Defining the lung outline from a gamma camera transmission attenuation map

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Processing

Background:

• Accurate lung segmentation from gamma camera transmission images is crucial for attenuation correction and quantitative analysis in thoracic imaging.
• Existing methods require evaluation for precision and accuracy in diverse imaging scenarios.

Purpose of the Study:

• To develop and compare two threshold-based lung segmentation techniques: global thresholding (GT) and context-sensitive thresholding (CST).
• To evaluate the performance of GT and CST in segmenting lung outlines from simulated and real human subject transmission images.

Main Methods:

• Development of GT and CST algorithms for lung segmentation.
• Generation of simulated gamma camera transmission images with varying resolution and chest wall thickness.
• Evaluation using simulated realistic thoraces and three real human subject transmission images.
• Comparison of segmented lung volumes against true values and calculation of surface distances.

Main Results:

• CST demonstrated significantly superior segmentation compared to GT in simulated data.
• In human subjects, GT underestimated lung volumes by 13% and missed lung areas.
• Both methods showed reasonable accuracy and precision, but CST was notably better for real subject images.

Conclusions:

• Context-sensitive thresholding (CST) is a more effective method for lung segmentation in gamma camera transmission imaging than global thresholding (GT).
• CST offers improved accuracy and reliability, especially for clinical applications involving human subjects.