The Miller Atlas, also known as Lopo Homem-Reineis Atlas, is a richly illustrated Portuguese partial world atlas dated from 1519, including a dozen charts. It is a joint work of the cartographers Lopo Homem, Pedro Reinel and Jorge Reinel, and illustrated by miniaturist António de Holanda.
The regions represented are the North Atlantic Ocean, Northern Europe, the Azores Archipelago, Madagascar, Horn of Africa, the Indian Ocean, Indonesia, the China Sea, the Moluccas, Brazil and the Mediterranean Sea. It was acquired by the librarian Bénigne Emmanuel Clement Miller in 1855 at a bookseller in Santarém, Portugal, hence the name Miller Atlas. In 1897, his widow sold it to the National Library of France,[1] where it has stayed ever since.
Atlas Da Terra Media.pdfl
Functional magnetic resonance imaging (fMRI) BOLD signal is commonly localized by using neuroanatomical atlases, which can also serve for region of interest analyses. Yet, the available MRI atlases have serious limitations when it comes to imaging subcortical structures: only 7% of the 455 subcortical nuclei are captured by current atlases. This highlights the general difficulty in mapping smaller nuclei deep in the brain, which can be addressed using ultra-high field 7 Tesla (T) MRI. The ventral tegmental area (VTA) is a subcortical structure that plays a pivotal role in reward processing, learning and memory. Despite the significant interest in this nucleus in cognitive neuroscience, there are currently no available, anatomically precise VTA atlases derived from 7 T MRI data that cover the full region of the VTA. Here, we first provide a protocol for multimodal VTA imaging and delineation. We then provide a data description of a probabilistic VTA atlas based on in vivo 7 T MRI data.
Neuroanatomical atlases can be used to localize functional magnetic resonance imaging (fMRI) BOLD signal in univariate (voxel-wise) analyses, or to extract BOLD signal in the region of interest analyses (Poldrack 2006). Currently, only 7% of the 455 subcortical nuclei are included in the available magnetic resonance imaging (MRI) atlases. This highlights the general difficulty in mapping smaller subcortical structures (Alkemade et al. 2013; Forstmann et al. 2017). One prominent subcortical nucleus is the ventral tegmental area (VTA). The VTA is located in the midbrain and contains dopaminergic neurons, which are crucial in reward-based learning and motor functions (Schultz 1998, 2015). Although some digital VTA atlases are available for neuroimaging purposes, there is a lack in availability of anatomically precise atlases derived from high-resolution 7 Tesla (T) MRI data that emphasize taking into account individual anatomical variability and precision.
Due to anatomical variability in the subcortex (Keuken et al. 2014), creating probabilistic atlases that capture this variability is crucial and particularly important for the VTA. The VTA is relatively small, with a size of circa 140 mm3, has a complex shape (Halliday and Törk 1986), and lacks a clear anatomical border with the surrounding nuclei. For all these reasons, delineating the VTA on individual MRI scans is very challenging. Despite these challenges, some efforts have been made to provide digital VTA atlases (Murty et al. 2014; Pauli et al. 2018). Eapen et al. (2011) and Ballard et al. (2011) set remarkable groundwork for segmenting the VTA on MRI data and Murty et al. (2014) provided the first publicly available probabilistic atlas of the VTA. Recently, Pauli et al. (2018) released another probabilistic atlas of the VTA by making use of the large publicly available data the field has gathered since. Accordingly, the atlas by Pauli et al. (2018) was aimed to emphasize anatomical variability, but interestingly the authors also applied a different VTA terminology compared to Murty et al. (2014).
The atlas introduced in this paper adds to the collection of probabilistic VTA atlases by providing another atlas that emphasizes anatomical precision next to anatomical variability which was the focus of previous atlases. Importantly, we acknowledge previous efforts; however, there are several challenges in the construction of VTA atlases that can be facilitated today due to developments in the field of ultra-high-field MRI. For example, given the relatively small size and location of the VTA, a scan protocol with a submillimeter voxel-size resolution is crucial (Ewert et al. 2018). Also, a 7T scan protocol helps to acquire improved signal-to-noise ratios (SNR) and contrast-to-noise ratios (CNR) compared to lower field strength (van der Zwaag et al. 2015). Furthermore, the optimization of the MR sequences by means of tailoring them to subcortical structures such as the VTA is required for individual anatomical precision (Trutti et al. 2019).
