Employing Xception convolutional neural network through high-precision MRI analysis for brain tumor diagnosis.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Medicine

Background:

• Accurate brain tumor classification from medical imaging is crucial but challenging.
• Current methods using machine learning and deep learning often face overfitting and limited generalizability on small datasets.
• Computational demands hinder real-time application of existing diagnostic tools.

Purpose of the Study:

• To develop an advanced AI model for enhanced brain tumor classification accuracy and reliability.
• To overcome limitations of existing models, including overfitting and poor generalizability.
• To create a robust diagnostic tool for distinguishing between meningioma, glioma, and pituitary tumors.

Main Methods:

• Utilized the Xception architecture with added batch normalization and dropout layers to mitigate overfitting.
• Employed transfer learning with large-scale data to leverage pre-trained network features.
• Implemented a customized dense layer setup for specific classification of tumor types.

Main Results:

• Achieved a classification accuracy of 98.039% on the test dataset.
• Demonstrated precision and recall rates exceeding 96% for all tumor categories.
• Showcased improved generalization capacity across different imaging conditions.

Conclusions:

• The proposed hybrid AI model significantly enhances brain tumor diagnostic performance.
• The model shows potential as a reliable clinical tool, surpassing current diagnostic protocols.
• This approach offers a more accurate and generalizable solution for brain tumor classification.