No abstract

P rognosis of functional outcome after ischemic stroke is influenced by a variety of factors already assessable in the acute phase and within the first days after symptom onset. In clinical trials, stroke outcome is most commonly rated by the modified Rankin Scale (mRS) 1 because of the validity and rapid application of this rating scale and its ability to discriminate clinically relevant levels of disability and recovery.2-4 Brain imaging in the early phase after stroke onset provides valuable information related to individual functional recovery. 5,6 In particular, structural MRI identifies injured brain regions and allows for assessment of extent and location, both known to influence and predict functional outcome measured by the mRS.3 However, infarct volume from early MRI correlates only moderately with the mRS at later time points, 7,8 indicating that additional factors, such as lesion location, influence functional outcome. It is therefore of major interest to elucidate the relationship between early lesion patterns and functional impairment in the later course of stroke.Clinical impact of lesion locations can be inferred from voxel-based lesion symptom mapping (VLSM). This statistical method examines effects of brain lesions on behavioral scores on a voxel-by-voxel base. Therefore, a statistical test is conducted for each voxel to detect differences in a behavioral score based on the presence or absence of injury.9 VLSM produces statistical results that map structural lesions to a behavioral scale. In patients with chronic stroke, it has been Background and Purpose-In the early days after ischemic stroke, information on structural brain damage from MRI supports prognosis of functional outcome. It is rated widely by the modified Rankin Scale that correlates only moderately with lesion volume. We therefore aimed to elucidate the influence of lesion location from early MRI (days 2-3) on functional outcome after 1 month using voxel-based lesion symptom mapping. Methods-We analyzed clinical and MRI data of patients from a prospective European multicenter stroke imaging study (I-KNOW). Lesions were delineated on fluid-attenuated inversion recovery images on days 2 to 3 after stroke onset. We generated statistic maps of lesion contribution related to clinical outcome (modified Rankin Scale) after 1 month using voxel-based lesion symptom mapping. Results-Lesion maps of 101 patients with middle cerebral artery infarctions were included for analysis (right-sided stroke, 47%). Mean age was 67 years, median admission National Institutes of Health Stroke Scale was 11. Mean infarct volumes were comparable between both sides (left, 37.5 mL; right, 43.7 mL). Voxel-based lesion symptom mapping revealed areas with high influence on higher modified Rankin Scale in regions involving the corona radiata, internal capsule, and insula. In addition, asymmetrically distributed impact patterns were found involving the right inferior temporal gyrus and left superior temporal gyrus. Conclusions-In this group of patients with strok...

Precision medicine is an emerging approach to clinical research and patient care that focuses on understanding and treating disease by integrating multimodal or ‘multi-omics’ data from an individual to make patient-tailored decisions. With the large and complex datasets generated using precision medicine diagnostic approaches, novel techniques to process and understand these complex data were needed. At the same time, computer science has progressed rapidly to develop techniques that enable the storage, processing, and analysis of these complex datasets, a feat that traditional statistics and early computing technologies could not accomplish. Machine learning, a branch of artificial intelligence, is a computer science methodology that aims to identify complex patterns in data that can be used to make predictions or classifications on new unseen data or for advanced exploratory data analysis. Machine learning analysis of precision medicine’s multimodal data allows for broad analysis of large datasets and ultimately a greater understanding of human health and disease. This review focuses on machine learning utilization for precision medicine’s “big data”, in the context of genetics, genomics, and beyond.

P rogression to infarction after acute ischemic stroke onset is time-sensitive and has substantial intersubject variability. 1,2 Computed tomographic (CT) perfusion (CTP) measurement of brain parenchyma can be used to estimate ischemic core and penumbra and, therefore, provide immediate information for treatment decision-making. Current CTP thresholds that estimate these tissue states are generally derived either by comparison with magnetic resonance (MR) diffusion-weighted imaging (DWI), often done within an hour of CTP, or with follow-up infarction in patients who have reperfused sometime within 24 hours.3-11 Because infarcts grow over time and final tissue fate depends greatly on what happens in the minutes to hours immediately after this imaging snapshot, CTP thresholds predicting infarction are likely to depend on the time from stroke symptom onset to imaging, time from imaging to reperfusion, and the quality of reperfusion.Background and Purpose-Among patients with acute ischemic stroke, we determine computed tomographic perfusion (CTP) thresholds associated with follow-up infarction at different stroke onset-to-CTP and CTP-to-reperfusion times. Methods-Acute ischemic stroke patients with occlusion on computed tomographic angiography were acutely imaged with CTP. Noncontrast computed tomography and magnectic resonance diffusion-weighted imaging between 24 and 48 hours were used to delineate follow-up infarction. Reperfusion was assessed on conventional angiogram or 4-hour repeat computed tomographic angiography. T max , cerebral blood flow, and cerebral blood volume derived from delayinsensitive CTP postprocessing were analyzed using receiver-operator characteristic curves to derive optimal thresholds for combined patient data (pooled analysis) and individual patients (patient-level analysis) based on time from stroke onset-to-CTP and CTP-to-reperfusion. One-way ANOVA and locally weighted scatterplot smoothing regression was used to test whether the derived optimal CTP thresholds were different by time. Results-One hundred and thirty-two patients were included. T max thresholds of >16.2 and >15.8 s and absolute cerebral blood flow thresholds of <8.9 and <7.4 mL•min −1•100 g −1 were associated with infarct if reperfused <90 min from CTP with onset <180 min. The discriminative ability of cerebral blood volume was modest. No statistically significant relationship was noted between stroke onset-to-CTP time and the optimal CTP thresholds for all parameters based on discrete or continuous time analysis (P>0.05). A statistically significant relationship existed between CTP-to-reperfusion time and the optimal thresholds for cerebral blood flow (P<0.001; r=0.59 and 0.77 for gray and white matter, respectively) and T max (P<0.001; r=−0.68 and −0.60 for gray and white matter, respectively) parameters. Conclusions-Optimal CTP thresholds associated with follow-up infarction depend on time from imaging to reperfusion.

