To determine whether the use of Machine Learning Model(MLM) would provide a more accurate prediction of morbidity and mortality in Acute Ischemic Stroke (AIS) patients immediately after Thrombectomy.

Clinical and radiological predictors of prognosis of acute ischemic stroke survivors have been extensively studied. Machine Learning (ML) has facilitated the quantification of the area of damage after stroke and improved prediction of Morbidity and Mortality. Using Machine Learning (ML) in Clinical and radiological practices led to the more sensitive diagnosis of cerebrovascular diseases and improved prediction of mortality in stroke patients. Prognostication immediately after thrombectomy shall help planning the treatment follow-up, clinical management and keep the patients well-informed. We devised a novel machine learning model(MLM) to predict dichotomized 90-day mRS (0-1, >=2) immediately after thrombectomy.

Forty acute ischemic stroke patients underwent Stroke Lesion Segmentation (ISLES data of 2015 and 2017) on Diffusion MRI. Input variables chosen are Thrombolysis in Cerebral Infarction Scale (TICI) score after Thrombectomy, Infarct and Penumbra lesions on ADC MRI scans at the onset. 744 Radiomics features were extracted from Infarct and Penumbra lesions. 744 Radiomics features along with TICI are provided as input to the Machine Learning Model(MLM). MLM comprises of LDA(Linear Discriminant Analysis) and Logistic Regression Model. The predictions are discussed in the results below.

The Machine Learning model(MLM) correctly classified all patients with Accuracy of 75% with F1-score of 70% as non-severe (mRS=0-1) and severe (mRS >=2) patient outcomes. The Machine Learning Model(MLM) would provide an improved prediction of Morbidity and Mortality in AIS patients immediately after Thrombectomy assisting clinical management.