TY - JOUR ID - TI - Pre Diagnosing the Stroke Using Deep Learning AU - Makarem S. Atshan , Zainab N. Nemer PY - 2022 VL - 14 IS - 3 SP - 33 EP - 43 JO - Journal of Al-Qadisiyah for Computer Science and Mathematics مجلة القادسية لعلوم الحاسوب والرياضيات SN - 20740204 25213504 AB - Stroke is a long-term disability that affects people all around the world. A stroke deprives the brain of oxygen and nutrients, causing brain cells to stop working or die. Radiologists classify them using Magnetic Resonance Imaging(MRI) of the brain for people with stroke. We've presented a solution based on Deep Learning Algorithms such as Convolution Neural Networks (CNN) and Transfer Learning (TL). We divided MRI into two categories in this work (stroke and normal). In this paper, a 4-layer CNN is used from beginning to end with a modification of the EfficientNetB0 network by adding two layers to make the extracted features more diverse. CNN and pre-trained EfficientNetB0 were used to train our architecture. The EfficientNetB0 achieves an accuracy of 99.6%, while CNN achieves an accuracy of 99.2%. The newly created EfficientNetB0 could be a highly important decision-making tool in stroke research and brain diagnostic testing. As a physician's assistant, the proposed method works.

ER -