Learning under Distributed Weak Supervision
This addresses the problem of limited training data for medical image analysis researchers, though it is incremental as it adapts existing crowdsourcing and neural network methods to a specific domain.
The paper tackles the bottleneck of obtaining expert annotations for medical image segmentation by using crowdsourced weak super-pixel annotations from non-experts to train a fully convolutional neural network for fetal brain segmentation in MRI, reporting encouraging results compared to fully supervised methods.
The availability of training data for supervision is a frequently encountered bottleneck of medical image analysis methods. While typically established by a clinical expert rater, the increase in acquired imaging data renders traditional pixel-wise segmentations less feasible. In this paper, we examine the use of a crowdsourcing platform for the distribution of super-pixel weak annotation tasks and collect such annotations from a crowd of non-expert raters. The crowd annotations are subsequently used for training a fully convolutional neural network to address the problem of fetal brain segmentation in T2-weighted MR images. Using this approach we report encouraging results compared to highly targeted, fully supervised methods and potentially address a frequent problem impeding image analysis research.