Abstract 133: FACET: Fractal Angiography for Continuous Revascularization Evaluation during Thrombectomy

Abstract

Introduction: The technical success of endovascular therapies in stroke may be visually evaluated with the Thrombolysis in Cerebral Ischemia score (0,1, 2A, 2B, 3), ranging from absent to complete downstream reperfusion. Subtle but possibly relevant reperfusion differences may not be detected with the observer dependent TICI score. The development of a quantitative and continuous measure to assess revascularization would provide a powerful tool to evaluate efficacy of clot retrieval. We present FACET, a fractal-based computational framework that can infer a continuous measure of revascularization from DSA. Methods: The DSA dataset used in this study was established retrospectively from patients treated with endovascular therapy for acute stroke. An expert reader manually delineated the MCA territory on the AP view of the angiogram using a region of interest (ROI) and TICI scores were obtained by consensus between 3 readers. The computational framework then processed each ROI to first extract the vessels using a vessel detector, followed by a skeletonization of the output responses that was used to compute the fractal dimension (FD). Evaluation was performed using nonlinear regression model to test the ability of the model in predicting degree of revascularization. Results: Among 97 patients included, 61 (63%) were female, mean age was 65.8 (range 22-93). TICI scores were distributed as [0:14, 1:5, 2a:31, 2b:40, 3:7]. FACET responses strongly correlated with TICI (r=.66, p <0.001) mapped to a continuous scale. In multivariate analysis, TICI 0 did not differ significantly from TICI 1 (p=.94) and 2a (p=.32), but did differ from group 2b (p<.0001) and 3 (p<.0001). Similarly, TICI 1 did not differ from TICI 2a (p=.46), but 2b (p<.0001) and 3 (p<.0001). Significant difference was found between TICI 2a and 2b, 3; but not between 2b and 3. Despite strong correlation, multivariate analysis of variance between TICI groups indicate non-strict equivalence between the TICI and FACET. Conclusions: FACET, or fractal analysis of DSA provides a continuous and objective measure that mirrors the TICI score and enables semi-automated high-throughput analysis of angiographic reperfusion.