Primary central nervous system tumors cause an estimated 13,000 deaths per year in the United States. Glioblastoma (GBM), the most aggressive malignant brain tumor in humans, has an especially poor prognosis with a median survival time of only 9-15 months following standard therapies. In addition to patient survival, evaluation of treatment response commonly relies on radiographic response rates, specifically changes in the area of contrast enhanced tumor as seen on a single brain MRI slice. This is problematic for several reasons. First, subjects'heads are placed differently in the scanner at each visit, inducing variability in the angle and position of the image slices, leading to an apparent change in tumor area unrelated to treatment. Second, agents used in anti-angiogenic therapies (AAT) cause a pseudoresponse, i.e. a rapid and profound decrease in contrast enhancement due to reduced vascular permeability rather than a true anti-tumor effect. Advanced MRI techniques such as perfusion, permeability, and diffusion imaging, as well as positron emission tomography show promise in predicting response and quantifying changes in non-enhancing tumor components. However, the Response Assessment in Neuro-Oncology (RANO) Group concluded that these techniques are still too limited to derive an objective measure of non-enhancing disease and require rigorous validation before they can be incorporated into response criteria used in clinical trials in high-grade gliomas. This project's core objective is to improve response assessment in AAT. We will address current limitations by (i) reducing acquisition variability via a novel automatic slice prescription method that ensures acquisition of identical image slices as in previous scans of the same patient, (ii) identifying pseudoresponse via quantification of deformations across visits (mass effects) in peritumoral tissue, and (iii) assessing perfusion and correlation with local response in tumor components at the voxel level through a temporal multi-parametric model combining several sequences/modalities and information across visits. Both simulated and retrospective clinical data (40 patients, 10-30 time points) will inform the development of the computational tools and markers. Subsequently, response assessment and outcome prediction in prospective data (48 patients, monthly scans) will be used for validation to address the need for rigorous evaluation. This project will deliver advanced tools that will be made freely available. The candidate brings a strong background in math and computer science and a unique skill set related to the development of computational tools, e.g. image registration. A strong team of mentors and collaborators will support the candidate's training in neuro-oncology, angiogenesis, and physiological and functional imaging modalities such as PET and MR. This award will enable him to focus his future research primarily on questions of health and disease, specifically, to establish an interdisciplinary tumor imaging research program, and to successfully transfer computational methods and imaging biomarkers into clinical practice.
The proposed project will promote the rigorous clinical validation of existing and novel MRI markers, and enable progress along three directions relevant to tumor treatment: (1) support individualized treatment decisions and help to eliminate ineffective regimens more rapidly via early and sensitive response evaluation, (2) facilitate differentiation of progressive non-enhancing disease from responding disease, and (3) help to understand AAT induced temporal and spatial vascular modifications and effect on tumor response.
|Wachinger, Christian; Reuter, Martin; Klein, Tassilo (2018) DeepNAT: Deep convolutional neural network for segmenting neuroanatomy. Neuroimage 170:434-445|
|Saygin, Z M; Kliemann, D; Iglesias, J E et al. (2017) High-resolution magnetic resonance imaging reveals nuclei of the human amygdala: manual segmentation to automatic atlas. Neuroimage 155:370-382|
|Aganj, Iman; Iglesias, Juan Eugenio; Reuter, Martin et al. (2017) Mid-space-independent deformable image registration. Neuroimage 152:158-170|
|Wachinger, Christian; Salat, David H; Weiner, Michael et al. (2016) Whole-brain analysis reveals increased neuroanatomical asymmetries in dementia for hippocampus and amygdala. Brain 139:3253-3266|
|Ge, Tian; Reuter, Martin; Winkler, Anderson M et al. (2016) Multidimensional heritability analysis of neuroanatomical shape. Nat Commun 7:13291|
|Yendiki, Anastasia; Reuter, Martin; Wilkens, Paul et al. (2016) Joint reconstruction of white-matter pathways from longitudinal diffusion MRI data with anatomical priors. Neuroimage 127:277-286|
|Tisdall, M Dylan; Reuter, Martin; Qureshi, Abid et al. (2016) Prospective motion correction with volumetric navigators (vNavs) reduces the bias and variance in brain morphometry induced by subject motion. Neuroimage 127:11-22|
|Iglesias, Juan Eugenio; Van Leemput, Koen; Augustinack, Jean et al. (2016) Bayesian longitudinal segmentation of hippocampal substructures in brain MRI using subject-specific atlases. Neuroimage 141:542-555|
|Wachinger, Christian; Reuter, Martin; Alzheimer's Disease Neuroimaging Initiative et al. (2016) Domain adaptation for Alzheimer's disease diagnostics. Neuroimage 139:470-479|
|Wachinger, Christian; Golland, Polina; Magnain, Caroline et al. (2015) Multi-modal robust inverse-consistent linear registration. Hum Brain Mapp 36:1365-80|
Showing the most recent 10 out of 14 publications