Rigid and non-rigid registration of polarized light imaging data for 3D reconstruction of the temporal lobe of the human brain at micrometer resolution

• Verfasst von: Ali, Sharib [VerfasserIn]
• Verfasst von: Wörz, Stefan [VerfasserIn]
• Verfasst von: Amunts, Katrin [VerfasserIn]
• Verfasst von: Eils, Roland [VerfasserIn]
• Verfasst von: Axer, Markus [VerfasserIn]
• Verfasst von: Rohr, Karl [VerfasserIn]
• Titel: Rigid and non-rigid registration of polarized light imaging data for 3D reconstruction of the temporal lobe of the human brain at micrometer resolution
• Verf.angabe: Sharib Ali, Stefan Wörz, Katrin Amunts, Roland Eils, Markus Axer, Karl Rohr
• E-Jahr: 2018
• Jahr: 6 July 2018
• Umfang: 17 S.
• Fussnoten: Gesehen am 04.05.2020
• Titel Quelle: Enthalten in: NeuroImage
• Ort Quelle: Orlando, Fla. : Academic Press, 1992
• Jahr Quelle: 2018
• Band/Heft Quelle: 181(2018), Seite 235-251
• ISSN Quelle: 1095-9572
• DOI: doi:10.1016/j.neuroimage.2018.06.084
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: 3D reconstruction
• Sach-SW: Canonical correlation transform
• Sach-SW: Hippocampus
• Sach-SW: Human brain
• Sach-SW: Image registration
• Sach-SW: Polarized light imaging
• K10plus-PPN: 1697106218

Abstract

To understand the spatial organization as well as long- and short-range connections of the human brain at microscopic resolution, 3D reconstruction of histological sections is important. We approach this challenge by reconstructing series of unstained histological sections of multi-scale (1.3μm and 64μm) and multi-modal 3D polarized light imaging (3D-PLI) data. Since spatial coherence is lost during the sectioning procedure, image registration is the major step in 3D reconstruction. We propose a non-rigid registration method which comprises of a novel multi-modal similarity metric and an improved regularization scheme to cope with deformations inevitably introduced during the sectioning procedure, as well as a rigid registration approach using a robust similarity metric for improved initial alignment. We also introduce a multi-scale feature-based localization and registration approach for mapping of 1.3μm sections to 64μm sections and a scale-adaptive method that can handle challenging sections with large semi-global deformations due to tissue splits. We have applied our registration method to 126 consecutive sections of the temporal lobe of the human brain with 64μm and 1.3μm resolution. Each step of the registration method was quantitatively evaluated using 10 different sections and manually determined ground truth, and a quantitative comparison with previous methods was performed. Visual assessment of the reconstructed volumes and comparison with reference volumes confirmed the high quality of the registration result.