The effect of errors in segmented attenuation maps on PET quantification.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Accurate attenuation correction is critical for quantitative Positron Emission Tomography (PET).
• MRI-based attenuation correction often relies on segmented attenuation maps, which can be prone to segmentation errors.
• Understanding the impact of these errors on PET quantification is essential for reliable imaging.

Purpose of the Study:

• To evaluate the quantitative impact of segmentation errors in MRI-derived attenuation maps on PET imaging.
• To identify specific segmentation errors that lead to significant deviations in PET quantification.
• To provide recommendations for improving the accuracy of attenuation correction in PET.

Main Methods:

• Simulated digital XCAT phantom data using the GATE Monte Carlo framework and a Philips Gemini TF model.
• Performed whole-body PET scan simulations with fifteen lesions placed in various organs and tissues.
• Reconstructed simulated PET data using a reference attenuation map and several modified maps reflecting common segmentation errors.

Main Results:

• Segmentation errors below 5% were observed when distinguishing five tissue classes: cortical bone, spongy bone, soft tissue, lung, and air.
• Ignoring lung tissue resulted in errors up to 45%, and ignoring cortical bone up to 17%.
• Interpatient variability in lung attenuation coefficients caused errors exceeding 10%; limited misclassification between bone/soft tissue and lung/air had minimal impact.

Conclusions:

• A minimum of five tissue types (cortical bone, spongy bone, soft tissue, lung, air) must be considered in segmented attenuation maps for accurate PET quantification.
• Interpatient variability of lung attenuation coefficients necessitates careful consideration.
• Minor misclassifications between bone and soft tissue, or lung and air, are generally acceptable and do not significantly affect quantitative results.