278 Intraoperative Anastomotic Flow Measurements During STA-MCA Bypass Using FLOW800 Hemodynamics

INTRODUCTION: Accurate intraoperative flow measurements are important for cerebral revascularization surgeries. Traditionally, these require direct manipulations of the vessels to measure with transonic flow probes. This study evaluates the feasibility of estimating flow using ICG-videoangiography-FLOW800 technology – a hands-off method of quantifying hemodynamics. METHODS: We conducted a retrospective analysis of 39 cases, assessing FLOW800 features (Speed, Delay Time, Rise Time, Time to Peak) from three Regions of Interests (Donor, Proximal and Distal recipient vessels) with respect to transonic flow probe rates. We engineered additional features by computing feature ratios and interactions (multiplication) between ROIs. Statistical models including General Linear Models (GLM), Random Forest (RF), and Multi-Layer-Perceptron (MLP) were used to estimate flow probe rates using these features. We extracted the models feature importance with a threshold of 0.01 and used Shapiro-Wilk Test to determine goodness of fit for each model. RESULTS: The combination of original and engineered features with statistical models demonstrated success (Shapiro-Wilk Test p>0.05) in estimating flow rates, with the best results achieved using RF (Root-Mean-Squared-Error=13.71, 5-fold cross validation). The model was best estimated the true blood flow when assessing flows in the range of 25-45mL/min, which was also the typical range of blood flow post-revascularization. Below or above this range, the model overestimated or underestimated flow, respectively. All three models used Proximal and Distal Rise Time as important features for calculation, and Takeoff Time-to-Peak were important for RF model. CONCLUSIONS: With appropriate mathematical manipulation and statistical models, we suggest that FLOW800 features can be utilized to estimate the post-revascularization blood flow measured by transonic flow probes, despite the different aspects of hemodynamics being captured by each method. This study highlights the utility of FLOW800 as an alternative approach to estimate cerebrovascular flow without direct vessel manipulation.