Abstract 023: Quantification Of White Matter Lesion Volume From Clinical T1 Weighted Magnetic Resonance Images To Predict Functional Outcomes Post-thrombectomy

Abstract

Background: Endovascular treatment (EVT) has immense benefits for patients with large vessel occlusion (LVO). However, many patients do not achieve functional independence despite successful EVT. This may be due to concurrent cerebral small vessel disease (CSVD) mitigating recovery. CSVD manifests as white matter hypointensities on T1-weighted (T1w) MRI and is usually qualitatively assessed from clinical-quality MRI, which does not allow for objective quantification. Objectives: Implement a processing pipeline to quantify white matter hypointensity volume (WMHV) from clinical images and examine WMHV's influence on functional outcomes post successful EVT. Methods: We performed a retrospective analysis of the Atrium Wake Forest Stroke Thrombectomy and Aneurysm Registry (n=602) collected between 2015 - 2021. We selected patients with LVO who underwent successful EVT, defined as a Thrombolysis in Cerebral Infarction score of 2B or better. Clinical T1w images were transformed into high-resolution images using the convolutional neural network SynthSR. Then, FreeSurfer was used to quantify baseline WMHV from the side of the brain contralateral to the stroke to minimize stroke interference. To correct for head size, WMHV was adjusted to the estimated total intracranial volume and then log-transformed to address skewness. Results: The analysis included 213 patients (mean age 67.5 ± 14.6, 49.3% (105 of 213) female) who had MRI of sufficient quality for assessment and 90-day mRS. Baseline WMHV was significantly predictive of 90-day mRS in an ordinal regression model adjusted for baseline mRS (p<0.001). After adjusting for confounders and comorbidities, WMHV remained an independent predictor of 90-day mRS (p<0.001). Conclusions: Increased volume of white matter hypointensities correlates with poorer functional outcomes after mechanical thrombectomy and may attenuate EVT benefit. Furthermore, advances in neural networks enable the quantification of cerebral small vessel disease from clinical T1w MRI. Such approaches expand the utility of clinical imaging for research purposes.