FIBRASCAN: A novel method for 3D white matter tract reconstruction in MR space from cadaveric dissection

Highlights

• We developed a method for 3D reconstruction of cerebral fiber tracts from dissection.

• Iterative surface acquisitions of human specimens were used to monitor dissection.

• Surface acquisitions were obtained with a laser scanner and digital cameras.

• As an example, six association tracts were interactively segmented and reconstructed.

• The 3D-reconstructed tracts were then ported to the ex vivo MR reference space.

Abstract

Introduction

Diffusion tractography relies on complex mathematical models that provide anatomical information indirectly, and it needs to be validated. In humans, up to now, tractography has mainly been validated by qualitative comparison with data obtained from dissection. No quantitative comparison was possible because Magnetic Resonance Imaging (MRI) and dissection data are obtained in different reference spaces, and because fiber tracts are progressively destroyed by dissection. Here, we propose a novel method and software (FIBRASCAN) that allow accurate reconstruction of fiber tracts from dissection in MRI reference space.

Method

Five human hemispheres, obtained from four formalin-fixed brains were prepared for Klingler's dissection, placed on a holder with fiducial markers, MR scanned, and then dissected to expose the main association tracts. During dissection, we performed iterative acquisitions of the surface and texture of the specimens using a laser scanner and two digital cameras. Each texture was projected onto the corresponding surface and the resulting set of textured surfaces was coregistered thanks to the fiducial holders. The identified association tracts were then interactively segmented on each textured surface and reconstructed from the pile of surface segments. Finally, the reconstructed tracts were coregistered onto ex vivo MRI space thanks to the fiducials. Each critical step of the process was assessed to measure the precision of the method.

Results

We reconstructed six fiber tracts (long, anterior and posterior segments of the superior longitudinal fasciculus; Inferior fronto-occipital, Inferior longitudinal and uncinate fasciculi) from cadaveric dissection and ported them into ex vivo MRI reference space. The overall accuracy of the method was of the order of 1 mm: surface-to-surface registration = 0.138 mm (standard deviation (SD) = 0.058 mm), deformation of the specimen during dissection = 0.356 mm (SD = 0.231 mm), and coregistration surface-MRI = 0.6 mm (SD = 0.274 mm). The spatial resolution of the method (distance between two consecutive surface acquisitions) was 0.345 mm (SD = 0.115 mm).

Conclusion

This paper presents the robustness of a novel method, FIBRASCAN, for accurate reconstruction of fiber tracts from dissection in the ex vivo MR reference space. This is a major step toward quantitative comparison of MR tractography with dissection results.

Introduction

Cerebral white matter fiber tract anatomy is of great importance for greater understanding of brain physiology and physiopathology, and for diagnosis and follow-up of degenerative, inflammatory or tumoral diseases. Diffusion tractography is the only non-invasive technique for localizing cerebral white matter fiber tracts in humans. Diffusion Tensor Imaging (DTI), which models diffusion as an ellipsoid for each voxel (Basser et al., 1994, Le Bihan and Breton, 1985), is widely used in both clinical practice and research to evaluate fiber tract pathways (Mori and van Zijl, 2002). DTI-based tractography oversimplifies fiber tract anatomy because it only considers a single direction of diffusion per voxel. Alternative methods for data acquisition, such as High Angular Resolution Diffusion Imaging (Tuch et al., 2002), and post-processing, such as spherical deconvolution (Alexander, 2005, Tournier et al., 2004) and probabilistic algorithms (Behrens et al., 2003, Behrens et al., 2007, Parker and Alexander, 2005), have been proposed for the detection and reconstruction of several populations of fibers in a single voxel. However, although tractography has already been used for pre-operative planification and per-operative navigation in patients operated on for cerebral tumors (Nimsky et al., 2006, Yu et al., 2005), it has sometimes been shown to lack reliability (Kinoshita et al., 2005, Nimsky et al., 2005). Indeed, diffusion tractography relies on complex acquisition methods and post-processing mathematical models, which provide anatomical information indirectly, and therefore needs to be validated. Many validation methods on phantoms, animal and human anatomical specimens have been proposed, but none are entirely satisfactory.

Phantoms include biological and synthetic objects modeling water diffusion in biological tissues, and artificial data. Tractography is possible in biological phantoms mimicking diffusion in the brain, such as muscle, spinal cord or plants (Basser et al., 1994, Latt et al., 2007). However, the fiber route is relatively simple in these specimens and they cannot be used over a long period of time due to their biological nature. Synthetic phantoms made of capillaries (Lin et al., 2003, Yanasak and Allison, 2006) remain very simple compared to the complex microarchitecture of the brain. For this reason, phantoms including crossing or angular fibers that are closer to the organization of white matter have been built using various textile fibers with known directions and calibers: rayon (Perrin et al., 2005), polyester (Pullens et al., 2010), and acrylic (Fillard et al., 2011, Poupon et al., 2008). Finally, artificial data are computer-generated and are thought to be similar to DTI (Chen and Song, 2008). They can be used to assess algorithms but of course not to check the anatomical validity of the acquired data.

