Improving Minimum Cross-Entropy Thresholding for Segmentation of Infected Foregrounds in Medical Images Based on Mean...

Area of Science:

• Computer Vision
• Medical Image Analysis
• Image Processing

Background:

• Mean-based thresholding is crucial for image segmentation, separating image intensities effectively.
• Minimum Cross-Entropy Thresholding (MCET) is a popular mean-based method but relies on a classical mean.
• Classical means can be sensitive to noise and outliers, potentially limiting segmentation accuracy.

Purpose of the Study:

• To enhance the Minimum Cross-Entropy Thresholding (MCET) method for improved image segmentation.
• To introduce and evaluate novel mean estimation approaches for MCET.
• To demonstrate the effectiveness of the proposed methods on medical imaging datasets.

Main Methods:

• Proposed alternative mean estimation techniques including alpha trim, harmonic, contraharmonic, and geometric means.
• Integrated these dedicated mean estimators with the MCET algorithm.
• Applied the enhanced MCET to segment three medical image datasets: Alzheimer's MRI, brain tumor MRI, and skin lesions.
• Evaluated segmentation performance using both unsupervised and supervised metrics.

Main Results:

• The proposed mean estimation approaches effectively mitigate the negative impact of noise and outliers in mean computation.
• The enhanced MCET method demonstrated improved accuracy in segmenting medical images compared to the classical MCET.
• Accurate segmentation results were achieved across diverse medical imaging datasets, including MRI and skin lesion images.

Conclusions:

• Dedicated mean estimation approaches significantly improve the performance of Minimum Cross-Entropy Thresholding for image segmentation.
• The proposed technique offers a robust solution for accurate medical image segmentation.
• The method shows potential for extension to other applications like object detection.