Of course, all efforts to provide probabilistic VTA atlases are limited by some factors and therefore tailored to different approaches. In the atlas of Pauli et al. (2018) anatomical variability was taken into account before manual segmentation was carried out by means of computing eight, unilateral group templates on which three raters manually segmented the VTA. This resulted in eight, unilateral VTA segmentations for each rater despite the abundance of individual data, as the data originated from the Human Connectome Project (HCP). Yet, manual segmentations are very time consuming and the authors provided many subcortical nuclei in the published atlas. Thus, their efforts were not limited to segmenting the VTA. Further, since they worked with HCP data, they were also restricted to the available scans. To get images that allow segmentation of subcortical nuclei, the construction of group templates was required.
In contrast, the atlas from Murty et al. (2014) is based on manual segmentation of the VTA on a large number of participants. Nonetheless, as the field rapidly develops the data quality is not state-of-the-art anymore. The resolution (1 mm isotropic) and CNR of the used T1-weighted MRI sequence render the incorporation of small anatomical differences needed for the segmentation of the VTA very challenging. These issues can be addressed again today with new scan protocols and higher field strengths.
In this study, we first present an optimized 7T MRI imaging protocol to delineate the VTA based on a well-established VTA terminology (Trutti et al. 2019). According to Halliday and Törk (1986) as well as Mai et al. (2016), the VTA covers a region that includes the parabrachial pigmented nucleus (PBP), paranigral nucleus (PN), interfascicular nucleus (IF), the caudal linear nucleus (CLi), and the rostral linear nucleus (RLi). Hence, the segmentation protocol was built so that the VTA masks could contain the different VTA component nuclei rather than to identify the individual component nuclei themselves, which is hardly possible even at 7T. We then demonstrate the capabilities of this optimized protocol to define a probabilistic VTA atlas based on 7T in vivo MRI data from 27 healthy participants.
The probabilistic atlas is based on twenty-seven healthy, young, right-handed participants (19 females) with a mean age of 24.5 (SD 4.8). All participants had normal or corrected-to-normal vision, and none of them had a history of neurological, major medical, or psychiatric disorders. The study was approved by the local ethics committee at the University of Amsterdam, the Netherlands. All participants gave their written informed consent prior to scanning and received monetary compensation for participating. This study was performed in line with the principles of the Declaration of Helsinki.
All linear registration steps were done using MIPAV 5.4.4. ( ) with the optimized automated linear registration algorithm. Whole-brain images were skull-stripped using the standard FSL BET tool (Smith 2002). The registration to atlas space was done by means of non-linearly aligning the individual whole-brain images to a group average template of the AHEAD database (Alkemade et al. 2020) which itself was co-registered to the MNI152 2009b template with the ANTs algorithm (Avants et al. 2008), using conservative deformation parameters as recommended for subcortex (Ewert et al. 2018). The individual VTA masks and midbrain contrasts were non-linearly transformed into MNI152 2009b space using the computed transformations. All registration steps were visually checked for misalignments by comparing several landmarks: fourth ventricle, pons, corpus callosum, and lateral ventricles. Because the AHEAD template is based on the same MP2RAGE-ME sequence that provides additional subcortical contrast compared to the MNI152 2009b template, the automated co-registration of the subcortical regions was satisfactory.
The unthresholded probabilistic atlas volume was 2226.375 mm3 for the left and 2368.00 mm3 for the right hemisphere. To increase the probability of the atlas voxels belonging to the VTA, thresholds can be applied. Each voxel has a particular probability of belonging a structure, based on the overlap across the 27 individually segmented masks. A threshold of 50%, for example, excludes all VTA atlas voxels that are not shared across at least 50% of the subjects (see Fig. 3 for the volume of the probabilistic atlas across different lower thresholds and Fig. 4 for three-dimensional reconstructions of the atlas).
VTA mask and atlas volume. a Distribution of the volume of individual VTA conjunction masks. b VTA atlas volume across different thresholds. The size of the VTA atlas decreases when discarding the voxels that have lower probabilities of belonging to the VTA across participants. Markers are colored according to the thresholds in c. c Effect of thresholding on the VTA atlas volume, shown on the right VTA atlas. Voxels inside the differently colored outlines fall within the atlas mask when the thresholds corresponding to those colors are applied. For example, most of the darker voxels suggesting a lower probability, and only those, would be excluded with a threshold of 0.2 as indicated by the bright red outline. In contrast, picking a high threshold (blue colored lines) would mostly include voxels in the ventromedial VTA region 2ff7e9595c
Comments