A Clinical and Imaging Fused Deep Learning Model Matches Expert Clinician Prediction of 90-Day Stroke Outcomes.

BACKGROUND AND PURPOSE: Predicting long-term clinical outcome in acute ischemic stroke is beneficial for prognosis, clinical trial design, resource management, and patient expectations. This study used a deep learning-based predictive model (DLPD) to predict 90-day mRS outcomes and compared its predictions with those made by physicians. MATERIALS AND METHODS: A previously developed DLPD that incorporated DWI and clinical data from the acute period was used to predict 90-day mRS outcomes in 80 consecutive patients with acute ischemic stroke from a single-center registry. We assessed the predictions of the model alongside those of 5 physicians (2 stroke neurologists and 3 neuroradiologists provided with the same imaging and clinical information). The primary analysis was the agreement between the ordinal mRS predictions of the model or physician and the ground truth using the Gwet Agreement Coefficient. We also evaluated the ability to identify unfavorable outcomes (mRS >2) using the area under the curve, sensitivity, and specificity. Noninferiority analyses were undertaken using limits of 0.1 for the Gwet Agreement Coefficient and 0.05 for the area under the curve analysis. The accuracy of prediction was also assessed using the mean absolute error for prediction, percentage of predictions ±1 categories away from the ground truth (±1 accuracy [ACC]), and percentage of exact predictions (ACC). RESULTS: To predict the specific mRS score, the DLPD yielded a Gwet Agreement Coefficient score of 0.79 (95% CI, 0.71-0.86), surpassing the physicians' score of 0.76 (95% CI, 0.67-0.84), and was noninferior to the readers (P < .001). For identifying unfavorable outcome, the model achieved an area under the curve of 0.81 (95% CI, 0.72-0.89), again noninferior to the readers' area under the curve of 0.79 (95% CI, 0.69-0.87) (P < .005). The mean absolute error, ±1ACC, and ACC were 0.89, 81%, and 36% for the DLPD. CONCLUSIONS: A deep learning method using acute clinical and imaging data for long-term functional outcome prediction in patients with acute ischemic stroke, the DLPD, was noninferior to that of clinical readers.

Poor venous outflow is associated with hyperintense acute reperfusion marker on follow-up MRI in patients with acute ischemic stroke with a large vessel occlusion.

BACKGROUND: Hyperintense acute reperfusion marker (HARM) refers to delayed enhancement in the subarachnoid or subpial space on post-contrast fluid attenuated inversion recovery (FLAIR) images. HARM is a measure of blood-brain barrier breakdown, which has been correlated with poor outcomes in patients with acute ischemic stroke with large vessel occlusion (AIS-LVO). We hypothesized that unfavorable venous outflow (VO) would be correlated with HARM after thrombectomy treatment of AIS-LVO. OBJECTIVE: To determine whether poor VO is associated with HARM on follow-up MRI after stroke in patients with AIS-LVO. METHODS: Patients with AIS-LVO from the prospective CRISP2 and DEFUSE2 studies with a baseline CT angiography (CTA) scan and a follow-up MRI with FLAIR sequence were screened for enrollment. VO was measured on the baseline CTA scan using the cortical venous opacification score (COVES). HARM was determined on FLAIR sequences at the follow-up MRI. The primary outcome was the occurrence of HARM between those with good VO (VO+; COVES 3-6) and bad VO (VO-; COVES 0-2). RESULTS: 121 patients were included; 60.3% (n=73) had VO+ and 39.7% (n=48) had VO-. Patients with VO- had higher presentation National Institutes of Health Stroke Scale scores (18 (IQR 12-20) vs 12 (IQR 8-16) in VO+; P<0.001). Middle cerebral artery M1 segment occlusions were more common in VO- patients (65% vs 43% VO+; P=0.028). VO- patients also had a larger pre-treatment ischemic core (23 (4-44) mL vs 12 (3-22) mL in VO+; P=0.049) and Tmax >6 s volumes (105 (72-142) mL vs 66 (35-95) mL in VO+; P<0.001). VO- patients were more likely to develop HARM after thrombectomy (31% vs 10% in VO+; P=0.003). On multivariable regression analysis, VO- (OR=3.6 (95% CI 1.2 to 10.6); P=0.02) and the presence of any ICH (OR=3.6 (95% CI 1.2 to 10.5); P=0.02) were independently associated with the occurrence of HARM. CONCLUSIONS: In patients with AIS-LVO, VO- correlated with HARM on post-thrombectomy MRI.