Analysis of Brain MRI Images Using Improved CornerNet Approach

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Neuroscience

Background:

• Brain tumors are a serious disease requiring timely and precise detection for effective treatment and surgical planning.
• Manual brain tumor detection is labor-intensive, time-consuming, and relies heavily on expert availability.
• Automated systems are crucial for accurate and efficient brain tumor detection and classification due to variations in tumor size, position, and structure.

Purpose of the Study:

• To develop an accurate and automated system for the detection and classification of brain tumors.
• To address the challenges of exact localization and categorization of brain tumors.
• To improve upon existing methods for brain tumor diagnosis.

Main Methods:

• A novel DenseNet-41-based CornerNet framework was developed.
• The framework involves initial annotation for region of interest identification.
• Deep features are extracted using a custom CornerNet with DenseNet-41, followed by tumor localization and classification using a one-stage detector.

Main Results:

• The proposed method achieved an average accuracy of 98.8% on the Figshare dataset.
• An average accuracy of 98.5% was attained on the Brain MRI dataset.
• Both qualitative and quantitative analyses demonstrated the approach's proficiency and consistency.

Conclusions:

• The DenseNet-41-based CornerNet framework offers a proficient and consistent solution for brain tumor detection and classification.
• The automated system outperforms other state-of-the-art techniques in accuracy and reliability.
• This approach can significantly aid in early diagnosis and surgical planning for brain tumors.