A voxel-based investigation for MRI-only radiotherapy of the brain using ultra short echo times

Abstract

Radiotherapy (RT) based on magnetic resonance imaging (MRI) as the only modality, so-called MRI-only RT, would remove the systematic registration error between MR and computed tomography (CT), and provide co-registered MRI for assessment of treatment response and adaptive RT. Electron densities, however, need to be assigned to the MRI images for dose calculation and patient setup based on digitally reconstructed radiographs (DRRs). Here, we investigate the geometric and dosimetric performance for a number of popular voxel-based methods to generate a so-called pseudo CT (pCT). Five patients receiving cranial irradiation, each containing a co-registered MRI and CT scan, were included. An ultra short echo time MRI sequence for bone visualization was used. Six methods were investigated for three popular types of voxel-based approaches; (1) threshold-based segmentation, (2) Bayesian segmentation and (3) statistical regression. Each approach contained two methods. Approach 1 used bulk density assignment of MRI voxels into air, soft tissue and bone based on logical masks and the transverse relaxation time T2 of the bone. Approach 2 used similar bulk density assignments with Bayesian statistics including or excluding additional spatial information. Approach 3 used a statistical regression correlating MRI voxels with their corresponding CT voxels. A similar photon and proton treatment plan was generated for a target positioned between the nasal cavity and the brainstem for all patients. The CT agreement with the pCT of each method was quantified and compared with the other methods geometrically and dosimetrically using both a number of reported metrics and introducing some novel metrics. The best geometrical agreement with CT was obtained with the statistical regression methods which performed significantly better than the threshold and Bayesian segmentation methods (excluding spatial information). All methods agreed significantly better with CT than a reference water MRI comparison. The mean dosimetric deviation for photons and protons compared to the CT was about 2% and highest in the gradient dose region of the brainstem. Both the threshold based method and the statistical regression methods showed the highest dosimetrical agreement.Generation of pCTs using statistical regression seems to be the most promising candidate for MRI-only RT of the brain. Further, the total amount of different tissues needs to be taken into account for dosimetric considerations regardless of their correct geometrical position.

Future of medical physics: Real‐time MRI‐guided proton therapy

“…In 2014, Edmund et al investigated five MRI-based brain tumor cases with proton beam plans. 25 Pseudo CT datasets were generated by three methods including threshold-based segmentation, Bayesian segmentation, and statistical regression. Dose agreement to within 2% was observed with reference CT-based plans for the most promising statistical regression method.…”

Generation of brain pseudo‐CTs using an undersampled, single‐acquisition UTE‐mDixon pulse sequence and unsupervised clustering

“…Most of these approaches focused on developing different data analysis algorithms, e.g., clustering/classification, 24,45,46 supervised learning, 8,47 supervised regression, 25,[41][42][43]45,48 and atlas-based approaches. [49][50][51] However, despite the importance F.…”

Monte Carlo calculation based on hydrogen composition of the tissue for MV photon radiotherapy

“…Although mathematical or statistical operations could be used to smooth registration errors due to interpatient anatomical differences, 19 , 20 these methods fail for patients presenting high anatomy dissimilarities. The voxel‐based methods convert MRI intensity into pseudo‐CT number using either a bulk density assignment after segmenting tissues or using a regression model to compute continuous values of CT number from a set of MRI data 22 , 23 , 24 , 25 . Ultrashort echo time sequences (26) are usually used in voxel based methods to distinguish the signal of bone from air.…”