TH‐C‐141‐04: Feasibility of MR Relaxometry‐Guided Synthetic CT Image Formation for MRI‐Based Radiation Treatment Planning

Abstract

Purpose: Recently, several groups have begun investigating methods to generate synthetic CT images from MRI images to facilitate MRI‐based radiation treatment planning. The goal of this work was to determine the feasibility of a novel approach for generating synthetic CT images using MR relaxometry guidance. Methods: MR relaxometry images of the head and pelvis were acquired on three volunteers using a Siemens 3T scanner. Rapid T1 mapping was performed by linear least squares fitting of the Ernst model to multiple flip angle (2, 5, 10, 15 deg) images acquired with a 3D FLASH sequence and corrected for B1+ inhomogeneities using a custom, 3D actual flip angle imaging (AFI) sequence. Rapid T2* mapping was performed by linear least squares fitting of eight echo time (2.9–29.1 ms) images acquired with a 3D FLASH sequence. Vendor‐provided 3D distortion correction was applied to all images to correct for gradient nonlinearity distortions. Tissue classification was performed based on published values of T1 and T2* at 3T. Electron densities (EDs), obtained from ICRU Report 46, were assigned to the segmented structures. Synthetic CT images were generated by combining the segmented ED structures and applying an inverted CT‐ED conversion table from a clinical radiation treatment planning system. Results: Total scan time for the MR relaxometry and AFI images was approximately ten minutes. T1 maps generated with B1+ correction demonstrated high uniformity compared to T1 maps obtained without correction. The combination of T1 and T2* maps was necessary to classify tissues in the head and pelvis regions. Synthetic CT images were successfully generated and transferred onto a radiation treatment planning system. Conclusion: Generation of synthetic CT images for MRI‐based radiation treatment planning using MR relaxometry‐guidance is feasible. Further investigations will compare synthetic CT images generated using the proposed approach to actual CT images in patients.