Acute Ischemic Brain Stroke Classification Using Attention-Augmented Ensemble Approach

Acute Ischemic Brain Disorder, also known as an (AIS) Acute Ischemic Stroke, takes place when a blood clot blocks a brain blood vessel, leading to oxygen deprivation and potential brain damage. Clinical analysis techniques with the help of Computer Tomography (CT) scans with Magnetic Resonance Imaging (MRI) helps in visualizing the brain and in the assessment of blockage related characteristics. This procedure is not always quick and accurate for the detection of mild, atypical or the variety of other patterns related to symptoms present in an MRIs. This research is an attempt to address these limitations with the development of AI-based automated deep learning technique. In this work, an ensemble consists of fine-tuned ResNet50 and EfficientNetB0 models and the output obtained from these models present in an ensemble are passed to the self-attention module which helps in focusing the relevant features. Finally, these features were used for the classification of MRI images affected with the Acute Ischemic brain stroke. The experimental results imply that the proposed method has compassed a classification accuracy of 95.00% in classifying the brain stroke MRI images which proves the effectiveness of the proposed method in the classification of Acute Ischemic stroke.