Evaluating a radiotherapy deep learning synthetic CT algorithm for PET-MR attenuation correction in the pelvis

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Artificial Intelligence in Medicine

Background:

• Positron emission tomography-magnetic resonance (PET-MR) attenuation correction is difficult due to MR signal limitations in representing tissue density and imaging bone.
• A novel zero echo time (ZTE) MR sequence can image cortical bone, and a deep learning model generates synthetic CT (sCT) for MR-only radiotherapy.

Purpose of the Study:

• To evaluate a ZTE-based synthetic CT (sCT) algorithm for PET-MR attenuation correction in pelvic radiotherapy.
• To compare the accuracy of sCT-based attenuation correction (sCTAC) against standard MR-based attenuation correction (MRAC) and CT-based attenuation correction (CTAC).

Main Methods:

• Ten patients undergoing ano-rectal radiotherapy received [Formula: see text]F-FDG-PET-MR scans.
• Attenuation maps were generated using ZTE-based sCT (sCTAC) and vendor-supplied MRAC.
• PET images were reconstructed using sCTAC, MRAC, and CTAC (gold standard) for comparison of SUV, GTV delineation, and metabolic parameters.

Main Results:

• Mean whole-image SUV differences were -0.02% for sCTAC versus -3.0% for MRAC compared to CTAC.
• No significant differences were observed in thresholded GTVs (Dice similarity coefficients [Formula: see text]).
• sCTAC demonstrated statistical equivalence to CTAC for GTV metabolic parameters ([Formula: see text] and [Formula: see text]), while MRAC did not.

Conclusions:

• Attenuation correction using the radiotherapy ZTE-based sCT algorithm is substantially more accurate than current MRAC methods.
• This approach minimally increases MR acquisition time (40s) and improves SUV accuracy without impacting tumor delineation.
• ZTE-based sCTAC enables accurate quantitative PET imaging on PET-MR scanners, offering a viable alternative for radiotherapy applications.