Use of Deep Learning to Predict Final Ischemic Stroke Lesions From Initial Magnetic Resonance Imaging

Abstract

IMPORTANCE Predicting infarct size and location is important for decision-making and prognosis in patients with acute stroke. OBJECTIVES To determine whether a deep learning model can predict final infarct lesions using magnetic resonance images (MRIs) acquired at initial presentation (baseline) and to compare the model with current clinical prediction methods. DESIGN, SETTING, AND PARTICIPANTS In this multicenter prognostic study, a specific type of neural network for image segmentation (U-net) was trained, validated, and tested using patients from the Imaging Collaterals in Acute Stroke (iCAS) study from April 14, 2014, to April 15, 2018, and the Diffusion Weighted Imaging Evaluation for Understanding Stroke Evolution Study-2 (DEFUSE-2) study from July 14, 2008, to September 17, 2011 (reported in October 2012). Patients underwent baseline perfusion-weighted and diffusion-weighted imaging and MRI at 3 to 7 days after baseline. Patients were grouped into unknown, minimal, partial, and major reperfusion status based on 24-hour imaging results. Baseline images acquired at presentation were inputs, and the final true infarct lesion at 3 to 7 days was considered the ground truth for the model. The model calculated the probability of infarction for every voxel, which can be thresholded to produce a prediction. Data were analyzed from July 1, 2018, to March 7, 2019. MAIN OUTCOMES AND MEASURES Area under the curve, Dice score coefficient (DSC) (a metric from 0-1 indicating the extent of overlap between the prediction and the ground truth; a DSC of Ն0.5 represents significant overlap), and volume error. Current clinical methods were compared with model performance in subgroups of patients with minimal or major reperfusion. RESULTS Among the 182 patients included in the model (97 women [53.3%]; mean [SD] age, 65 [16] years), the deep learning model achieved a median area under the curve of 0.92 (interquartile range [IQR], 0.87-0.96), DSC of 0.53 (IQR, 0.31-0.68), and volume error of 9 (IQR, −14 to 29) mL. In subgroups with minimal (DSC, 0.58 [IQR, 0.31-0.67] vs 0.55 [IQR, 0.40-0.65]; P = .37) or major (DSC, 0.48 [IQR, 0.29-0.65] vs 0.45 [IQR, 0.15-0.54]; P = .002) reperfusion for which comparison with existing clinical methods was possible, the deep learning model had comparable or better performance. CONCLUSIONS AND RELEVANCE The deep learning model appears to have successfully predicted infarct lesions from baseline imaging without reperfusion information and achieved comparable performance to existing clinical methods. Predicting the subacute infarct lesion may help clinicians prepare for decompression treatment and aid in patient selection for neuroprotective clinical trials.

Novel Estimation of Penumbra Zone Based on Infarct Growth Using Machine Learning Techniques in Acute Ischemic Stroke

“…In all the external validation subjects, the presented ML model had a median Dice similarity coefficient (DSC) of 0.49 (IQR, 0.37–0.59), which was comparable to the DSC of 0.53 (IQR, 0.31–0.68) in the study by Yu et al [ 19 ]. The presented ML model had a median DSC of 0.43 (IQR, 0.20–0.52) in the SR external validation subjects and a median DSC of 0.58 (IQR, 0.55–0.67) in the UR external validation subjects.…”

Artificial Intelligence and Acute Stroke Imaging

“…55 A multicenter study showed that an attentiongated U-Net DL algorithm with DWI and MRP as inputs could predict final infarct volume regardless of reperfusion status, with a median AUC of 0.92 (IQR, 0.87-0.96) and significant overlap with the ground truth of a FLAIR sequence obtained 3-7 days after baseline presentation (Dice score, 0.53; IQR, 0.31-0.68). 56 The e-ASPECTS software was able to predict poor clinical outcomes after thrombectomy (Spearman correlation ¼ À0.15; P ¼ .027) and was an independent predictor of poor outcome in a multivariate analysis (OR, 0.79; 95% CI, 0.63-0.99) while also demonstrating high consensus with 3 expert ASPECTS readers (ICC ¼ 0.72, 0.74, and 0.76). 57 Traditional ML techniques combining clinical data and core-penumbra mismatch ratio derived from MR imaging and MRP to determine postthrombolysis clinical outcomes performed with an AUC of 0.863 (95% CI, 0.774-0.951) for short-term (day 7) outcomes and 0.778 (95% CI, 0.668-0.888) for long-term (day 90) outcomes.…”

Impact of the reperfusion status for predicting the final stroke infarct using deep learning

“…We used a U-Net architecture, a multi-scale network that has already shown its potential for infarct prediction tasks ( Winzeck et al, 2018 , Yu et al, 2020 ). Perfusion and diffusion MRI were used as inputs, as both modalities are complementary to evaluate the risk of infarction ( Barber et al, 1998 ).…”

“…A well-acknowledged limitation of CNNs is the large quantity of data required for their training and validation. Only a limited number of studies, with heterogeneous treatment paradigms and evaluations metrics, have evaluated CNNs for the prediction of the final stroke lesion from baseline MRI ( Winzeck et al, 2018 , Pinto et al, 2018 , Nielsen et al, 2018 , Yu et al, 2020 ) or CT ( Robben et al, 2020 ). Sample size and performance were modest ( 50 to 200 patients, Dice similarity coefficient 0.50 or lower), illustrating both the inherent difficulty of prediction tasks and scarcity of high-quality data, compared to simpler image segmentation tasks.…”