Deep Learning-Based Noncontrast CT Imaging Markers Enhance Posttransfer DWI Core Volume Prediction

Deep learning enables the extraction of ischemic lesion size and hypodensity imaging markers from deep learning-based NCCT (DLNCCT) assessment in patients with acute ischemic stroke, but it remains unclear whether those markers can predict posttransfer core volume.

We performed a post hoc analysis of prospectively enrolled patients transferred from a primary stroke center (PSC) to a comprehensive stroke center (CSC) for endovascular treatment (EVT). Using a validated DLNCCT segmentation method, we quantified total lesion volume (per 10 mL), modified net water uptake (mNWU %), and severely hypodense volume (≤26 Hounsfield units [HU] per 10 mL) and compared these markers with core lab rated ASPECTS (per point decrease) and CTP-based evaluation for their association with (adjusted regression coefficient [95% CI]) and predictive performance in addition to baseline variables (R ² ± SEs) for posttransfer CSC admission DWI core volume.

We included 420 patients (239 [57%] men) with a median age of 72 years (interquartile range: 61-80). We observed 11.2 mL (95% CI, 8.3-14.1] larger posttransfer core volumes per point decrease in ASPECTS, and 10.0 mL (95% CI, 6.8-13.3), and 20.0 mL (95% CI, 12.7-27.2) larger posttransfer core volumes per 10-mL increase in total and severely hypodense DLNCCT volume, respectively. mNWU was not associated with posttransfer core volume (P = .63). In addition to clinical baseline and CTA variables, posttransfer core volume prediction with ASPECTS (R ²: 0.49 ± 0.02) and DLNCCT (R ²: 0.50 ± 0.02) did not differ significantly (P = .58). Compared with using CTP imaging markers (R ²: 0.56 ± 0.02), adding ASPECTS (R ²: 0.63 ± 0.02; P < .01) and DLNCCT (R ²: 0.65 ± 0.01; P < .01) improved performance for posttransfer core volume prediction.

Total and severely hypodense DLNCCT volumes are independent predictors for posttransfer core volume. These DLNCCT markers improved CTP-based posttransfer core volume prediction.