Abstract: Acute ischemic stroke is one of the fatal brain dysfunction diseases caused by the interruption of blood supply to the brain tissue. Digital Subtract Angiography (DSA) is the gold standard for diagnosing such cerebrovascular diseases. Based on the frontal and lateral DSA images of the patients, a dual-path image classification intelligent model, Dual-Path Vision Transformer (DPVF), is constructed in this paper to evaluate the treatment effectiveness of acute ischemic stroke in a graded manner. In order to improve the speed of auxiliary diagnosis, the model is constructed based on the lightweight design idea of EdgeViT. And in order to make the model have high accuracy, the spatial-channel self-attention module is proposed to promote the transformer model to capture more comprehensive feature information and improve the model representation. In addition, for the feature fusion of two branches of DPVF, a cross-attention module is constructed to cross-fuse the outputs of the two branches, which promotes the model to extract richer features and thus improves the model performance. The experimental results show that the accuracy of DPVF on the test set reaches 98.5%, which can effectively meet the practical requirements.