A Hybrid Deep Learning Model with Data Augmentation to Improve Tumor Classification Using MRI Images.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Computational Neuroscience

Background:

• Cancer is a leading global cause of mortality, with brain cancer having a particularly low survival rate.
• Accurate brain tumor classification is crucial for effective treatment planning and patient outcomes.
• Deep learning models are increasingly utilized for disease diagnosis, with a focus on optimizing neural network performance.

Purpose of the Study:

• To develop and evaluate a novel hybrid deep learning approach for enhanced brain tumor classification.
• To improve the accuracy and efficiency of diagnosing various brain tumor types using medical imaging.

Main Methods:

• A hybrid model was developed, integrating Convolutional Neural Networks (CNNs) and EfficientNetV2B3 for robust feature extraction.
• K-nearest neighbors (KNN), a supervised learning algorithm, was employed for the final classification of extracted features.
• The model's performance was enhanced by concatenating flattened outputs from CNN and EfficientNetV2B3 before KNN classification.

Main Results:

• The proposed hybrid framework was tested on two benchmark MRI datasets, including one with 3064 images of meningiomas, gliomas, and pituitary tumors.
• The second dataset comprised images of healthy brains and brain tumors.
• The model achieved exceptional accuracy rates of 99.51% and 99.8% on the respective datasets.

Conclusions:

• The novel hybrid approach effectively enhances brain tumor classification accuracy.
• Concatenating features from CNN and EfficientNetV2B3 with KNN provides a powerful diagnostic tool.
• This method shows significant potential for improving clinical diagnosis and patient prognosis in neuro-oncology.