Description of BrainRAT


You can find the documentation of the toolbox on brainvisa web site.


BrainRAT results from collaborative work of image processing methodologists and biologists of MIRCen, a new CEA facility headed by Philippe Hantraye and dedicated to preclinical research.

BrainRAT is mainly developed by the MIRCen image processing team of Thierry Delzescaux, PhD (Julien Dauguet, PhD, Jessica Lebenberg, PhD student, N. Souedet, software engineer and Didier Thenadey, IT specialist) as well as by some close collaborators (Albertine Dubois, PhD, Inserm U803, Vincent Frouin, CEA/SCSR, Denis Rivière, PhD and Yann Cointepas, PhD, CEA/NeuroSpin).

Expert biologists have been associated with this project, from the beginning, to guarantee the relevance of developed methods and the compatibility/performances of processing times with biological experiments (Anne-Sophie Hérard, PhD, Gilles Bonvento, PhD, and Françoise Condé, PhD).

All the above-mentioned people can be contacted by e-mail at


BrainRAT's goal is to provide an automated and generic toolbox for the processing of histological and autoradiographic sections based on a two-step approach, which involves:


At the beginning, these methods have been developed to analyse autoradiographic data in 3D using complementary histological information. The 3D reconstruction of biological volumes arising from several histological and/or autoradiographic sections enables to take advantage of the high resolution of ex vivo autoradiography as well as of the diversity of available histological stainings.

The analysis of autoradiographic data, 3D reconstructed or not, remains traditionally limited to conventional region of interest (ROI)-based analysis. Although the users have the corresponding histological stained sections available to consider anatomical information, the delineation of ROI(s) to be analyzed is usually directly performed on the autoradiographic sections, which may be inaccurate and subject to observer bias. To overcome this limitation, BrainRAT enables to perform multimodal co-registration between slices or 3D reconstructed volumes.

Recent studies have highlighted the ability of voxel-wise statistical analysis to deal with 3D reconstructed autoradiographic volumes for group comparisons. This automated, exploratory, whole-brain approach can be used to analyze local functional differences between population groups without the need for prior assumptions. BrainRAT provides useful tools to carry out such studies in optimal conditions (optimized digitization, 3D reconstruction algorithms, activity conversion and intensity normalization procedures).

These methods have been applied to Nissl stained (cresyl violet) sections and [14C]-2DG autoradiography and intensively validated both in rats and mice. They have been also successfully tested on other post mortem imaging modalities such as immunohistochemistry, fluorescence, etc. They could now be extended to various biological applications (activation models, morphology, lesion extension, development, etc.) and to domains of application including pathological and therapeutic studies involving animal models of various brain diseases (neurodegenerative diseases, ischemia, and cancer).


For the time being, the only reference to BrainRAT is related to proceedings of the SFN 2008:

BrainRAT: Brain Reconstruction and Analysis Toolbox. A freely available toolbox for the 3D reconstruction of anatomo-functional brain sections in rodents.
A. Dubois, J. Dauguet, N. Souedet, A.-S. Hérard, D. Rivière, Y. Cointepas, G. Bonvento, P. Hantraye, V. Frouin, T. Delzescaux. In Proc. 38th annual meeting of the Society for Neuroscience, Washington, USA, 2008.

The main reference relating to the computerized treatments and procedures gathered in BrainRAT to be mentioned is:

Automated three-dimensional analysis of histological and autoradiographic rat brain sections: application to an activation study.
A. Dubois, J. Dauguet, A-S. Hérard, L. Besret, E. Duchesnay, V. Frouin, P. Hantraye, G. Bonvento and T. Delzescaux. Journal of Cerebral Blood Flow and Metabolism, 2007, 27(10), 1742-1755. Abstract.

Yet, these developments were the basis of other works published by our team:

They enabled to provide evidence that the metabolic response to synaptic activation (i.e. enhancement of glucose uptake) is decreased in the superior colliculus during visual stimulation in young adult mice deficient for the glial glutamate transporter GLT-1 (Figure 1.1).

Decreased metabolic response to visual stimulation in the superior colliculus of mice lacking the glial glutamate transporter GLT-1.
A-S. Herard, A. Dubois, C. Escartin, K. Tanaka, T. Delzescaux, P. Hantraye and G.Bonvento. European Journal of Neuroscience, 2005, 22(7):1807-11. Abstract.

Fig. 1.1: (Left) 3-D reconstruction of cresyl violet-stained sections clearly outlines the superior colliculus. (Right) The corresponding 3-D reconstruction of autoradiographic images indicates the extent of the metabolic activation during checkerboard stimulation in a control C57BL/6 mouse with one eye shut and one eye open.

They enabled to accurately study stimulus-driven synaptic activity in vivo in a restricted rat brain region, the superior colliculus (Figure 1.2).

siRNA targeted against amyloid precursor protein impairs synaptic activity in vivo.
A-S. Hérard, L. Besret, A. Dubois, J. Dauguet, T. Delzescaux, P. Hantraye, G. Bonvento and K.L. Moya. Neurobiology of Aging, 2006, 27(12):1740-50. Abstract.

