To determine the use of the Machine Learning Model (MLM) to provide an accurate prediction of functional outcome in patients with Acute Ischemic Stroke (AIS) in terms of dichotomized 7-day modified Rankin Scale (mRS).

In patients with acute ischemic stroke secondary to Large Vessel Occlusion (LVO) functional outcomes can be assessed using clinical and imaging criteria with the help of MLM. We devised an improved MLM to predict dichotomized 7-day mRS (0-1, >=2) using neuroimaging scores.

Computerized Tomographic Angiography (CTA) scans of AIS patients with LVO admitted in a tertiary care hospital, have been graded for Clot quantification metrics-clot burden score (CBS) and Collateral quantification metrics- regional leptomeningeal score (rLMC), MAAS, MITTEFF, and TAN. Various clinical parameters, and NIHSS were correlated with functional outcomes. The neuroimaging scores and clinical parameters were combined in the prediction model to construct MLM.

CTAs of 42 AIS subjects were Analyzed (Age, 64.1+/-15.7; NIHSS 16[1QR, 0-22]; ASPECTS 8[IQR 6-10]). Machine learning models ensemble of support vector machine (SVM), random forest (RF) and neural network (NN) demonstrated that neuroimaging scores-rLMC, MAAS and TAN are effective in prognostication of dichotomous mRS at 7 days with an accuracy of 84.75% without clinical scores/parameters, whereas 83.40% with clinical scores/parameters.

The machine learning model based on objective imaging criteria of CBS and Cerebral collaterals yielded a robust prognostication model in AIS subjects with LVO than subjective scores based on clinical parameters and NIHSS which can be less robust.