NIMG-29. IN SILLICO BRAIN TUMOR MODELS FOR VALIDATING NEW DYNAMIC MR IMAGING METHODS PRIOR TO CLINICAL TRIALS. (5th November 2018)

• Record Type: Journal Article
• Authors: Chen, Junzhou; Henze Bancroft, Leah; Jimenez, Jorge; Strigel, Roberta; Ahmed, Azam; Block, Walter
• Is Part Of: Neuro-oncology. Volume 20(2018)Supplement 6
• Page Start: vi182
• Page End: vi182
• Publication Date: 2018-11-05
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noy148.755
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: New dynamic MRI methods hold promise in better characterization of the complex structure and heterogeneous treatment response of brain tumors. These methods utilize assumptions to reduce the data samples needed to produce high temporal and spatial resolution images with complex, iterative reconstructions. Determining how these methods depict actual brain tumor pathology is difficult due to lack of a gold standard. We therefore tailor an existing digital tool, built originally to simulate dynamic MRI breast studies, to assess the performance of new dynamic methods if applied to brain cancer. The tool permits a tumor of customizable shape featuring an inner core and an adjustable surrounding rim to be overlaid on a digital 3-D brain model, which we acquired from an online database called BrainWeb, with segmented brain tissue layers. We used existing capabilities in the tool to assign pharmacokinetic parameters to each tumor region to simulate necrotic cores and rims of varying width that represent spatially varying levels of enhancement due to tumor pathology and/or radiation necrosis. The tool then computes the corresponding MR k-space data for any proposed MR sampling trajectory during a simulated passage of a contrast agent. The simulated MR k-space data can then be input to any proposed reconstruction to produce a simulated time course of image volumes. Performance of proposed dynamic MR imaging methods can be measured against the digital truth prior to clinical trials through methods including: the structural similarity (SSIM) index over the lesion region-of-interest, comparing calculated vs. assigned pharmacokinetic parameters, and root-mean-square error. We are using this tool to estimate the performance, in brain applications, of a 3-D radial acquisition and reconstruction method with compressed sensing and local low rank that was previously demonstrated to produce 0.8 mm resolution in a 10 second frame rate in bilateral breast screening.