Beyond FA
We are pleased to announce the Beyond FA (BFA) challenge, a diffusion MRI grand challenge to extract meaningful features to better understand white matter integrity within Alzheimer's Disease (AD) cohorts.
Background: FA¶
Fractional anisotropy (FA) is a metric frequently used to interpret white matter integrity, but it exhibits high sensitivity and low specificity in pathological interpretation. These properties of FA make it difficult to use as a consistent biomarker as it makes different ages, cognitive statuses, and pathologies difficult to discern.
The Need to Consider Other Biomarkers¶
With the development of dozens of other white matter models and metrics, such as complex network measures, tractography bundle analysis, and NODDI metrics, there are thousands of combinations to analyze for white matter integrity analysis, especially in the context of lower-quality clinical data that FA’s voxel-wise computation can be sensitive to. The widespread collection of DW-MRI data necessitates an analysis of these metrics to interpret white matter integrity beyond what FA can capture.
Data¶
(a) For training, participants may use any dataset they choose, as long as it includes diffusion MRI. We will provide subjects from the MASIVar dataset for the Phase 1 test so that participants can make sure that their submissions align with our interfaces.
(b) Participants may use as many validating and testing cases as they choose, though we ask that they limit b-values to be a single shell of approximately 1000 (some datasets use 995 or 1005 -- these are acceptable). We will train and test our neural network based on hidden data. Before uploading the hidden data, we run PreQual. We encourage teams to perform the same preprocessing on their images this way.
(c) We will train based on 500 images (210 CN, 210 MCI, and 80 AD) and test based on 100 images -- 34 with normal cognition (CN), 33 with mild cognitive impairment (MCI), and 33 with AD. We do this to encourage teams to rise to the challenge of the scarcity of AD data among highly accessible CN data.
The Challenge Task¶
We ask participants to use the white matter model of their choice to extract meaningful features. We provide an evaluation Docker that trains and tests submissions in a shallow MLP.
Phases¶
Phase 1 includes images from the open MASiVar dataset for teams to debug their Dockers with our platform. Predictions from these Dockers are random -- they are ONLY a tool to make sure your Dockers are in the correct format. The outputs are random to deal with the small number of items in the archive. Teams will have access to debugging logs and allowed 5 attempts to upload Dockers. If you have multiple failures, please get in touch with the BFA team so that we can help you.
Phase 2 is the evaluation phase. Participants may upload their Docker once during this phase. Please see our testing MLP in the Additional Resources section.
Additional Resources¶
We provide a baseline algorithm to help with algorithm I/O. THIS HAS BEEN UPDATED 3/12! You can use the Docker to help build yours, too.
We also provide an example of our testing MLP.
See the timeline page for schedule updates.