Brain tumor segmentation using DeepAttRes-Unet based transfer learning

Early-stage brain tumor identification is a difficult issue that is well handled by machine learning (ML) and deep learning (DL) classifiers. In order to identify the best classifier, a comparison analysis of the various ML classifier types is conducted in this work.Transfer learning allows us to reuse previously taught models. Transfer learning refers to applying previously acquired information to new activities.The identification and detection of brain tumors with MRIs is tough to perform manually. When done manually, detecting and classifying brain tumors with MRIs is a difficult process. Automation of these processes has become necessary. However, developing deep learning models from scratch on large datasets may be computationally intensive. This research offers a transfer learning for image segmentation and prediction of tumors. We utilized models and formed an Unet to segment the tissues in this work. We utilized the LGG dataset, which has 110 patient records, to conduct segmentation. We developed a unique residual model as the foundation for creating the unet from scratch and doing multispectral image segmentation. The suggested model works well, and its performance is measured by the Dice coefficient, Jaccard index, and IoU. The model produces an accuracy mean of 0.99 and a Dice coefficient mean of 0.8428.