Towards a Framework for Validating Machine Learning Results in Medical Imaging Opening the black box

In the medical imaging domain, non-linear warping has enabled pixel by pixel mapping of one image dataset to a reference dataset. This co-registration of data allows for robust, pixel-wise, statistical maps to be developed in the domain, leading to new insights regarding disease mechanisms [20]. Deep learning technologies have given way to some impressive discoveries. In some applications, deep learning algorithms have surpassed the abilities of human image readers to classify data. As long as endpoints are clearly defined, and the input data volume is large enough, deep learning networks can often converge and reach prediction, classification, and segmentation with success rates as high or higher than human operators [13]. However, machine learning, and deep learning algorithms are complex and interpretability is not always a straightforward byproduct of the classification performed. Visualization techniques have been developed to add a layer of interpretability. The work presented here compares a simplified machine learning workflow for medical imaging to a statistical map from a previous study to validate that the machine learning model used does indeed focus its attention on known important regions.