Next Generation Hosting

for Scientific Algorithm Challenges

in Medical Imaging

 
Today the scientific community puts several efforts in order to improve the quality of published scientific results, so named reproducible research. All scientific data (from raw to derived data) associated to a published paper should be easily accessible for a long time. In the best case this just means adding a link into the paper to corresponding data, and being sure that every reader can access and reproduce the described results afterwards.

To achieve this, in the best case authors store the original data, the derived data, as well as the computational software used to produce the results. This would keep the hole loop safely stored and available for later investigations.
 
Even one generation further, one could combine different software to achieve better consensus results to solve scientific problems. 
 

What we do?

We help scientific challenge organizers and their challengers by providing a platform to organize scientific algorithm challenges within the field of medical imaging. Our FLI-IAM platform stores the original data that researchers can share, encapsulates the challenger pipelines on using docker images, computes and keeps the results. A detailled evaluation of all the pipelines is provided afterwards. All valuable information is retrievable and kept in one platform. As the data remain unchanged after the challenge day, same for the docker images, all the scientific work is stored safely in our system, and can be accessible easily and for long term afterwards with specific rights policy.
 

With what we can help you?

  • We have experience in organizing challenges in medical imaging and share it with you.
  • We support you to prepare your data, to have them ready for a challenge.
    • Converting your data from DICOM to NIfTI
  • Our portal website allows to describe your challenge in details and offers a subscription process.
  • Our portal is a perfect means to communicate with your challengers,
    to import your data and share them  (DICOM and/or NIfTI) to produce the challenger pipeline results.
  • We import using docker all your challenger pipelines/computational methods and execute them on our platform.
  • We evaluate the produced results with multiple metrics
    • Dice, Jaccard, and many more
  • We also evaluate additionnal performance parameters, like CPU, RAM usage etc.
  • We have a results dashboard to be used during the workshop day.
 

What you have to do as challenge organizer?

  • Take care of all scientific aspects, the work you like
  • Bring your data with you or we can support you for that

What you have to do as challenger?

  • Take care of all scientific aspects, the work you like
  • Provide us your pipeline/computational method in a docker and we will integrate it

Why we organized two scientific challenges for MICCAI 2016?

The aim of the “MICCAI 2016” challenges organized by FLI-IAM was to demonstrate the relevance of our plateform as a service (PAAS) in real life and to exhibit how our IT einfrastructure allows the evaluation of segmentation algorithm on medical images. These challenges assessed the state of the art of segmentation methods provided by the participants and running on data on our infrastructure. The databases were provided by communities of users together with FLI-IAM:
 
  • OFSEP[1] for the competition in Multiple Sclerosis and the
  • American Association of Physicists in Medicine (AAPM and its working group 211)[2] for the PET competition.

The segmentation solutions were uploaded by the participants to our infrastructure and automatically executed on the databases.
The results were recorded and evaluation metrics computed.

In summary, our e-Infra provides several services:
  • Hosting imaging data
  • Hosting and executing processing tools
  • Computing relevant metrics for solution comparison
  • Dissemination of the results

A web portal was implemented  to manage different aspects such as user authentication and navigation in infrastructure services.
 
Links, see menu above:
  • MS Segmentation
  • PET Segmentation
[1] OFSEP : Observatoire Français pour la Sclérose en Plaques (http://www.ofsep.org)
[2] ]http://aapm.org/org/structure/default.asp?committee_code=TG211
 

Want to see more details?

Please see the video below.
 

Want to know more?

Please write an email to michael.kain@inria.fr
 

Challenges Miccai 2016

Dear Miccai 2016 challengers, welcome on our home page!

Please find your required informations with the next links:

Challenges results presentation

Services

FLI-IAM, the software platform node of France Life Imaging (FLI), called Information Analysis and Management (IAM), provides services for in-vivo medical imaging.

Software as a Service (SaaS)

for clinical and preclinical research

  • Subcontracting (turnkey service)
  • Resource service for clinical and preclinical research

Platform as a Service (PaaS)

"open resource" for professionals

  • Storage of data (DaaS)
  • Processing pipelines defined for the needs of the large (open) community of medical imaging research

Storage and Computing as a Service (SCaaS)

Outsourced resources sharing

  • for imaging data & processing (with backup) for human or animal in-vivo imaging

Data Storage

  • Big data storage of DICOM or NIfTI files
  • Any other file formats for clinical and preclinical research
  • Access to data from everywhere and on all devices

Computation

  • Large computations within grid infrastructure
  • Integrations of developed algorithms within infrastructure
  • Easy computation results access

Small Animals

  • Web-oriented management system of preclinical brain imaging data
  • Catalog of software and processing systems dedicated to data processing of small animals imaging