In animals, tractography has been compared with autoradiography in monkeys (Dauguet et al., 2007, Schmahmann et al., 2007), with a lack of perfect concordance, possibly because of difficulty reconstructing a histological volume from autoradiographic slices before registering it onto in vivo MRI. Manganese has been proposed to limit difficulty related to histological preparation (Dyrby et al., 2007, Lin et al., 2001, Lin et al., 2003); once applied on the animal cortex, it acts as a tracer that follows the neighboring axons. Marked axons gain paramagnetic properties from the manganese and appear hyperintense on T1-weighted MRI. Direct comparison of tractography with manganese tracing on T1-weighted MRI is therefore possible with no need for histological preparation. Nevertheless, even when manganese is stereotactically injected, regional diffusion occurs, marking a large area of the brain (Dyrby et al., 2007, Lin et al., 2001).

For obvious ethical reasons, only ex vivo techniques are acceptable in humans. Ex vivo axonal tracing using silver (Mesulam, 1979) or Di-I (1,1-dioctadecy 1–3,3,3′,3′-tetramethyl lindocarbocyanine perchlorate) (Sparks et al., 2000) have been proposed. Unfortunately, this stain only diffuses to 20 to 40 mm from the injection point, which is insufficient for large fiber tracts. Polarized light imaging (PLI) records light transmission through an anatomical slice; due to optical birefringence of the myelin sheaths, light transmission depends on the relative orientations of light polarization and fiber tracts (Axer et al., 2001, Dammers et al., 2010, Palm et al., 2010). Refinements of this promising technique can evaluate fiber direction at a micrometer-scale resolution but, as it uses slices, it suffers from the same limitation as histological preparation for axonal tracing, i.e. loss of 3D coherence of fiber tracts. Optical Coherence Tomography (OCT) directly images the fiber direction at the surface of the specimen with (Wang et al., 2011) or without (Magnain et al., 2014) polarized light. As it studies this orientation directly at the surface of the blockface, and not on slices, it does not suffer from artifacts induced by specimen slicing. Convincing images have been published for imaging the rat (Wang et al., 2011) but not the whole human brain, mainly because of technical limitations, especially specimen size. Direct comparison of ex vivo tractography and dissection in the same specimens is of course a very attractive but challenging approach (D'Arceuil et al., 2007). Post-mortem scanning of fixed specimens has at least two advantages: long scanning time and absence of physiological and motion noise. Recently, a pipeline for high-quality ex vivo scanning using limited b-value (4000 s/mm²) and gradient strength value (56 mT/m) was proposed by Dyrby et al. (2011) in sedated pigs transcardially perfused with formalin. As demonstrated by D'Arceuil and de Crespigny (2007), such an immediate fixation limits the drop of apparent diffusion coefficient (ADC), T2 and signal-to-noise ratio (SNR) occuring post-mortem but cannot be obtained in humans for evident reasons. Consequently, the resulting ADC drop needs to be corrected by modified scanning conditions. For instance, increasing gradient strength was proposed to preserve ex vivo DWI image quality (D'Arceuil and de Crespigny, 2007). It indeed allows high b-values to be achieved with much shorter echo times, and thus much higher SNR than on conventional scanners.

White matter dissection of formalin-fixed anatomical specimens after freezing and defrosting was first proposed by Klingler (Klingler, 1935, Klingler and Gloor, 1960, Ludwig and Klingler, 1956). This method, which helps the dissection of white matter tracts, is widely considered to be a gold standard to validate tractography results in the human, and good concordance has generally been found between tractography and dissection (Catani et al., 2002, Catani et al., 2005, Catani and Thiebaut de Schotten, 2008, Lawes et al., 2008, Mori et al., 2002, Mori et al., 2008, Oishi et al., 2008, Peltier et al., 2010a, Peltier et al., 2010b, Stieltjes et al., 2001, Wakana et al., 2004). Nevertheless, this finding has certain limitations and can only be considered at a coarse scale for two main reasons: (1) results from dissection and tractography are obtained in different reference spaces (anatomy laboratory and MR scanner, respectively), and often from different individuals, allowing qualitative but not quantitative comparisons; and (2) Klingler's method is destructive because superficial layers of the studied specimen and tract are removed to study inner anatomical structures, making it difficult to study the relationships of fiber tracts to the cortex and between fiber tracts.

We have developed an original method and software, FIBRASCAN: (1) to monitor dissection of white matter tracts by iterative surface and texture acquisitions of the specimen; (2) to interactively segment tracts on these surfaces; (3) to reconstruct tracts from these segmented surfaces; and (4) to port the reconstructed tracts into ex vivo MR anatomical space.

We present this method below and illustrate it with the reconstruction of four association fiber tracts: the arcuate fasciculus (AF), the Inferior fronto-occipital fasciculus (IFOF), the Inferior longitudinal fasciculus (ILF), and the uncinate fasciculus (UF).