Quantifying the margin sharpness of lesions on radiological images for content-based image retrieval

Area of Science:

• Medical Imaging
• Radiology
• Image Analysis

Background:

• Accurate quantification of lesion margin characteristics is crucial for medical image analysis.
• Existing methods for lesion margin characterization may lack precision and robustness.

Purpose of the Study:

• To develop and evaluate a novel method for quantifying lesion margin sharpness in computed tomography (CT) images.
• To assess the method's performance in simulated data and clinical scans of liver and lung lesions.

Main Methods:

• Computed lesion margin sharpness using intensity difference and transition sharpness from fitted sigmoid curves.
• Extracted feature vectors from histograms of these parameters.
• Validated the method using simulated CT scans and clinical datasets of liver and lung lesions, comparing against existing techniques and subjective assessments.

Main Results:

• Achieved high concordance correlation (0.994) for margin sharpness in simulated images.
• Demonstrated superior performance in image retrieval tasks (mean normalized discounted cumulative gain scores) compared to existing methods for both liver and lung lesions.
• Showed enhanced robustness against lesion margin deformations (p < 0.05).

Conclusions:

• Introduced a novel image feature for quantifying lesion margin sharpness on CT.
• The feature exhibits strong correlation and robust performance in simulated and clinical datasets.
• Potential applications include content-based image retrieval in medical imaging.