 |
Page History
...
Challenges are often conducted in conjunction with scientific conferences. The following Pilot Challenges, supported by the CTIIP project and described in the following table, were part of the Computational Brain Tumor Cluster of Event (CBTC) 2015 which was held on October 9, 2015 in Munich, Germany, in conjunction with MICCAI 2015.
http://miccai.cloudapp.net:8000/competitions/28
| MICCAI 2015 Challenges | Description | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Combined Radiology and Pathology Classification | The datasets for this challenge are Radiology and Pathology images obtained from the same patients. Each case corresponds to a single patient. There is one Radiology image and one whole slide tissue image for each case. The training set contains a total of 32 cases: 16 cases that are classified by pathologists as Oligodendroglioma and 16 cases classified as Astrocytoma. The test set will have 20 cases. Please note that the number of cases in the test set may not be equally partitioned between the two sub-types. The whole slide tissue images are stored in Aperio SVS format. There are open source tools and libraries that can read these images: OpenSlide
| ||||||||||||||||
| Segmentation of Nuclei in Pathology Images | The characteristics of cancer nuclei are central components in many aspects of pathology classification. Nuclear features, when combined with “omics,” have been shown by many research groups to be linked to patient outcome. Although there are many important and predictive features, characteristics of cancer nuclei are the most important. Virtually without exception, all commercial systems and academic groups in the digital Pathology area make use of nuclear segmentation algorithms. The ability to segment and then classify nuclei is therefore a key task. The goal of this challenge is to evaluate the performance of algorithms for detection and segmentation of nuclear material in a tissue image. Participants are asked to detect and segment all the nuclei in a given set of image tiles extracted from whole slide tissue images. The algorithm results will be compared with consensus pathologist segmented sub regions. Winners will be ranked based on their nuclei segmentation best matching the reference standards. The reference standard will be pathologist generated nuclear segmentation on select regions of TCGA Glioma whole slide images for the challenge. |
There are several platforms to host these challenges. Jayashree Kalpathy-Cramer’s team is setting up Medici for our CTIIP pilot challenges (and hopefully others after that).
A team A team from Massachusetts General Hospital will guide the pilot challenges, using the Pilot Challenges. They will use Medical Imaging Challenge Infrastructure (MedICI) , a system that supports medical imaging to support the challenges.The MedICI system will utilize MedICI, in turn, uses the CodaLab framework, a newly developed, an open-source challenge platform developed by Microsoft Research and others in the medical imaging and machine learning communities. Because CodaLab does not have built-in imaging handling, display or annotation capabilities, we the team will build on two application packages, ePad and caMicroscope, to provide those features.
MedICI
Jaysharee's program: Medical Imaging Challenge Infrastructure: MedICI
- Based on open-source CodaLab
- ePAD (created by Daniel Rubin's group at Stanford): tool for annotating images, creates AIM images
- caMicroscope
http://miccai.cloudapp.net:8000/competitions/28
- Competition #1: MICCAI challenge has a training phase where they train their algorithms. A test phase where they run their algorithms on images they have never seen before. They are compared to the ground truth that is determined beforehand. caMicroscope is used to see what is there before and to visualize the results. Overlap/completeness match determines the winner.
- Competition #2: They are given slides.
From PPT: Use titles of slides
Setting up a competition by an organizer. Organizer creates competition bundle.
Can go to cancerimagingarchive.net and create shared lists. Shared lists are pulled into CodaLab. That is how they get the test and training data.
Next is to create ground truth.
Regions of interest in a tumor for annotations are necrosis, adema, and active cancer. Radiologists create the ground truth.
Once participants upload their results, they can see them in ePad.
Notes
Medical Image Computational and computer-assisted Intervention: MICCAI
Ground truth: find the compatibility of the informatics that we need to run pilots. Take images out of TCIA, TCGA, clinical data and compare them.
Jasharee doing MICCAI Challenge in Munich. Segmentation of nuclear imaging in pathology. Combined radiology and pathology classification.
Want to be able to say that these informatics allow us to compare the pathology, rad, co-clinical findings.
Document the approach, technology, application to do a MICCAI challenge the way Jaysharee does itChallenge participants receive test and training data by creating shared lists in TCIA, then pulling those into CodaLab. Once participants upload their results, they can see them in ePad.
Scenarios
Need to generate proper therapy for a patient. Look at in vivo imaging, radiology and pathology, run a gene panel to look for abnormal. Look at co-clinical trials (model of a tumor in a mouse that is similar to a human. Experiment therapies on mice.) Run an integrative query to develop a sophisticated diagnosis. Search big data.
...
