Platforms

Jump to: navigation, search

Since 2016, the BFO team split into 2 new teams:

MUSTIC

Can be downloaded here

Ground Truth Agreement Toolbox (GTAT)

This Matlab toolbox accompanies the paper:

T. Lampert, A. Stumpf, and P. Gancarski, 'An Empirical Study into Annotator Agreement, Ground Truth Estimation, and Algorithm Evaluation'. (submitted).

It contains implementations of the functions described within the paper related to agreement analysis and the evaluation of detectors using different ground truth estimation techniques. It may also be used to recreate the figures for the fissure case study to gain a better understanding of the method (see the QUICK START section of the included readme file).

It is assumed that you have a number of annotations related to the same image.

To use the toolbox's functions, simply add the toolbox directory to Matlab's path. Within the header of each function may be found a short description of its purpose and in which section of the paper its mathematical derivation can be found. For more information see the toolbox's README file.

The toolbox is separated into three main functions:

  1. The agreement_analysis function calculates the statistics outlined in our paper for the collection of annotations passed to it.
  2. The calculate_GTs function calculates ground truths using the methods outlined below:
    • the LSML algorithm;
    • the agreement of any annotator;
    • the agreement of 50% of annotators;
    • the agreement of 75% of annotators;
    • the STAPLE algorithm;
    • by excluding outliers of the annotator clustering evaluation;
    • and by excluding outliers and then calculating the 50% agreement level.
  3. The detector_analysis function determines the detector's performance respective to each of the ground truths passed to the function, it then ranks the detectors based upon these performances.

The toolbox can be downloaded here.

Precision-Recall Toolbox (PRT)

This Matlab toolbox accompanies the paper:

T. Lampert and P. Gancarski, 'The Bane of Skew: Uncertain Ranks and Unrepresentative Precision'. Machine Learning, pp. (submitted), 2013.

It contains implementations of the functions described within the paper to calculate Precision-Recall, temporal Precision-Recall, integrated Precision-Recall and weighted Precision-Recall curves given a classifier's response and a corresponding ground truth. It may also be used to recreate the paper's figures to gain a better understanding of the method.

To use the toolbox's functions, simply add the toolbox directory to Matlab's path. Within the header of each function may be found a short description of its purpose and in which section of the paper its mathematical derivation can be found. For more information see the toolbox's README file.

The toolbox can be downloaded here.

BISTRO bioinformatics platform

BISTRO platform

The BISTRO bioinformatics platform has been recognized by the French Institute of Bioinformatics as a member of the national network of Bioinformatics platforms (Renabi)-East which includes Strasbourg, Lille, Vandoeuvre lès Nancy and Reims, in agreement with the IFB and Renabi policy to improve the national and international visibility of French bioinformatics achievements, by combining multidisciplinary and multi-Institute platforms. In this context, the BISTRO platform is developed by Strasbourg teams from the IGBMC, IBMC, IBMP, GMGM, IPHC, ICube and is able to provide coherent bioinformatics services including: expertise, tools and resources, data mining algorithms. These services are focused on scalable and functional analyses in various application domains, including biomedical studies, plants, yeast and bacteria.