Analysis of DL-Algorithms for Segmentation of Tumor

Precisely identifying brain tumors in MRI images is essential for medical image analysis, as it significantly aids ‎in diagnosis, therapy planning, and tracking disease progression. Despite the rapid advancement of Deep ‎learning methods, especially neural networks, which have added prominence for their effectiveness in image-related tasks, a systematic comparative analysis of leading DL architectures using the Figshare BTMR dataset ‎is still lacking. This study addresses this gap by experimentally evaluating six prominent DL models—VGG16, ‎VGG19, ResNet18, ResNet50, MobileNet, and Xception—used as backbone encoders for brain tumor ‎segmentation. Each model’s performance is objectively measured using“metrics such as Mean Intersection ‎over Union (mIoU) and Mean Dice Similarity Coefficient”(mDSC). The findings provide valuable insights into ‎the balance between segmentation accuracy and efficiency, assisting in the selection of suitable models for real-time and clinical implementation. This work serves as a benchmark reference, guiding future efforts in selecting ‎effective DL-based architectures for brain tumor segmentation tasks‎.