Ischemic Stroke Segmentation in FLAIR MRI using Modified U-Net Model
E Ruthra, A. Ruhan Bevi · 2023
An abrupt stoppage of blood flow to the brain characterizes an ischemic stroke. This potentially fatal medical condition presents a significant challenge for global healthcare systems because of its high death and long-term disability rates. The timely intervention that follows an ischemic stroke is essential for achieving substantial improvements in patient outcomes. Diagnostic medical imaging methods include Computed Tomography (CT) and Magnetic Resonance Imaging (MRI). Images from Fluid-Attenuated Inversion Recovery (FLAIR) are handy for evaluating stroke patients. This research study presents an altered U-Net architecture for ischemia segmentation based on deep learning to improve diagnosis accuracy and speed. The study demonstrated its efficacy in ischemic stroke segmentation with a Dice coefficient of 0.8890 during training and 0.8244 during validation. Several metrics were used to assess the system's performance, and it beat other U-Net models and attention-based deep neural networks, as well as other methods already in use. The results show that this model has the potential to improve the diagnosis of ischemic strokes due to its high levels of accuracy, sensitivity, specificity, and Dice coefficient. Subsequent research endeavors could encompass refining the system to classify lesions according to diverse attributes and augmenting its usefulness in medical imaging applications. This paper introduces a novel approach for converting NIfTI files into PNG images to enhance the accessibility and visualization of medical imaging data.