Diffusion tensor driven image registration: a deep learning approach

TY - CHAP

T1 - Diffusion tensor driven image registration

T2 - 9th International Workshop on Biomedical Image Registration, WBIR 2020

AU - Grigorescu, Irina

AU - Uus, Alena

AU - Christiaens, Daan

AU - Cordero-Grande, Lucilio

AU - Hutter, Jana

AU - Edwards, A. David

AU - Hajnal, Joseph V.

AU - Modat, Marc

AU - Deprez, Maria

PY - 2020/1/1

Y1 - 2020/1/1

N2 - Tracking microsctructural changes in the developing brain relies on accurate inter-subject image registration. However, most methods rely on either structural or diffusion data to learn the spatial correspondences between two or more images, without taking into account the complementary information provided by using both. Here we propose a deep learning registration framework which combines the structural information provided by $$T:2$$-weighted ($$T:2$$w) images with the rich microstructural information offered by diffusion tensor imaging (DTI) scans. This allows our trained network to register pairs of images in a single pass. We perform a leave-one-out cross-validation study where we compare the performance of our multi-modality registration model with a baseline model trained on structural data only, in terms of Dice scores and differences in fractional anisotropy (FA) maps. Our results show that in terms of average Dice scores our model performs better in subcortical regions when compared to using structural data only. Moreover, average sum-of-squared differences between warped and fixed FA maps show that our proposed model performs better at aligning the diffusion data.

AB - Tracking microsctructural changes in the developing brain relies on accurate inter-subject image registration. However, most methods rely on either structural or diffusion data to learn the spatial correspondences between two or more images, without taking into account the complementary information provided by using both. Here we propose a deep learning registration framework which combines the structural information provided by $$T:2$$-weighted ($$T:2$$w) images with the rich microstructural information offered by diffusion tensor imaging (DTI) scans. This allows our trained network to register pairs of images in a single pass. We perform a leave-one-out cross-validation study where we compare the performance of our multi-modality registration model with a baseline model trained on structural data only, in terms of Dice scores and differences in fractional anisotropy (FA) maps. Our results show that in terms of average Dice scores our model performs better in subcortical regions when compared to using structural data only. Moreover, average sum-of-squared differences between warped and fixed FA maps show that our proposed model performs better at aligning the diffusion data.

KW - Diffusion tensor imaging

KW - Image registration

UR - http://www.scopus.com/inward/record.url?scp=85087018200&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-50120-4_13

DO - 10.1007/978-3-030-50120-4_13

M3 - Conference paper

AN - SCOPUS:85087018200

SN - 9783030501198

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 131

EP - 140

BT - Biomedical Image Registration - 9th International Workshop, WBIR 2020, Proceedings

A2 - Spiclin, Ziga

A2 - McClelland, Jamie

A2 - Kybic, Jan

A2 - Goksel, Orcun

PB - SPRINGER

Y2 - 1 December 2020 through 2 December 2020

ER -