Beyond Recanalization: Machine Learning-Based Insights into Postthrombectomy Vascular Morphology in Patients with Stroke

Abstract

Many patients with stroke have poor outcomes despite successful endovascular therapy (EVT). We hypothesized that machine learning (ML)-based analysis of vascular changes post-EVT could identify macrovascular perfusion deficits such as residual hypoperfusion and distal emboli. Patients with anterior circulation large vessel occlusion stroke, pre- and post-EVT MRI, and successful recanalization (modified TICI 2b/3) were included. An ML algorithm extracted vascular features from pre- and 24-hour post-EVT MRA. A ≥100% increase in ipsilateral arterial branch length was considered significant. Perfusion deficits were defined by using PWI, MTT, or distal clot presence and early neurologic improvement (ENI) by a 24-hour NIHSS decrease ≥4 or NIHSS 0-1. Among 44 patients (median age 63 years), 71% had complete reperfusion. Those with distal clot had smaller arterial length increases (51% versus 134%, P = .05). ENI patients showed greater arterial length increases (161% versus 67%, P = .023). ML-based vascular analysis post-EVT correlates with perfusion deficits and may guide adjunctive therapy.