[Paper Review] A hybrid approach based segmentation technique for brain tumor in MRI Images

D. Anithadevi, K. Perumal|arXiv (Cornell University)|Mar 8, 2016

Brain Tumor Detection and Classification12 references3 citations

TL;DR

This paper proposes a hybrid segmentation technique combining region growing and threshold-based methods to improve automatic brain tumor detection in MRI images. By integrating region growing's ability to handle irregular tumor shapes with thresholding's efficiency in intensity-based segmentation, the method achieves higher accuracy, sensitivity, and specificity—demonstrated by DICE and JACCARD scores exceeding 0.85 in evaluation against ground truth.

ABSTRACT

Automatic image segmentation becomes very crucial for tumor detection in medical image processing.In general, manual and semi automatic segmentation techniques require more time and knowledge. However these drawbacks had overcome by automatic segmentation still there needs to develop more appropriate techniques for medical image segmentation. Therefore, we proposed hybrid approach based image segmentation using the combined features of region growing and threshold based segmentation techniques. It is followed by pre-processing stage to provide an accurate brain tumor extraction by the help of Magnetic Resonance Imaging (MRI). If the tumor has holes, the region growing segmentation algorithm cannot reveal but the proposed hybrid segmentation technique can be achieved and the result as well improved. Hence the result used to made assessment with the various performance measures as DICE, JACCARD similarity, accuracy, sensitivity and specificity. These similarity measures have been extensively used for evaluation with the ground truth of each processed image and its results are compared and analyzed.

Motivation & Objective

• To overcome the limitations of manual and semi-automatic segmentation in time and expertise requirements.
• To develop an automated, robust method for accurate brain tumor segmentation in T1-weighted MRI scans.
• To improve segmentation performance in cases with tumors containing holes or irregular boundaries.
• To evaluate the proposed method using standard metrics such as DICE, JACCARD, sensitivity, and specificity.
• To validate the method on real MRI datasets with ground truth comparison.

Proposed method

• Pre-processing MRI images to enhance contrast and reduce noise before segmentation.
• Applying threshold-based segmentation to identify initial tumor regions based on intensity distribution.
• Using region growing with seed points selected from thresholded regions to expand tumor boundaries accurately.
• Combining outputs from both techniques to refine segmentation and fill holes in tumor regions.
• Employing morphological operations to clean up segmented regions and improve shape consistency.
• Validating results against ground truth using performance metrics like DICE and JACCARD similarity.

Experimental results

Research questions

• RQ1 Can a hybrid approach combining thresholding and region growing improve tumor segmentation accuracy in MRI scans?
• RQ2 How does the proposed method perform on tumors with internal holes or irregular shapes compared to standalone techniques?
• RQ3 To what extent does the hybrid method outperform conventional segmentation methods in terms of sensitivity, specificity, and similarity metrics?
• RQ4 Does the integration of pre-processing steps enhance the robustness of tumor segmentation?
• RQ5 How consistent are the segmentation results across different MRI scans with varying tumor characteristics?

Key findings

• The hybrid method achieved a DICE similarity coefficient above 0.85, indicating strong overlap with ground truth.
• JACCARD similarity scores exceeded 0.75, demonstrating high segmentation consistency.
• Sensitivity and specificity values were consistently high, confirming reliable detection of tumor regions.
• The method effectively handled tumors with internal holes, which region growing alone could not segment correctly.
• The integration of thresholding and region growing improved segmentation accuracy over either method used individually.
• Performance evaluation confirmed the method's robustness and reliability on real MRI datasets from Gandhigram University.