A Study on Self-Attention-Based ResNet and EfficientNet inBrain Tumor Detection

This paper explores Self-Attention-Based ResNet and EfficientNet models for brain tumordetection using MRI images. Through ۴-fold cross-validation, models are trained and evaluatedon both original and augmented datasets. The augmentation process includes random affinetransformations, color jitter adjustments, horizontal and vertical flips, and random perspectivetransformations. EfficientNet demonstrates improved accuracy on augmented data, whileResNet remains competitive. The study elucidates the mechanisms of these models, providinginsights into their effectiveness in medical imaging. The findings underscore the potential ofthese models for brain tumor detection

Keywords:

Brain tumor detection , Self-Attention-Based ResNet , EfficientNet , Medicalimaging.