Left-ventricle boundary detection from nuclear medicine images

Area of Science:

• Medical Imaging
• Image Segmentation
• Nuclear Medicine

Background:

• Accurate segmentation of the left-ventricle (LV) boundary in nuclear medicine images is crucial for cardiac assessment.
• Existing image segmentation techniques like edge detection and region growing have limitations in accurately defining the LV boundary.

Purpose of the Study:

• To develop and evaluate a novel, robust algorithm for left-ventricle (LV) boundary segmentation in nuclear medicine images.
• To compare the performance of the proposed algorithm against traditional edge detection and region growing methods.

Main Methods:

• Exploration of edge detection limitations based on radial orientations and brightness functions.
• Evaluation of various region growing criteria, including intensity and gradient changes.
• Introduction and implementation of the Contour-Modified Region Growing (CMRG) algorithm.

Main Results:

• Edge detection methods were insufficient due to variations in LV boundary characteristics.
• Individual region growing criteria failed to reliably detect the LV boundary.
• The proposed CMRG algorithm demonstrated superior performance, handling low signal-to-noise ratios and low contrast effectively.

Conclusions:

• The Contour-Modified Region Growing (CMRG) algorithm offers a robust, rapid, and user-friendly solution for left-ventricle (LV) segmentation.
• CMRG overcomes limitations of previous methods, performing well without assumptions on LV shape.
• This algorithm significantly advances automated segmentation in challenging nuclear medicine imaging scenarios.