Search

Using magnetic resonance imaging (MRI) as the sole imaging modality for patient modeling in radiation therapy (RT) is a challenging task due to the need to derive electron density information from MRI and construct a so-called pseudo-computed tomography (pCT) image. We have previously published a new method to derive pCT images from head T1-weighted (T1-w) MR images using a single-atlas propagation scheme followed by a post hoc correction of the mapped CT numbers using local intensity information. The purpose of this study was to investigate the performance of our method with head zero echo time (ZTE) MR images. To evaluate results, the mean absolute error in bins of 20 HU was calculated with respect to the true planning CT scan of the patient. We demonstrated that applying our method using ZTE MR images instead of T1-w improved the correctness of the pCT in case of bone resection surgery prior to RT (that is, an example of large anatomical difference between the atlas and the patient)., (Copyright © 2017. Published by Elsevier Ltd.)

Purpose: Magnetic resonance imaging (MRI)-only radiotherapy treatment planning requires accurate pseudo-CT (pCT) images for precise dose calculation. The current work introduced an atlas-based method combined with MR intensity information. pCT analyses and Monte Carlo dose calculations for intracranial stereotactic treatments were performed., Methods: Twenty-two patients, representing 35 tumor targets, were scanned using a 3D T1-weighted MRI sequence according to the clinical protocol. The MR atlas image was registered to the MR patient image using a deformable algorithm, and the deformation was then applied to the atlas CT. Two methods were applied. The first method (MRdef) was based on deformations only, while the second (MRint) also used the actual MR intensities. pCT analysis was performed using the mean (absolute) error, as well as an in-house tool based on a gamma index. Dose differences between pCT and true CT were analyzed using dose-volume histogram (DVH) parameters, statistical tests, the gamma index, and probability density functions. An unusual case, where the patient underwent an operation (part of the skull bone was removed), was studied in detail., Results: Soft tissues presented a mean error inferior to 50 HUs, while low-density tissues and bones presented discrepancies up to 600 HUs for hard bone. The MRdef method led to significant dose differences compared with the true CT (p-value < 0.05; Wilcoxon-signed-rank test). The MRint method performed better. The DVH parameter differences compared with CT were between -2.9% and 3.1%, except for two cases where the tumors were located within the sphenoid bone. For these cases, the dose errors were up to 6.6% and 5.4% (D 98 and D 95 ). Furthermore, for 85% of the tested patients, the mean dose to the planning target volume agreed within 2% with the calculation using the actual CT. Fictitious bone was generated in the unusual case using atlas-based methods., Conclusions: Generally, the atlas-based method led to acceptable dose distributions. The use of common T1 sequences allows the implementation of this method in clinical routine. However, unusual patient anatomy may produce large dose calculation errors. The detection of large anatomic discrepancies using MR image subtraction can be realized, but an alternative way to produce synthetic CT numbers in these regions is still required.

Delineation of organs at risk (OARs) is a crucial step in surgical and treatment planning in brain cancer, where precise OARs volume delineation is required. However, this task is still often manually performed, which is time-consuming and prone to observer variability. To tackle these issues a deep learning approach based on stacking denoising auto-encoders has been proposed to segment the brainstem on magnetic resonance images in brain cancer context. Additionally to classical features used in machine learning to segment brain structures, two new features are suggested. Four experts participated in this study by segmenting the brainstem on 9 patients who underwent radiosurgery. Analysis of variance on shape and volume similarity metrics indicated that there were significant differences (p<0.05) between the groups of manual annotations and automatic segmentations. Experimental evaluation also showed an overlapping higher than 90% with respect to the ground truth. These results are comparable, and often higher, to those of the state of the art segmentation methods but with a considerably reduction of the segmentation time., (Copyright © 2016 Elsevier Ltd. All rights reserved.)

Purpose: To constrain the risk of severe toxicity in radiotherapy and radiosurgery, precise volume delineation of organs at risk is required. This task is still manually performed, which is time-consuming and prone to observer variability. To address these issues, and as alternative to atlas-based segmentation methods, machine learning techniques, such as support vector machines (SVM), have been recently presented to segment subcortical structures on magnetic resonance images (MRI)., Methods: SVM is proposed to segment the brainstem on MRI in multicenter brain cancer context. A dataset composed by 14 adult brain MRI scans is used to evaluate its performance. In addition to spatial and probabilistic information, five different image intensity values (IIVs) configurations are evaluated as features to train the SVM classifier. Segmentation accuracy is evaluated by computing the Dice similarity coefficient (DSC), absolute volumes difference (AVD) and percentage volume difference between automatic and manual contours., Results: Mean DSC for all proposed IIVs configurations ranged from 0.89 to 0.90. Mean AVD values were below 1.5 cm(3), where the value for best performing IIVs configuration was 0.85 cm(3), representing an absolute mean difference of 3.99% with respect to the manual segmented volumes., Conclusion: Results suggest consistent volume estimation and high spatial similarity with respect to expert delineations. The proposed approach outperformed presented methods to segment the brainstem, not only in volume similarity metrics, but also in segmentation time. Preliminary results showed that the approach might be promising for adoption in clinical use.