Faculty Collaboration Database

A Technique to Rapidly Generate Synthetic Computed Tomography for Magnetic Resonance Imaging-Guided Online Adaptive Replanning: An Exploratory Study. Int J Radiat Oncol Biol Phys 2019 Apr 01;103(5):1261-1270

Date

12/15/2018

Pubmed ID

30550817

DOI

10.1016/j.ijrobp.2018.12.008

Scopus ID

2-s2.0-85060989536 (requires institutional sign-in at Scopus site)   16 Citations

Abstract

PURPOSE: To develop an automatic, accurate, atlas-based technique for synthetic computed tomography (sCT) generation to be used for online adaptive replanning during magnetic resonance imaging (MRI)-guided radiation therapy (RT).

METHODS AND MATERIALS: The proposed method uses deformable image registration (DIR) of daily MRI and reference computed tomography (CT) with additional corrections to maintain bone rigidity and to transfer random air regions by thresholding. The DIR is performed with constraints on the bony structures using a special algorithm of ADMIRE (Elekta). The air regions are delineated from low-signal regions on the daily MRI and forced to air density. The bone regions in the MRI (already determined from the CT) are separated from the air regions because both bone and air have low signal density in MRI. All these steps are automated. The generated sCT is compared with reference CT and the alternative voxel-based CT (bCT) for 4 extracranial sites (head and neck, thorax, abdomen, pelvis) in terms of mean absolute error (MAE), gamma analysis of 3-dimensional doses, and dose volume histogram parameters.

RESULTS: Both MAE and dosimetric analysis results were favorable for the proposed sCT generation method. The average MAE for the sCT/bCT were 25.5/66.7, 25.9/65.3, 24.8/44.2 and 16.6/47.7 for head and neck, thorax, abdomen, and pelvis, respectively, and the gamma analysis (1.5%, 2 mm) yielded 98.7/97.1, 99.1/93.9, 99.5/99.4, 99.7/99.4, respectively, for those sites.

CONCLUSIONS: The proposed method generates equal or more accurate sCT than those from the bulk density assignment, without the need for multiple MRI sequences. This method can be fully automated and applicable for online adaptive replanning.

Author List

Ahunbay EE, Thapa R, Chen X, Paulson E, Li XA

Authors

Ergun Ahunbay PhD Professor in the Radiation Oncology department at Medical College of Wisconsin
Eric Paulson PhD Chief, Professor in the Radiation Oncology department at Medical College of Wisconsin

MESH terms used to index this publication - Major topics in bold

Abdominal Neoplasms
Air
Algorithms
Automation
Bone and Bones
Connective Tissue
Head and Neck Neoplasms
Humans
Image Processing, Computer-Assisted
Intestines
Magnetic Resonance Imaging
Magnetic Resonance Imaging, Interventional
Particle Accelerators
Pelvic Neoplasms
Radiotherapy Dosage
Radiotherapy Planning, Computer-Assisted
Radiotherapy, Image-Guided
Software
Thoracic Neoplasms
Tomography, X-Ray Computed