Team parietal

Overall Objectives
Scientific Foundations
Application Domains
New Results
Other Grants and Activities

Section: Dissemination

Scientific Community animation

Invited conferences

Participants : Bertrand Thirion, Jean-Baptiste Poline.

Scientific animation

Participants : Bertrand Thirion, Jean-Baptiste Poline, Pierre Fillard, Gaël Varoquaux.

Conference organization

Participants : Gaël Varoquaux, Bertrand Thirion, Pierre Fillard.

Scipy 2009

Scipy 2009 conference: G. Varoquaux was the president of the scientific committee for the 8th international scipy conference (Caltech, August 2009) on scientific use of python language. Articles selected by a review committee are published on

fMRI workshop at MICCAI'2009

B. Thirion was the main organizer, with A. Roche, P.Ciuciu (CEA, Neurospin) and T.Nichols (GSK) of the fMRI workshop at MICCAI 2009, that took place in London, 24 September (30 participants, 9 papers):

Functional MRI (fMRI) provides a unique view on brain activity, which is used both for a better understanding of brain functional anatomy and the assessment of various mental diseases. The analysis of fMRI data entails detection issues, in which it has to be decided whether certain regions shows an activity significantly correlated to some variables of interest. This problem can be formulated in a given individual dataset (in which case the variable of interest is the experimental paradigm) or in a multi-subject dataset (the variable of interest is then a behavioral, clinical, or genetic factor of interest). Moreover, this problem can be handled as a modeling problem when addressing the temporal structure of the BOLD response and various fluctuations observed in fMRI datasets, or when delineating brain regions, especially across individuals, as well as a statistical problem: for instance, a typical concern is to warrant a certain control over false positives (specificity) for a testing procedure, or to achieve an optimal compromise between sensitivity and specificity by using judicious decision statistics.

While some of these questions may be familiar to the medical imaging community, partly for historical reasons, the neuroimaging community has developed specific contributions to solve these issues, and all the questions mentioned above are still the object of active research. This workshop was an opportunity to discuss and evaluate several solutions that have been proposed to solve these questions, and to confront different points of view.

Fibre cup at MICCAI'2009

The emergence of numerous models and fiber tracking techniques during the last decade raises the need for a comprehensive comparison of available methods on a common ground truth dataset. Objectives of the Fiber Cup were threefold: 1) to provide a MR phantom containing a plethora of crossing, kissing, splitting and bending fiber configurations, 2) to offer a set of quantitative criteria for algorithm performance evaluation with online results, and 3) to attract many participants by organizing it as a contest during the MICCAI conference in London. Participants were asked to run their algorithms on the phantom dataset and return their results along with a 2-page paper summarizing their method for quantitative evaluation.

6 datasets were made available: 3 of resolution 3x3x3mm (image size: 64x64x3) and 3 b-values (650, 1500 and 2000), and 3 of resolution 6x6x6mm (image size: 64x64x1) and 3 b-values (650, 1500, 2650). Participants could use any of these datasets for the contest, but only one submission was taken into consideration.

Contest rules:

Participants were free to use any combination of available tools, or algorithms, that could help them to reach the best result. It may consists in a combination of preprocessing steps and tractography algorithms. Only manual drawing of the fibers was not allowed.


The contest was a success and we received 10 submissions from international groups, with a result quality ranging from poor to very good results. The most striking result is undoubtedly the evidence of a large disparity of results (Fig. 8 ). This large inter-method variability is in itself an very interesting finding as it shows how tractography is method-dependent. The best 3 methods were awarded a diploma during the conference. The Fiber Cup allowed to collect a database of results on the same dataset and to create a common reference to evaluate new algorithms. Thus, we expect the phantom and the comparison methodology to be used thoroughly by groups to evaluate new algorithms. A web application to submit tractography results and obtain instantly the scoring is being investigated.

Figure 8. Left: ground truth. Right: all 10 submissions overlapped. 16 fibers were selected for comparison For each fiber on the left image, 10 fibers with the exact same color are displayed, each fiber corresponding to a submission. This illustrates the inter-method variability.


Logo Inria