The fast and accurate assessment of cerebral perfusion is fundamental for the diagnosis and successful treatment of stroke patients. Magnetic particle imaging (MPI) is a new radiation-free tomographic imaging method with a superior temporal resolution, compared to other conventional imaging methods. In addition, MPI scanners can be built as prehospital mobile devices, which require less complex infrastructure than computed tomography (CT) and magnetic resonance imaging (MRI). With these advantages, MPI could accelerate the stroke diagnosis and treatment, thereby improving outcomes. Our objective was to investigate the capabilities of MPI to detect perfusion deficits in a murine model of ischemic stroke. Cerebral ischemia was induced by inserting of a microfilament in the internal carotid artery in C57BL/6 mice, thereby blocking the blood flow into the medial cerebral artery. After the injection of a contrast agent (superparamagnetic iron oxide nanoparticles) specifically tailored for MPI, cerebral perfusion and vascular anatomy were assessed by the MPI scanner within seconds. To validate and compare our MPI data, we performed perfusion imaging with a small animal MRI scanner. MPI detected the perfusion deficits in the ischemic brain, which were comparable to those with MRI but in real-time. For the first time, we showed that MPI could be used as a diagnostic tool for relevant diseases in vivo, such as an ischemic stroke. Due to its shorter image acquisition times and increased temporal resolution compared to that of MRI or CT, we expect that MPI offers the potential to improve stroke imaging and treatment.

In this preliminary study involving young adults without clinical evidence of cerebrovascular disease, a greater number of modifiable cardiovascular risk factors at recommended levels was associated with higher cerebral vessel density and caliber, higher cerebral blood flow, and fewer white matter hyperintensities. Further research is needed to verify these findings and determine their clinical importance.

Background: Mild behavioral impairment (MBI) and subjective cognitive decline (SCD) are dementia risk states, and potentially represent neurobehavioral and neurocognitive manifestations, respectively, of early stage neurodegeneration. Both MBI and SCD predict incident cognitive decline and dementia, are associated with known dementia biomarkers, and are both represented in the NIA-AA research framework for AD in Stage 2 (preclinical disease). Objective: To assess the associations of MBI and SCD, alone and in combination, with incident cognitive and functional decline in a population of older adults. We tested the hypothesis that MBI and SCD confer additive risk for decline. Methods: Cognitively normal participants were followed up annually at Alzheimer’s Disease Centers. Logistic regression assessed the relationship between baseline classification (MBI-SCD-, MBI-SCD+, MBI+SCD-, or MBI+SCD+) and 3-year outcome. Results: Of 2,769 participants (mean age=76), 1,536 were MBI-SCD-, 254 MBI-SCD+, 743 MBI+SCD-, and 236 MBI+SCD+. At 3 years, 349 (12.6%) declined to CDR >0, including 23.1% of the MBI+group, 23.5% of the SCD+group, and 30.9% of the intersection group of both MBI+and SCD+participants. Compared to SCD-MBI-, we observed an ordinal progression in risk (ORs [95% CI]): 3.61 [2.42–5.38] for MBI-SCD+ (16.5% progression), 4.76 [3.57–6.34] for MBI+SCD- (20.7%), and 8.15 [5.71–11.64] for MBI+SCD+(30.9%). Conclusion: MBI and SCD together were associated with the greatest risk of decline. These complementary dementia risk syndromes can be used as simple and scalable methods to identify high-risk patients for workup or for clinical trial enrichment.

We present DeepVesselNet, an architecture tailored to the challenges faced when extracting vessel trees and networks and corresponding features in 3-D angiographic volumes using deep learning. We discuss the problems of low execution speed and high memory requirements associated with full 3-D networks, high-class imbalance arising from the low percentage (&lt;3%) of vessel voxels, and unavailability of accurately annotated 3-D training data—and offer solutions as the building blocks of DeepVesselNet. First, we formulate 2-D orthogonal cross-hair filters which make use of 3-D context information at a reduced computational burden. Second, we introduce a class balancing cross-entropy loss function with false-positive rate correction to handle the high-class imbalance and high false positive rate problems associated with existing loss functions. Finally, we generate a synthetic dataset using a computational angiogenesis model capable of simulating vascular tree growth under physiological constraints on local network structure and topology and use these data for transfer learning. We demonstrate the performance on a range of angiographic volumes at different spatial scales including clinical MRA data of the human brain, as well as CTA microscopy scans of the rat brain. Our results show that cross-hair filters achieve over 23% improvement in speed, lower memory footprint, lower network complexity which prevents overfitting and comparable accuracy that does not differ from full 3-D filters. Our class balancing metric is crucial for training the network, and transfer learning with synthetic data is an efficient, robust, and very generalizable approach leading to a network that excels in a variety of angiography segmentation tasks. We observe that sub-sampling and max pooling layers may lead to a drop in performance in tasks that involve voxel-sized structures. To this end, the DeepVesselNet architecture does not use any form of sub-sampling layer and works well for vessel segmentation, centerline prediction, and bifurcation detection. We make our synthetic training data publicly available, fostering future research, and serving as one of the first public datasets for brain vessel tree segmentation and analysis.