Fig. 1.2: 3-D reconstruction of the left and right manually segmented superior colliculus. The automatic extraction of the "activated" volume using the threshold method is represented in red and the symmetric volume in green.

They also enabled the investigation of new methods of analysis of 3D biologic data in groups of rodents, based on voxel-wise statistical approaches without the need for prior assumptions (Figure 1.3).

Quantitative validation of voxel-wise statistical analyses of autoradiographic rat brain volumes: application to unilateral visual stimulation.
A. Dubois, A-S. Hérard, E. Duchesnay, G. Flandin, L. Besret, P. Hantraye, G. Bonvento and T. Delzescaux. NeuroImage, 2008, 40(2), 482-494. Abstract.

Fig. 1.3: Brain area (the superior colliculus) in which CMRGlu was significantly higher or lower in visually stimulated hemibrains than in the corresponding control hemibrains automatically detected by a voxel-wise statistical analysis. Significance is indicated with t statistic color scales, corresponding to the level of significance at voxel level.

Lastly, these developments were also the basis of works performed in collaboration with other teams :

Metabolic and vascular changes induced in mice olfactory glomeruli by odor presentation (Figure 1.4).
Gurden et al. IMNC, UMR8165 CNRS-University of Paris 7 and Paris 11, Orsay, France. 5th Forum of European Neuroscience (FENS), Vienna, Austria, 2006.

Fig. 1.4: Anatomo-functional superimposition and 3D reconstruction of the olfactory bulb.

The potential of the radiosensitive β-microprobe to monitor the [18F]-MPPF binding in the mouse hippocampus in vivo (Figure 1.5).
Desbrée et al. IMNC, UMR8165 CNRS-University of Paris 7 and Paris 11, Orsay, France. Journal of Nuclear Medicine, 2008, 49(7):1155-61. Abstract.

Fig. 1.5: 3D postmortem imaging of radioligand distribution in mouse brain after injection of [18F]-MPPF. Background black and white anatomic volumetric reconstruction of cresyl violet-stained sections shows bilateral location of 5-HT1A binding in mouse hippocampus. Pseudocolored digitized autoradiographs illustrate level of [18F]-MPPF binding and are color-coded from low (blue) to high (red) binding.

Ongoing work

The next releases of BrainRAT toolbox will contain new processes allowing to:

  1. Speed up of registration processes (optimization, parallelization);
  2. Use blockface images (photographs of the brain face taken during the sectioning process) as a consistent 3D geometric reference for reconstruction of post mortem volumes;
  3. Bring in vivo data such as MR images into registration with 3D reconstructed post mortem volumes;
  4. Use a MRI-based 3D digital atlas as a template for fully-automated brain structure segmentation of in vivo and post mortem data;
  5. Configure dedicated databases by defining a reliable organization description (ontology).


BrainRAT toolbox is a BrainVISA add-on module package providing additional processes and functionalities relating to 3D reconstruction and analysis of post mortem rodent brain sections.

BrainVISA, in-house image processing software, is already available for free download on the Internet ( Since then, the core of BrainVISA is mainly developed in CEA/NeuroSpin by the LNAO lead by Jean-François Mangin, PhD (Yann Cointepas, PhD and D. Rivière, PhD) and by software engineers of the IFR 49 (Isabelle Denghien and Dominique Geffroy). The first prototype was developed by Yann Cointepas during years 2000/2001. BrainVISA also relies on Anatomist software visualization capacities.

Anatomist has been designed in the Service Hospitalier Frederic Joliot of CEA and is now mainly developed by Denis Rivière, PhD. The current version stems from ten years of work on several prototypes. One of its main original features is the generic module dedicated to structural data, namely sets of objects linked one another into a graph structure. It also provides a toolbox for regions of interest manual drawing.

For a complete and more detailed description of BrainVISA software, please refer to the handbook of BrainVISA.

General support

If you have any problems or queries relating to BrainRAT, BrainVISA or Anatomist, please contact us through the BrainVISA forum at:

Yet, before asking questions of reporting problems, please make sure your query is not solved in the FAQ. Thank you to consider that we have very few resources for technical support at the moment.

If you have any problems with your data, please check the procedure described in the How to scan section.

Demonstration data

For training and educational purposes, 2 datasets can be downloaded on demand at the following e-mail address: Please mention your institution, lab, research area, email address and operating system (Windows, MacOS or Linux).

Demonstration data enable to test all the BrainRAT functionalities through two applications:

  1. A unilateral stimulation of the rat visual system ([14C]-2DG autoradiography, Nissl staining -> ~150 adjacent sections)
  2. A whole rat brain anatomical dataset(Nissl staining -> ~200 adjacent sections)
Demonstration data will be made available electronically via anonymous FTP from the CEA FTP server as .zip or .tar.gz files according to our operating systems.

Remember that they are CEA property and are made available for educational purposes only. The use of these data for other purposes are submitted to CEA authorization. In case of use of BrainRAT for academic researches, please mention Dubois et al., 2007, J. Cereb. Blood Flow Metab.