skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: SU-F-J-93: Automated Segmentation of High-Resolution 3D WholeBrain Spectroscopic MRI for Glioblastoma Treatment Planning

Abstract

Purpose: We report on an automated segmentation algorithm for defining radiation therapy target volumes using spectroscopic MR images (sMRI) acquired at nominal voxel resolution of 100 microliters. Methods: Wholebrain sMRI combining 3D echo-planar spectroscopic imaging, generalized auto-calibrating partially-parallel acquisitions, and elliptical k-space encoding were conducted on 3T MRI scanner with 32-channel head coil array creating images. Metabolite maps generated include choline (Cho), creatine (Cr), and N-acetylaspartate (NAA), as well as Cho/NAA, Cho/Cr, and NAA/Cr ratio maps. Automated segmentation was achieved by concomitantly considering sMRI metabolite maps with standard contrast enhancing (CE) imaging in a pipeline that first uses the water signal for skull stripping. Subsequently, an initial blob of tumor region is identified by searching for regions of FLAIR abnormalities that also display reduced NAA activity using a mean ratio correlation and morphological filters. These regions are used as starting point for a geodesic level-set refinement that adapts the initial blob to the fine details specific to each metabolite. Results: Accuracy of the segmentation model was tested on a cohort of 12 patients that had sMRI datasets acquired pre, mid and post-treatment, providing a broad range of enhancement patterns. Compared to classical imaging, where heterogeneity in the tumor appearance andmore » shape across posed a greater challenge to the algorithm, sMRI’s regions of abnormal activity were easily detected in the sMRI metabolite maps when combining the detail available in the standard imaging with the local enhancement produced by the metabolites. Results can be imported in the treatment planning, leading in general increase in the target volumes (GTV60) when using sMRI+CE MRI compared to the standard CE MRI alone. Conclusion: Integration of automated segmentation of sMRI metabolite maps into planning is feasible and will likely streamline acceptance of this new acquisition modality in clinical practice.« less

Authors:
;  [1]; ; ; ; ;  [2]
  1. Department of Radiation Oncology, Winship Cancer Institute of Emory University, Atlanta, GA (United States)
  2. Department of Radiology and Imaging Sciences, Emory University, Atlanta, GA (United States)
Publication Date:
OSTI Identifier:
22634702
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 43; Journal Issue: 6; Other Information: (c) 2016 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
60 APPLIED LIFE SCIENCES; 61 RADIATION PROTECTION AND DOSIMETRY; ACCURACY; ALGORITHMS; BIOMEDICAL RADIOGRAPHY; CHOLINE; COMPARATIVE EVALUATIONS; CORRELATIONS; CREATINE; DATASETS; GLIOMAS; IMAGES; METABOLITES; NEUTRON ACTIVATION ANALYSIS; NMR IMAGING; PATIENTS; PLANNING; RADIOTHERAPY; SKULL

