Classification of Brain Tumors in MRI Images with Brain-CNXSAMNet: Integrating Hybrid ConvNeXt and Spatial Attention...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Brain tumors (BT) pose significant health risks, necessitating accurate early detection through magnetic resonance imaging (MRI).
• Classifying BT types is challenging due to cellular complexity, impacting treatment decisions.
• AI-powered systems are increasingly vital for efficient and accurate BT identification from MRI data.

Purpose of the Study:

• To develop and evaluate a novel hybrid deep learning model for classifying three common brain tumor types: meningioma, pituitary, and glioma.
• To enhance the accuracy and efficiency of automated brain tumor classification using MRI images.
• To leverage advanced deep learning architectures for improved diagnostic capabilities in neuro-oncology.

Main Methods:

• A hybrid model integrating ConvNeXt and a spatial attention mechanism (SAM) was developed.
• ConvNeXt was utilized to broaden the receptive field for capturing extensive spatial information.
• SAM was applied post-ConvNeXt to enable focused attention on critical image regions for improved classification.

Main Results:

• The hybrid model achieved high classification accuracies of 99.39% on the BSF dataset and 98.86% on the Figshare dataset.
• The proposed model demonstrated superior performance compared to recent studies.
• The model achieved these results with significantly reduced training periods, indicating high efficiency.

Conclusions:

• The hybrid ConvNeXt-SAM model represents a significant advancement in the automatic classification of brain tumors from MRI.
• The model offers a highly accurate and efficient solution for distinguishing between meningioma, pituitary, and glioma.
• This research highlights the potential of integrated deep learning approaches for improving diagnostic accuracy in neuroimaging.