Deep learning on pre-procedural computed tomography and clinical data predicts outcome following stroke thrombectomy.

BACKGROUND: Deep learning using clinical and imaging data may improve pre-treatment prognostication in ischemic stroke patients undergoing endovascular thrombectomy (EVT). METHODS: Deep learning models were trained and tested on baseline clinical and imaging (CT head and CT angiography) data to predict 3-month functional outcomes in stroke patients who underwent EVT. Classical machine learning models (logistic regression and random forest classifiers) were constructed to compare their performance with the deep learning models. An external validation dataset was used to validate the models. The MR PREDICTS prognostic tool was tested on the external validation set, and its performance was compared with the deep learning and classical machine learning models. RESULTS: A total of 975 patients (550 men; mean±SD age 67.5±15.1 years) were studied with 778 patients in the model development cohort and 197 in the external validation cohort. The deep learning model trained on baseline CT and clinical data, and the logistic regression model (clinical data alone) demonstrated the strongest discriminative abilities for 3-month functional outcome and were comparable (AUC 0.811 vs 0.817, Q=0.82). Both models exhibited superior prognostic performance than the other deep learning (CT head alone, CT head, and CT angiography) and MR PREDICTS models (all Q<0.05). CONCLUSIONS: The discriminative performance of deep learning for predicting functional independence was comparable to logistic regression. Future studies should focus on whether incorporating procedural and post-procedural data significantly improves model performance.

Active conductive head cooling of normal and infarcted brain: A magnetic resonance spectroscopy imaging study.

Active conductive head cooling is a simple and non-invasive intervention that may slow infarct growth in ischemic stroke. We investigated the effect of active conductive head cooling on brain temperature using whole brain echo-planar spectroscopic imaging. A cooling cap (WElkins Temperature Regulation System, 2nd Gen) was used to administer cooling for 80 minutes to healthy volunteers and chronic stroke patients. Whole brain echo-planar spectroscopic imaging scans were obtained before and after cooling. Brain temperature was estimated using the Metabolite Imaging and Data Analysis System software package, which allows voxel-level temperature calculations using the chemical shift difference between metabolite (N-acetylaspartate, creatine, choline) and water resonances. Eleven participants (six healthy volunteers, five post-stroke) underwent 80 ± 5 minutes of cooling. The average temperature of the coolant was 1.3 ± 0.5°C below zero. Significant reductions in brain temperature (ΔT = -0.9 ± 0.7°C, P = 0.002), and to a lesser extent, rectal temperature (ΔT = -0.3 ± 0.1°C, P = 0.03) were observed. Exploratory analysis showed that the occipital lobes had the greatest reduction in temperature (ΔT = -1.5 ± 1.2°C, P = 0.002). Regions of infarction had similar temperature reductions to the contralateral normal brain. Future research could investigate the feasibility of head cooling as a potential neuroprotective strategy in patients being considered for acute stroke therapies.

Vertebrobasilar Artery Calcification and Outcomes in Posterior Circulation Large Vessel Occlusion Thrombectomy.

Background and Purpose- Intracranial carotid artery calcification is associated with worse outcome in anterior circulation stroke patients who undergo endovascular thrombectomy. We investigated the association between vertebrobasilar artery calcification (VBAC) and outcome in patients undergoing endovascular thrombectomy for posterior circulation large vessel occlusion. Methods- Consecutive patients treated for posterior circulation large vessel occlusion from a prospective single-center registry were studied. VBAC was manually segmented on computed tomography brain scans. The associations between VBAC and VBAC volume, functional independence (90-day modified Rankin Scale score of 0-2), and 90-day mortality were assessed using propensity score-adjusted logistic regression. Results- Sixty-four posterior circulation large vessel occlusion patients were included. Twenty-five (39.1%) patients had VBAC, and of these, the median (interquartile range) VBAC volume was 19.8 (6.65-23.4) mm3. VBAC was associated with reduced functional independence (OR, 0.19 [95% CI, 0.04-0.78]; P=0.03) and increased mortality (OR, 9.44 [95% CI, 2.43-36.62]; P=0.005). Larger VBAC volumes were a significant predictor of reduced functional independence and increased mortality. Conclusions- VBAC is an independent predictor of outcome in patients undergoing endovascular thrombectomy for posterior circulation large vessel occlusion. Considering the presence of VBAC might improve prognostication and shared treatment decision-making between patients, families, and physicians.

Automated Measurement of Cerebral Atrophy and Outcome in Endovascular Thrombectomy.

Background and Purpose- Methods of identifying ischemic stroke patients with a greater probability of poor outcome following endovascular thrombectomy (EVT) might improve shared treatment decision-making between patients, families, and physicians. We used an objective, automated method to measure cerebral atrophy and investigated whether this was associated with outcome in EVT patients. Methods- Consecutive EVT patients from a single-center registry were studied. CT brain scans were segmented with a combination of a validated U-Net and Hounsfield unit thresholding. Intracranial cerebrospinal fluid (CSF) volume was used as a marker of cerebral atrophy and calculated as a proportion of total intracranial volume. The primary outcome was functional independence, defined as a 3-month modified Rankin Scale score of 0 to 2. Results- Three-hundred sixty EVT patients were included. Functional independence was achieved in 204 (56.7%) patients. The mean±SD CSF volume was 9.0±4.7% of total intracranial volume. Multivariable regression demonstrated that increasing CSF volume was associated with reduced functional independence (OR=0.65 per 5% increase in CSF volume; 95% CI, 0.48-0.89; P=0.007) and higher 3-month modified Rankin Scale scores (common OR, 1.59 per 5% increase in CSF volume; 95% CI, 1.05-2.41; P=0.03). Conclusions- Cerebral atrophy determined by automated measurement of intracranial CSF volume is associated with functional outcome in patients undergoing EVT. If validated in future studies, this simple, objective, and automated imaging marker could potentially be incorporated into decision-support tools to improve shared treatment decision-making.