Segmentation methods for volume determination with 111In/99Tcm SPET

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Analysis

Background:

• Accurate quantification of radionuclide volume distributions in single photon emission tomographic (SPET) images relies on objective determination of regions of interest (ROIs).
• Various segmentation methods exist, each with potential strengths and weaknesses for SPET image analysis.

Purpose of the Study:

• To compare the performance of four different image segmentation methods for determining ROIs in SPET images.
• To evaluate the accuracy and reliability of fixed thresholding (FT), grey level histogram (GL), region growing (RG), and combined region growing and edge detection (RGE) methods.

Main Methods:

• An elliptical phantom with varying volumes and activities of 111In and 99Tcm was used to assess segmentation accuracy.
• Four methods (FT, GL, RG, RGE) were applied to phantom data and SPET images of 25 patients.
• Correlation coefficients (r) were calculated to compare true and measured volumes.

Main Results:

• FT and RG demonstrated high correlation coefficients (r=0.95 for 111In, r=0.85 for 99Tcm) in phantom studies.
• Significant correlation was observed between spleen volumes measured with FT, RG, and RGE in patient SPET images, unlike the GL method.
• FT and RG showed robustness to image filtering cut-off frequencies.

Conclusions:

• Fixed thresholding (FT) and region growing (RG) are recommended for clinical ROI determination in SPET imaging due to their accuracy.
• These methods may face challenges with low or variable signal-to-noise ratios, potentially requiring multimodal imaging.
• Phantom measurements are crucial for optimizing parameters before clinical application of any segmentation method.