Improved brain tumor classification through DenseNet121 based transfer learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Early and accurate brain tumor diagnosis is critical for effective treatment.
• Current diagnostic methods are often time-consuming and rely on manual interpretation.
• Limitations in existing methods necessitate advanced automated solutions.

Purpose of the Study:

• To develop an automated system for brain tumor classification using MRI data.
• To leverage DenseNet121 architecture with transfer learning for improved diagnostic accuracy.
• To enhance the speed and precision of brain tumor detection and classification.

Main Methods:

• Utilized the DenseNet121 architecture with transfer learning for brain tumor detection.
• Trained the model on a Kaggle dataset of MRI scans.
• Preprocessed MRI images, including resizing, to minimize noise and improve model performance.
• Classified brain tissues into four categories: benign tumors, gliomas, meningiomas, and pituitary gland malignancies.

Main Results:

• Achieved an average accuracy of 96.90% in multi-class brain tumor classification.
• Demonstrated superior performance compared to manual interpretation and other machine learning models.
• The DenseNet121 approach offers increased accuracy, reduced analysis time, and minimal human input.

Conclusions:

• The proposed automated method significantly improves the accuracy and speed of brain tumor diagnosis from MRI scans.
• This approach offers a consistent and predictable alternative to conventional diagnostic techniques, reducing the impact of human error.
• The developed deep learning model holds promise for advancing MRI-based classification research and clinical applications in neuro-oncology.