Segmentation of serial MRI of TBI patients using personalized atlas construction and topological change estimation

Abstract

Traumatic brain injury (TBI) due to falls, car accidents, and warfare affects millions of people annually. Determining personalized therapy and assessment of treatment efficacy can substantially benefit from longitudinal (4D) magnetic resonance imaging (MRI). In this paper, we propose a method for segmenting longitudinal brain MR images with TBI using personalized atlas construction. Longitudinal images with TBI typically present topological changes over time due to the effect of the impact force on tissue, skull, and blood vessels and the recovery process. We address this issue by defining a novel atlas construction scheme that explicitly models the effect of topological changes. Our method automatically estimates the probability of topological changes jointly with the personalized atlas. We demonstrate the effectiveness of this approach on MR images with TBI that also have been segmented by human raters, where our method that integrates 4D information yields improved validation measures compared to temporally independent segmentations.

Integration of Multimodal Neuroimaging and Electroencephalography for the Study of Acute Epileptiform Activity After Traumatic Brain Injury

“…For example, the presence of lesions can be associated with regions of substantial hyper-or hypo-intensities across imaging modalities, which makes the use of histogram-matching algorithms problematic. In our studies, the challenges described above are typically addressed using sophisticated, TBI-tailored interpolation algorithms available within the LONI Pipeline environment which are described in detail elsewhere [20][21][22].…”

Modeling 4D pathological changes by leveraging normative models

“…A common procedure for supervised classification is to let a user define regions of interest of the different types of pathology categories in an affine-registered normative atlas, for example by drawing spherical regions for lesion types (Prastawa et al, 2004; Wang et al, 2012). The expert input, number of lesion types, and an affine-registered atlas are then used to initialize the parameters of the multivariate Gaussian probability distribution (${\mu }_t^c$ and ${\sum }_t^c$), and spatial prior ${\prod }_t^c$ via these user-input spheres S t for each tissue/lesion class c = 1, …, k t , k l , …, K where c = 1, …, k t are normal tissue classes and c = k l , …, K are lesion classes.…”

Low-Rank to the Rescue – Atlas-Based Analyses in the Presence of Pathologies

“…Other methods attempt to address this problem by joint registration and segmentation which tolerates missing correspondences [1], geometric metamorphosis that separates estimating healthy tissue deformation from modeling tumor change [5], or personalized atlas construction that accounts for diffeomorphic and non-diffeomorphic changes [9]. While effective, these methods require explicit lesion segmentations or initial lesion localizations, which, in this case, is actually the goal of the process.…”