Non-invasive grading of brain tumors using online support vector machine with dynamic fuzzy rule-based parameters...

Area of Science:

• Medical imaging analysis
• Artificial intelligence in oncology
• Neurosurgery and neurology

Background:

• Non-invasive brain tumor grading is crucial for treatment planning.
• Accurate tumor segmentation and grading from MR images remain challenging.

Purpose of the Study:

• To propose a fully automated method for brain tumor grading using MR images.
• To develop an innovative, fast tumor segmentation technique.
• To implement an online support vector machine with dynamic fuzzy rule-based optimization for grading.

Main Methods:

• Tumor segmentation using intensity and edge information.
• Feature extraction from segmented tumor regions.
• Online Support Vector Machine with Kernel (OSVMK) and dynamic fuzzy rule-based parameter optimization for grading.

Main Results:

• The proposed segmentation method showed good correlation with manual segmentation.
• The automated grading achieved high performance: 95.20% accuracy, 97.87% precision, 96.48% recall, and 96.45% specificity.
• The online method demonstrated significantly reduced execution times compared to batch methods.

Conclusions:

• The developed method offers a potential for fully automated, non-invasive brain tumor grading.
• This approach can assist physicians in determining optimal, individualized treatment strategies for brain tumor patients.