Utilizing Deep Learning Via The 3d U-net Neural Network For The Delineation Of Brain Stroke Lesions In Mri Image

This article introduces a novel deep fully convolutional neural network model designed for segmenting stroke lesions using MRI images. The proposed model employs a deep learning algorithm to focus on decrypting the lesion zone. The suggested technique utilizes a fully convolutional neural network for semantic image segmentation, which can be integrated into treatment decision workflows to quickly estimate the locations of lesion cores. The dataset used for this study was the Sub-Acute Ischemic Stroke Lesion Segmentation (SISS) challenge, which is a subset of the larger Ischemic Stroke Lesion Segmentation (ISLES) dataset. All 3D MRIs were transformed into 2D image segments along the axial direction for implementation of the segmentation algorithm.