Document Type

Article

Publication Date

12-13-2014

Journal

Journal of neuroscience methods

Volume

241C

First Page

53

Last Page

65

URL with Digital Object Identifier

10.1016/j.jneumeth.2014.12.005

Abstract

BACKGROUND: Advances in MRI have the potential to improve surgical treatment of epilepsy through improved identification and delineation of lesions. However, validation is currently needed to investigate histopathological correlates of these new imaging techniques. The purpose of this work is to develop and evaluate a protocol for deformable image registration of in-vivo to ex-vivo resected brain specimen MRI. This protocol, in conjunction with our previous work on ex-vivo to histology registration, completes a registration pipeline for histology to in-vivo MRI, enabling voxel-based validation of novel and existing MRI techniques with histopathology.

NEW METHOD: A combination of image-based and landmark-based 3D registration was used to register in-vivo MRI and the ex-vivo MRI from patients (N=10) undergoing epilepsy surgery. Target registration error (TRE) was used to assess accuracy and the added benefit of deformable registration.

RESULTS: A mean TRE of 1.35±0.11 and 1.41±0.33mm was found for neocortical and hippocampal specimens respectively. Statistical analysis confirmed that the deformable registration significantly improved the registration accuracy for both specimens.

COMPARISON WITH EXISTING METHODS: Image registration of surgically resected brain specimens is a unique application which presents numerous technical challenges and that have not been fully addressed in previous literature. Our computed TRE are comparable to previous attempts tackling similar applications, as registering in-vivo MRI to whole brain or serial histology.

CONCLUSION: The presented registration pipeline finds dense and accurate spatial correspondence between in-vivo MRI and histology and allows for the spatially local and quantitative assessment of pathological correlates in MRI.


Find in your library

Share

COinS