Ground truth delineation for medical image segmentation based on Local Consistency and Distribution Map analysis

Area of Science:

• Medical Imaging
• Computer-Aided Detection (CAD)
• Image Segmentation

Background:

• Computer-aided detection (CAD) systems are vital in medical imaging for enhancing efficiency and accuracy.
• Accurate image segmentation is a critical preprocessing step in CAD systems.
• Current ground truth delineation methods, manual or automated, have limitations in precision and consistency.

Purpose of the Study:

• To propose a systematic ground truth delineation method using Local Consistency Set Analysis.
• To establish an accurate ground truth representation for medical image segmentation.
• To provide a method for assessing the accuracy of CAD segmentation algorithms.

Main Methods:

• Development of a computational model based on Local Consistency Set Analysis.
• Validation of the model using medical imaging data.
• Analysis of consistency at the distributed boundary pixel level.

Main Results:

• The proposed Local Consistency Set Analysis method demonstrates robustness in establishing accurate ground truth.
• The approach provides consistency information at the pixel level, offering finer detail than global methods.
• Experimental results validate the effectiveness of the computational model.

Conclusions:

• The Local Consistency Set Analysis offers a superior method for ground truth delineation in medical image segmentation.
• This approach enhances the accuracy assessment of computer-aided detection (CAD) algorithms.
• The method's invariance to global compensation errors marks a significant advancement.