Personalized Modeling to Improve Pseudo-Computed Tomography Images for Magnetic Resonance Imaging-Guided Adaptive Radiation Therapy

Abstract

Purpose: Magnetic resonance imaging-guided adaptive radiation therapy (MRIgART) greatly improves daily tumor localization and enables online replanning to obtain maximum dosimetric benefits. However, accurately predicting patient-specific electron density maps for adaptive radiation therapy planning remains a challenge. Therefore, this study proposes a personalized modeling framework for generating pseudo-computed tomography (pCT) in MRIgART.

Methods and materials: Eighty-three patients who received MRIgART were included and computed tomography (CT) simulations were performed on all the patients. Daily T2-weighted 1.5 T magnetic resonance imaging (MRI) was acquired using the Unity MR-linac for adaptive planning. Pairs of coregistered CT and daily MRI images of the randomly selected training set (68 patients) were inputted into a generative adversarial network to establish a population model. The personalized model for each patient in the test set (15 patients) was acquired using model fine-tuning, which adopted the pair of the deformable-registered CT and the first daily MRI to fine-tune the population model. The pCT quality was quantitatively evaluated in the second and the last fractions with 3 metrics: intensity accuracy using mean absolute error; anatomic structure similarity using dice similarity coefficient; and dosimetric consistency using gamma-passing rate.

Results: The image generation speed was 65 slices/s. For the last fractions, and for head-neck, thoracoabdominal, and pelvic cases, the average mean absolute errors were 76.8 HU versus 123.6 HU, 38.1 HU versus 52.0 HU, and 29.5 HU versus 39.7 HU, respectively. Furthermore, the average dice similarity coefficients of bone were 0.92 versus 0.80, 0.85 versus 0.73, and 0.94 versus 0.88; and the average gamma-passing rates (1%/1 mm) were 95.5% versus 84.7%, 97.7% versus 92.8%, and 95.5% versus 88.7%, for personalized versus population models, respectively. The results of the second fractions were similar.

Conclusions: The proposed personalized modeling framework remarkably improved pCT quality for multiple treatment sites and was well suited for the MRIgART clinical setting.