A Novel Deep Learning Method for Recognition and Classification of Brain Tumors from MRI Images

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Computational neuroscience

Background:

• Brain tumors require early and accurate diagnosis for effective treatment and surgical planning.
• Manual segmentation of brain tumors from MRI scans is time-consuming, complex, and requires specialized expertise.
• Automated tumor detection faces challenges due to variations in tumor location, shape, and boundaries.

Purpose of the Study:

• To develop and evaluate a custom Mask Region-based Convolution neural network (Mask RCNN) with a densenet-41 backbone for precise brain tumor classification and segmentation.
• To improve the accuracy and efficiency of brain tumor detection compared to existing methods.

Main Methods:

• Implementation of a custom Mask RCNN architecture utilizing a densenet-41 backbone.
• Training the model using transfer learning on benchmark datasets.
• Evaluation using quantitative measures for segmentation and classification accuracy.

Main Results:

• The custom Mask RCNN model demonstrated high precision in detecting tumor locations via bounding boxes and generating accurate segmentation masks.
• Achieved an accuracy of 96.3% for segmentation and 98.34% for classification.
• Outperformed state-of-the-art approaches in robustness and accuracy.

Conclusions:

• The proposed Mask RCNN model offers a robust and accurate solution for automated brain tumor segmentation and classification.
• This approach can aid clinicians in surgical planning and treatment by providing precise tumor region identification.
• The study highlights the potential of deep learning in advancing neuro-oncology diagnostics.