Automatic lung segmentation in functional SPECT images using active shape models trained on reference lung shapes...

Area of Science:

• Medical Imaging
• Radiology
• Pulmonology

Background:

• Accurate lung segmentation is crucial for quantifying lung function, especially in patients with obstructive lung disease.
• Ventilation/Perfusion (V/P) SPECT imaging provides functional information about the lungs.
• Computed Tomography (CT) offers detailed anatomical information.

Purpose of the Study:

• To develop and validate a novel automated tool for lung segmentation in V/P SPECT images.
• To compare the performance of the automated SPECT lung segmentation against manual SPECT segmentation.
• To evaluate both automatic and manual SPECT lung segmentations against reference CT volumes.

Main Methods:

• Developed active shape models for right and left lung segmentation using 57 low-dose CT images.
• Trained and validated the automated segmentation algorithm on 77 subjects (69 with obstructive lung disease).
• Compared automated SPECT segmentations with manual SPECT segmentations and reference CT volumes using Dice coefficients and volumetric analysis.

Main Results:

• The automated SPECT lung segmentation achieved Dice coefficients of 0.83 ± 0.04 for the right lung and 0.82 ± 0.05 for the left lung, comparable to manual segmentation.
• Statistically significant volumetric differences were observed between SPECT delineations (both automatic and manual) and reference CT volumes.
• These findings highlight the inherent challenges in segmenting functional SPECT images compared to anatomical CT.

Conclusions:

• Automated lung segmentation on SPECT images demonstrates performance on par with manual segmentation.
• The study confirms that segmenting functional SPECT images is challenging, with notable volumetric discrepancies compared to anatomical CT.
• The developed algorithm represents a significant initial step towards automated quantification of various lung measurements from SPECT imaging.