Citation Formats

Schreibmann, E, Shu, H, Cordova, J, Gurbani, S, Holder, C, Cooper, L, and Shim, H. SU-F-J-93: Automated Segmentation of High-Resolution 3D WholeBrain Spectroscopic MRI for Glioblastoma Treatment Planning. United States: N. p., 2016. Web. doi:10.1118/1.4956001.
Schreibmann, E, Shu, H, Cordova, J, Gurbani, S, Holder, C, Cooper, L, & Shim, H. SU-F-J-93: Automated Segmentation of High-Resolution 3D WholeBrain Spectroscopic MRI for Glioblastoma Treatment Planning. United States. doi:10.1118/1.4956001.
Schreibmann, E, Shu, H, Cordova, J, Gurbani, S, Holder, C, Cooper, L, and Shim, H. 2016. "SU-F-J-93: Automated Segmentation of High-Resolution 3D WholeBrain Spectroscopic MRI for Glioblastoma Treatment Planning". United States. doi:10.1118/1.4956001.
@article{osti_22634702,
title = {SU-F-J-93: Automated Segmentation of High-Resolution 3D WholeBrain Spectroscopic MRI for Glioblastoma Treatment Planning},
author = {Schreibmann, E and Shu, H and Cordova, J and Gurbani, S and Holder, C and Cooper, L and Shim, H},
abstractNote = {Purpose: We report on an automated segmentation algorithm for defining radiation therapy target volumes using spectroscopic MR images (sMRI) acquired at nominal voxel resolution of 100 microliters. Methods: Wholebrain sMRI combining 3D echo-planar spectroscopic imaging, generalized auto-calibrating partially-parallel acquisitions, and elliptical k-space encoding were conducted on 3T MRI scanner with 32-channel head coil array creating images. Metabolite maps generated include choline (Cho), creatine (Cr), and N-acetylaspartate (NAA), as well as Cho/NAA, Cho/Cr, and NAA/Cr ratio maps. Automated segmentation was achieved by concomitantly considering sMRI metabolite maps with standard contrast enhancing (CE) imaging in a pipeline that first uses the water signal for skull stripping. Subsequently, an initial blob of tumor region is identified by searching for regions of FLAIR abnormalities that also display reduced NAA activity using a mean ratio correlation and morphological filters. These regions are used as starting point for a geodesic level-set refinement that adapts the initial blob to the fine details specific to each metabolite. Results: Accuracy of the segmentation model was tested on a cohort of 12 patients that had sMRI datasets acquired pre, mid and post-treatment, providing a broad range of enhancement patterns. Compared to classical imaging, where heterogeneity in the tumor appearance and shape across posed a greater challenge to the algorithm, sMRI’s regions of abnormal activity were easily detected in the sMRI metabolite maps when combining the detail available in the standard imaging with the local enhancement produced by the metabolites. Results can be imported in the treatment planning, leading in general increase in the target volumes (GTV60) when using sMRI+CE MRI compared to the standard CE MRI alone. Conclusion: Integration of automated segmentation of sMRI metabolite maps into planning is feasible and will likely streamline acceptance of this new acquisition modality in clinical practice.},
doi = {10.1118/1.4956001},
journal = {Medical Physics},
number = 6,
volume = 43,
place = {United States},
year = 2016,
month = 6
}
  • Purpose: Identification of a metabolite signature that shows significant tumor cell infiltration into normal brain in regions that do not appear abnormal on standard MRI scans would be extremely useful for radiation oncologists to choose optimal regions of brain to treat, and to quantify response beyond the MacDonald criteria. We report on integration of high-resolution magnetic resonance spectroscopic imaging (HR-MRSI) with radiation dose escalation treatment planning to define and target regions at high risk for recurrence. Methods: We propose to supplement standard MRI with a special technique performed on an MRI scanner to measure the metabolite levels within defined volumes.more » Metabolite imaging was acquired using an advanced MRSI technique combining 3D echo-planar spectroscopic imaging (EPSI) with parallel acquisition (GRAPPA) using a multichannel head coil that allows acquisition of whole brain metabolite maps with 108 μl resolution in 12 minutes implemented on a 3T MR scanner. Elevation in the ratio of two metabolites, choline (Cho, elevated in proliferating high-grade gliomas) and N-acetyl aspartate (NAA, a normal neuronal metabolite), was used to image infiltrating high-grade glioma cells in vivo. Results: The metabolite images were co-registered with standard contrast-enhanced T1-weighted MR images using in-house registration software and imported into the treatment-planning system. Regions with tumor infiltration are identified on the metabolic images and used to create adaptive IMRT plans that deliver a standard dose of 60 Gy to the standard target volume and an escalated dose of 75 Gy (or higher) to the most suspicious regions, identified as areas with elevated Cho/NAA ratio. Conclusion: We have implemented a state-of-the-art HR-MRSI technology that can generate metabolite maps of the entire brain in a clinically acceptable scan time, coupled with introduction of an imaging co-registration/ analysis program that combines MRSI data with standard imaging studies in a clinically useful fashion.« less
  • Purpose: Accurate image segmentation is a crucial step during image guided radiation therapy. This work proposes multi-atlas machine learning (MAML) algorithm for automated segmentation of head-and-neck CT images. Methods: As the first step, the algorithm utilizes normalized mutual information as similarity metric, affine registration combined with multiresolution B-Spline registration, and then fuses together using the label fusion strategy via Plastimatch. As the second step, the following feature selection strategy is proposed to extract five feature components from reference or atlas images: intensity (I), distance map (D), box (B), center of gravity (C) and stable point (S). The box feature Bmore » is novel. It describes a relative position from each point to minimum inscribed rectangle of ROI. The center-of-gravity feature C is the 3D Euclidean distance from a sample point to the ROI center of gravity, and then S is the distance of the sample point to the landmarks. Then, we adopt random forest (RF) in Scikit-learn, a Python module integrating a wide range of state-of-the-art machine learning algorithms as classifier. Different feature and atlas