reader study pilot to validate model observer performance

[bnel1201]

Goal is perform a reader study at max does with a few different recon options (fbp sharp, fbp smooth, REDCNN sharp, ...) to calibrate where our human performance is relative to this model observers

Available reader study tools to consider

1. Miguel has an online tool that could be useful: simplephy, here's it's gitlab wiki, paper
2. Steve has a 2AFC sample we could also use

By next meeting I will give a small demo with our signal present, signal absent images