A DBSCAN based approach for jointly segment and classify brain MR images.

Area of Science:

• Medical Imaging
• Image Processing
• Computational Biology

Background:

• Magnetic Resonance (MR) image segmentation is crucial for various applications.
• Existing techniques often rely on gray-level intensity, limiting robustness.

Purpose of the Study:

• To present a novel MR image segmentation approach.
• To enhance segmentation accuracy and robustness using advanced techniques.

Main Methods:

• The novel method utilizes estimated proton density and relaxation times per pixel, not gray-level intensity.
• An evolved version of the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is implemented.
• The approach is compared against an Euclidean distance-based method.

Main Results:

• The algorithm demonstrates robustness due to its feature exploitation.
• The evolved DBSCAN approach improves region estimation effectiveness.
• The proposed technique shows a higher correct classification rate compared to Euclidean distance methods.

Conclusions:

• The novel segmentation approach offers improved accuracy and robustness for MR images.
• The method's effectiveness is validated on a simulated dataset.
• Future work includes extending the validation to real-world MR imaging data.