strategies are used for different ROIs for improved performance, such as multi-atlas strategy with reference box for brainstem, and single-atlas strategy with reference landmark for optic chiasm. Results: The algorithm was validated on a set of 33 CT images with manual contours using a leave-one-out cross-validation strategy. Dice similarity coefficients between manual contours and automated contours were calculated: the proposed MAML method had an improvement from 0.79 to 0.83 for brainstem and 0.11 to 0.52 for optic chiasm with respect to multi-atlas segmentation method (MA). Conclusion: A MAML method has been proposed for automated segmentation of head-and-neck CT images with improved performance. It provides the comparable result in brainstem and the improved result in optic chiasm compared with MA. Xuhua Ren and Hao Gao were partially supported by the NSFC (#11405105), the 973 Program (#2015CB856000), and the Shanghai Pujiang Talent Program (#14PJ1404500).« less
  • Purpose: Magnetic resonance (MR) imaging and computed tomography (CT) are used almost exclusively in radiation therapy planning of glioblastoma multiforme (GBM), despite their well-recognized limitations. MR spectroscopic imaging (MRSI) can identify biochemical patterns associated with normal brain and tumor, predominantly by observation of choline (Cho) and N-acetylaspartate (NAA) distributions. In this study, volumetric 3-dimensional MRSI was used to map these compounds over a wide region of the brain and to evaluate metabolite-defined treatment targets (metabolic tumor volumes [MTV]). Methods and Materials: Volumetric MRSI with effective voxel size of ∼1.0 mL and standard clinical MR images were obtained from 19 GBM patients.more » Gross tumor volumes and edema were manually outlined, and clinical target volumes (CTVs) receiving 46 and 60 Gy were defined (CTV{sub 46} and CTV{sub 60}, respectively). MTV{sub Cho} and MTV{sub NAA} were constructed based on volumes with high Cho and low NAA relative to values estimated from normal-appearing tissue. Results: The MRSI coverage of the brain was between 70% and 76%. The MTV{sub NAA} were almost entirely contained within the edema, and the correlation between the 2 volumes was significant (r=0.68, P=.001). In contrast, a considerable fraction of MTV{sub Cho} was outside of the edema (median, 33%) and for some patients it was also outside of the CTV{sub 46} and CTV{sub 60}. These untreated volumes were greater than 10% for 7 patients (37%) in the study, and on average more than one-third (34.3%) of the MTV{sub Cho} for these patients were outside of CTV{sub 60}. Conclusions: This study demonstrates the potential usefulness of whole-brain MRSI for radiation therapy planning of GBM and revealed that areas of metabolically active tumor are not covered by standard RT volumes. The described integration of MTV into the RT system will pave the way to future clinical trials investigating outcomes in patients treated based on metabolic information.« less
  • Purpose We developed an automated treatment planning system based on a hierarchical goal programming approach. To demonstrate the feasibility of our method, we report the comparison of prostate treatment plans produced from the automated treatment planning system with those produced by a commercial treatment planning system. Methods In our approach, we prioritized the goals of the optimization, and solved one goal at a time. The purpose of prioritization is to ensure that higher priority dose-volume planning goals are not sacrificed to improve lower priority goals. The algorithm has four steps. The first step optimizes dose to the target structures, whilemore » sparing key sensitive organs from radiation. In the second step, the algorithm finds the best beamlet weight to reduce toxicity risks to normal tissue while holding the objective function achieved in the first step as a constraint, with a small amount of allowed slip. Likewise, the third and fourth steps introduce lower priority normal tissue goals and beam smoothing. We compared with prostate treatment plans from Memorial Sloan Kettering Cancer Center developed using Eclipse, with a prescription dose of 72 Gy. A combination of liear, quadratic, and gEUD objective functions were used with a modified open source solver code (IPOPT). Results Initial plan results on 3 different cases show that the automated planning system is capable of competing or improving on expert-driven eclipse plans. Compared to the Eclipse planning system, the automated system produced up to 26% less mean dose to rectum and 24% less mean dose to bladder while having the same D95 (after matching) to the target. Conclusion We have demonstrated that Pareto optimal treatment plans can be generated automatically without a trial-and-error process. The solver finds an optimal plan for the given patient, as opposed to database-driven approaches that set parameters based on geometry and population modeling.« less
  • Purpose: To describe approaches to four-dimensional (4D) treatment planning, including acquisition of 4D-CT scans, target delineation of spatio-temporal image data sets, 4D dose calculations, and their analysis. Methods and Materials: The study included patients with thoracic and hepatocellular tumors. Specialized tools were developed to facilitate visualization, segmentation, and analysis of 4D-CT data: maximum intensity volume to define the extent of lung tumor motion, a 4D browser to examine and dynamically assess the 4D data sets, dose calculations, including respiratory motion, and deformable registration to combine the dose distributions at different points. Results: Four-dimensional CT was used to visualize and quantitativelymore » assess respiratory target motion. The gross target volume contours derived from light breathing scans showed significant differences compared with those extracted from 4D-CT. Evaluation of deformable registration using difference images of original and deformed anatomic maps suggested the algorithm is functionally useful. Thus, calculation of effective dose distributions, including respiratory motion, was implemented. Conclusion: Tools and methods to use 4D-CT data for treatment planning in the presence of respiratory motion have been developed and applied to several case studies. The process of 4D-CT-based treatment planning has been implemented, and technical barriers for its routine use have been identified.« less