Synthetic CT generation using Zero TE MR for head-and-neck radiotherapy.

Area of Science:

• Medical Imaging
• Radiation Oncology
• Artificial Intelligence in Medicine

Background:

• MRI-based synthetic CTs (synCTs) are promising for replacing planning CT scans.
• The head-and-neck region presents challenges due to complex tissue interfaces.
• Zero-Echo-Time (ZTE) MRI offers fast, silent scanning with accurate bone and air discrimination.

Purpose of the Study:

• To evaluate the dosimetric accuracy of a novel, fast ZTE MRI sequence for synCT generation in the head-and-neck region.
• To assess the potential of ZTE-based synCTs to replace conventional CT scans for radiation therapy planning.

Main Methods:

• Developed and validated a synCT generation method using a multi-task 2D U-net from ZTE MRIs.
• Scanned 127 patients, with validation on 17 independent cases.
• Recalculated clinical treatment plans on synCTs and compared dose-volume-histogram metrics with CT data.

Main Results:

• Mean absolute error between synCT and CT was 94 ± 11 HU within the patient contour.
• Clinical plans showed excellent agreement: 98.8% of PTV metrics deviated <2% and all OAR metrics deviated <1 Gy.
• Synthetic plans exhibited dose differences influenced by PTV location.

Conclusions:

• ZTE-MR-based synCTs demonstrate excellent dosimetric accuracy in the head-and-neck region compared to conventional CT.
• Synthetic plans are valuable for assessing synCT scan accuracy in clinical practice.