A region-based segmentation of tumour from brain CT images using nonlinear support vector machine classifier

Area of Science:

• Medical Imaging
• Computer Vision
• Machine Learning

Background:

• Manual segmentation of brain tumors from computed tomography (CT) images is time-consuming and requires expert medical knowledge.
• Accurate tumor segmentation is crucial for diagnosis, treatment planning, and monitoring of brain tumors.
• Existing segmentation methods may struggle with smaller tumor dimensions and require significant computational resources.

Purpose of the Study:

• To develop an efficient and accurate system for segmenting benign and malignant brain tumors from CT images.
• To enhance tumor segmentation by combining grey, texture, and novel edge features.
• To evaluate the performance of a nonlinear support vector machine (SVM) classifier for this task.

Main Methods:

• A region-based segmentation approach was employed using combined grey, texture, and new edge features.
• A nonlinear support vector machine (SVM) classifier was modeled and trained with selected optimal features.
• The proposed method was applied to 80 real-world benign and malignant tumor CT images.

Main Results:

• The system achieved efficient and accurate segmentation of tumors, particularly in smaller dimensions.
• Quantitative analysis using segmentation accuracy and the Dice metric showed superior performance compared to fuzzy c-means clustering.
• The normalized cut segmentation method demonstrated higher segmentation accuracy and a better Dice metric.

Conclusions:

• The proposed method effectively segments brain tumors from CT images with improved accuracy and efficiency.
• Combining diverse features and employing a nonlinear SVM classifier enhances tumor segmentation capabilities.
• This approach offers a promising tool for medical experts in analyzing CT scans and improving patient care.