Abstract Body: Introduction:
Ischemic core growth rate in patients with a large vessel occlusion varies during transfer from a primary stroke center (PSC) to a comprehensive stroke center (CSC) for thrombectomy. Patients at risk for a high growth rate may be optimal candidates for neuroprotective treatments. We determined if deep learning based CT lesion (DLCT) volume and density measures are associated with growth rate.
Methods:
Patients with CT and CTP at the PSC and CTP or DWI at the CSC were included from the prospective CRISP 2 study. Growth rate per hour (mL/h) was measured as the difference between the CTP core volume (relative cerebral blood flow <30%) at the PSC and CTP ischemic core or manually segmented DWI lesion volume at the CSC. We used a validated deep learning model to segment the ischemic lesion on non-contrast CT and determined lesion volume, density in the lesion, and relative hypodensity; the percentage difference with contralateral voxels. Additional clinical, CT angiography, and CTP measurements were considered as independent variables (Figure 1). Stepwise regression with combined forward/backward selection optimized for the Akaike Information Criterion was used to select variables. For each variable, we report adjusted regression slopes with 95% confidence interval (aB[95%CI]) for their association with growth rate. R2 and adjusted R2 are reported to reflect model performance.
Results:
We included 269 patients; 113(42%) women, median age was 71(IQR:60-80) years, 185(69%) had witnessed onset of stroke symptoms. The stepwise optimized regression model had an R2 of 0.44 (adjusted R2: 0.41). DLCT variables with significant associations with growth rate were volume (aB/10mL:3.2[95%CI:2.3; 4.1]), >20% relative hypodensity volume (aB/10mL:-8.5[-11.6; -5.3]), and density standard deviation in the lesion (aB/HU:2[0.8; 3.2]). Penumbra volume (aB/10mL:-0.5[-0.8; -0.2]) was associated with core growth rate, while ischemic core volume (aB/10mL:0.62[-0.03; 1.27]) was not. Tan collaterals score (aB per point:-4.7[-6.6; -2.9]), clot burden score (aB/point:-1.6[-2.1; -1.0]), female sex (aB:-3.2[-5.7; -0.7]), and admission systolic blood pressure (aB/10mmHg:0.7[0.3; 1.2]) were also significantly associated with core growth rate.
Conclusion:
DLCT based volume and density measures have a strong association with ischemic core growth rate during interhospital transfer which might indicate that DLCT also identifies tissue at risk for irreversible ischemic damage.