Person:

Lev, Michael

The Massachusetts General Hospital Neuroradiology Division employed an experience and evidence based approach to develop a neuroimaging algorithm to best select patients with severe ischemic strokes caused by anterior circulation occlusions (ACOs) for intravenous tissue plasminogen activator and endovascular treatment. Methods found to be of value included the National Institutes of Health Stroke Scale (NIHSS), non-contrast CT, CT angiography (CTA) and diffusion MRI. Perfusion imaging by CT and MRI were found to be unnecessary for safe and effective triage of patients with severe ACOs. An algorithm was adopted that includes: non-contrast CT to identify hemorrhage and large hypodensity followed by CTA to identify the ACO; diffusion MRI to estimate the core infarct; and NIHSS in conjunction with diffusion data to estimate the clinical penumbra.

Purpose: To improve ischemic stroke outcome prediction using imaging information from a prospective cohort who received admission CT angiography (CTA). Methods: In a prospectively designed study, 649 stroke patients diagnosed with acute ischemic stroke had admission NIH stroke scale scores, noncontrast CT (NCCT), CTA, and 6-month outcome assessed using the modified Rankin scale (mRS) scores. Poor outcome was defined as mRS>2. Strokes were classified as “major” by the (1) Alberta Stroke Program Early CT Score (ASPECTS+) if NCCT ASPECTS was(\leq7); (2) Boston Acute Stroke Imaging Scale (BASIS+) if they were ASPECTS+ or CTA showed occlusion of the distal internal carotid, proximal middle cerebral, or basilar arteries; and (3) NIHSS for scores>10. Results: Of 649 patients, 253 (39.0%) had poor outcomes. NIHSS, BASIS, and age, but not ASPECTS, were independent predictors of outcome. BASIS and NIHSS had similar sensitivities, both superior to ASPECTS (p<0.0001). Combining NIHSS with BASIS was highly predictive: 77.6% (114/147) classified as NIHSS>10/BASIS+ had poor outcomes, versus 21.5% (77/358) with NIHSS(\leq10)/BASIS− (p<0.0001), regardless of treatment. The odds ratios for poor outcome is 12.6 (95% CI: 7.9 to 20.0) in patients who are NIHSS>10/BASIS+ compared to patients who are NIHSS(\leq10)/BASIS−; the odds ratio is 5.4 (95% CI: 3.5 to 8.5) when compared to patients who are only NIHSS>10 or BASIS+. Conclusions: BASIS and NIHSS are independent outcome predictors. Their combination is stronger than either instrument alone in predicting outcomes. The findings suggest that CTA is a significant clinical tool in routine acute stroke assessment.

Background: Major anterior circulation ischemic strokes caused by occlusion of the distal internal carotid artery (ICA) or proximal middle cerebral artery (MCA) or both account for about one-third of ischemic strokes with mostly poor outcomes. These strokes are treatable by IV-tPA and endovascular methods. However, dynamics of infarct growth in these strokes are poorly documented. The purpose was to help understand infarct growth dynamics by measuring acute infarct size with DWI at known times after stroke onset in patients with documented ICA/MCA occlusions. Methods: Retrospectively, we included 47 consecutive patients with documented ICA/MCA occlusions who underwent DWI within 30h of stroke onset. Prospectively, 139 patients were identified using the same inclusion criteria. DWI lesion volumes were measured and correlated to time since stroke onset. Perfusion data was reviewed in those who underwent perfusion imaging. Results: Acute infarct volumes ranged from 0.41-318.3ml. Infarct size and time did not correlate (R2=0.001). The majority of patients had DWI lesions that were less than 25% the territory at risk (<70ml) whether they were imaged < or >8h after stroke onset. DWI lesions corresponded to areas of greatly reduced perfusion. Conclusions: Poor correlation between infarct volume and time after stroke onset suggests that there are factors more powerful than time in determining infarct size within the first 30h. The observations suggest that highly variable cerebral perfusion via the collateral circulation may primarily determine infarct growth dynamics. If verified, clinical implications include the possibility of treating many patients outside traditional time windows.

BACKGROUND AND PURPOSE: To determine the effect of intravenous tissue plasminogen activator (IV-tPA) on outcomes in patients with severe major anterior circulation ischemic stroke. METHODS: Prospectively, 649 patients with acute stroke had admission National Institutes of Health stroke scale (NIHSS) scores, noncontrast computed tomography (CT), CT angiography (CTA), and 6-month outcome assessed using modified Rankin scale. IV-tPA treatment decisions were made before CTA, at the time of noncontrast CT scanning, as per routine clinical protocol. Severe symptoms were defined as NIHSS>10. Poor outcome was defined as modified Rankin scale >2. Major occlusions were identified on CTA. Univariate and multivariate stepwise-forward logistic regression analyses of the full cohort were performed. RESULTS: Of 649 patients, 188 (29%) patients presented with NIHSS>10, and 64 out of 188 (34%) patients received IV-tPA. Admission NIHSS, large artery occlusion, and IV-tPA all independently predicted good outcomes; however, a significant interaction existed between IV-tPA and occlusion (P<0.001). Of the patients who presented with NIHSS>10 with anterior circulation occlusion, twice the percentage had good outcomes if they received IV-tPA (17 out of 49 patients, 35%) than if they did not (13 out of 77 patients, 17%; P=0.031). The number needed to treat was 7 (95% confidence interval, 3-60). CONCLUSIONS: IV-tPA treatment resulted in significantly better outcomes in patients with severely symptomatic stroke with major anterior circulation occlusions. The 35% good outcome rate was similar to rates found in endovascular therapy trials. Vascular imaging may help in patient selection and stratification for trials of IV-thrombolytic and endovascular therapies.

Owing to improvements in image recognition via deep learning, machine-learning algorithms could eventually be applied to automated medical diagnoses that can guide clinical decision-making. However, these algorithms remain a 'black box' in terms of how they generate the predictions from the input data. Also, high-performance deep learning requires large, high-quality training datasets. Here, we report the development of an understandable deep-learning system that detects acute intracranial haemorrhage (ICH) and classifies five ICH subtypes from unenhanced head computed-tomography scans. By using a dataset of only 904 cases for algorithm training, the system achieved a performance similar to that of expert radiologists in two independent test datasets containing 200 cases (sensitivity of 98% and specificity of 95%) and 196 cases (sensitivity of 92% and specificity of 95%). The system includes an attention map and a prediction basis retrieved from training data to enhance explainability, and an iterative process that mimics the workflow of radiologists. Our approach to algorithm development can facilitate the development of deep-learning systems for a variety of clinical applications and accelerate their adoption into clinical practice.