chemoembolization.

BACKGROUND AIMS: Locoregional therapies, including transarterial chemoembolization (TACE), are recommended for the treatment of HCC; however, clinical trials evaluating their effectiveness have been complicated by a lack of validated surrogate outcomes. We aimed to evaluate if stage migration could serve as a potential surrogate of overall survival in patients undergoing TACE. APPROACH: We conducted a retrospective cohort study of adult patients with HCC who underwent TACE as initial therapy from 3 centers in the US from 2008 to 2019. The primary outcome was overall survival from the date of the first TACE treatment, and the primary exposure of interest was Barcelona Clinic Liver Cancer stage migration to a more advanced stage within 6 months of TACE. Survival analysis was completed using Kaplan-Meier and multiple Cox proportional hazard models adjusted by the site. RESULTS: Of 651 eligible patients (51.9% Barcelona Clinic Liver Cancer stage A and 39.6% stage B), 129 (19.6%) patients experienced stage migration within 6 months of TACE. Those with stage migration had larger tumors (5.6 vs. 4.2 cm, p < 0.01) and higher AFP levels (median 92 vs. 15 ng/mL, p < 0.01). In multivariate analysis, stage migration was significantly associated with worse survival (HR: 2.82, 95% CI: 2.66-2.98), with a median survival of 8.7 and 15.9 months in those with and without stage migration. Other predictors of worse survival included the White race, higher AFP levels, a higher number of tumors, and a larger maximum HCC diameter. CONCLUSION: Stage migration is associated with increased mortality after TACE in patients with HCC and could serve as a surrogate end point in clinical trials evaluating locoregional therapies such as TACE.

Paraspinal soft tissue edema ratio: An accurate marker for early lumbar spine spondylodiscitis on an unenhanced MRI.

PURPOSE: MRI is currently the gold standard imaging modality in the diagnosis of lumbar spine discitis/osteomyelitis. However, even with supportive clinical and laboratory data, the accuracy of MRI remains limited by several degenerative and inflammatory mimics, such that it continues to represent a challenge for radiologists. This study reports a new quantitative imaging marker of lumbar paraspinal soft tissue edema which shows significant accuracy for spondylodiscitis. METHODS: Thirty-five patients with equivocal MRI findings of lumbar discitis/osteomyelitis vs endplate degenerative changes were reviewed over a 24-month period. Patients with a history of surgery, fractures/recent trauma, signs of advanced infection such as abscesses, phlegmon or severe osseous destruction were excluded. Two ABR board certified neuroradiologists who were blinded to the final diagnosis evaluated a new marker; the superior-inferior paraspinal edema ratio (SI-PER). The SI-PER was obtained by measuring the superior-inferior extent of increased signal/edema in the paraspinal soft tissues on the paraspinal inversion recovery images divided by the vertebral body height measured at midpoint. Cases positive for spondylodiscitis were those confirmed by biopsy, aspiration/drainage, surgery, or clinical improvement following antibiotic treatment. The diagnostic sensitivity and specificity of SI-PER were determined by Receiver operating characteristic (ROC) analysis. RESULTS: In 23/35 (66%) patients, the diagnosis of discitis/osteomyelitis was confirmed. The SI-PER showed a significant association with a positive MRI diagnosis (p = 0.001). Inter-observer correlation for SI-PER was 0.92. ROC analysis showed an area under the curve of 0.84. A SI-PER of 2.5 was 96% sensitive and 75% specific for the diagnosis of discitis/osteomyelitis, with a PPV of 88% and a NPV of 90%. CONCLUSION: In this study, the superior inferior paraspinal edema ratio (SI-PER), a newly defined MRI marker, was found to have high sensitivity for differentiating spondylodiscitis from endplate degenerative changes on lumbar spine MRI.

Automated eloquent cortex localization in brain tumor patients using multi-task graph neural networks.

Localizing the eloquent cortex is a crucial part of presurgical planning. While invasive mapping is the gold standard, there is increasing interest in using noninvasive fMRI to shorten and improve the process. However, many surgical patients cannot adequately perform task-based fMRI protocols. Resting-state fMRI has emerged as an alternative modality, but automated eloquent cortex localization remains an open challenge. In this paper, we develop a novel deep learning architecture to simultaneously identify language and primary motor cortex from rs-fMRI connectivity. Our approach uses the representational power of convolutional neural networks alongside the generalization power of multi-task learning to find a shared representation between the eloquent subnetworks. We validate our method on data from the publicly available Human Connectome Project and on a brain tumor dataset acquired at the Johns Hopkins Hospital. We compare our method against feature-based machine learning approaches and a fully-connected deep learning model that does not account for the shared network organization of the data. Our model achieves significantly better performance than competing baselines. We also assess the generalizability and robustness of our method. Our results clearly demonstrate the advantages of our graph convolution architecture combined with multi-task learning and highlight the promise of using rs-fMRI as a presurgical mapping tool.

Role of C-reactive protein in effective utilization of emergent MRI for spinal infections.

PURPOSE: Emergent spinal MRI is recommended for patients with back pain and red flags for infection. However, many of these studies are negative due to low prevalence of spinal infections. Our purpose was to assess if C-reactive protein (CRP) can be used to guide effective utilization of emergent MRI for spinal infections. METHODS: 316/960 (33%) MRIs performed for infection by the emergency department over 75-month period had CRP levels obtained at presentation, after excluding patients receiving antibiotic or had spinal surgery in < 1 month. An MRI was considered positive when there was imaging evidence of spinal infection confirmed on follow-up by surgery/biopsy/drainage or definitive therapy. A CRP of ≤ 10 mg/L was considered normal and > 100 mg/L as highly elevated. RESULTS: CRP was normal in 95/316 (30%) and abnormal in 221/316 (70%) patients. MRI was positive in 43/316 (13.6%) patients, all of whom had abnormal CRP. CRP (p < 0.001) and intravenous drug use (IVDU; p = 0.002) were independently associated with a positive MRI. Receiver operator characteristic (ROC) analysis showed AUC of 0.76 for CRP, slightly improving with IVDU. Sensitivity, specificity, and negative predictive values for CRP level cut-off: 10 mg/L, 100%, 35%, and 100%, and 100 mg/L, 58%, 70% and 91%, respectively. CONCLUSION: Abnormal CRP, although extremely sensitive, lacks specificity in predicting a positive MRI for spinal infection unless highly elevated. However, a normal CRP (absent recent antibiotic or surgery) makes spinal infection unlikely, and its routine use as a screening test can help reducing utilization of emergent MRI for this purpose.

A Multi-Scale Spatial and Temporal Attention Network on Dynamic Connectivity to Localize The Eloquent Cortex in Brain Tumor Patients.

We present a deep neural network architecture that combines multi-scale spatial attention with temporal attention to simultaneously localize the language and motor areas of the eloquent cortex from dynamic functional connectivity data. Our multi-scale spatial attention operates on graph-based features extracted from the connectivity matrices, thus honing in on the inter-regional interactions that collectively define the eloquent cortex. At the same time, our temporal attention model selects the intervals during which these interactions are most pronounced. The final stage of our model employs multi-task learning to differentiate between the eloquent subsystems. Our training strategy enables us to handle missing eloquent class labels by freezing the weights in those branches while updating the rest of the network weights. We evaluate our method on resting-state fMRI data from one synthetic dataset and one in-house brain tumor dataset while using task fMRI activations as ground-truth labels for the eloquent cortex. Our model achieves higher localization accuracies than conventional deep learning approaches. It also produces interpretable spatial and temporal attention features which can provide further insights for presurgical planning. Thus, our model shows translational promise for improving the safety of brain tumor resections.