Replication Data for: Approximating the Hotelling Observer with Autoencoder-Learned Efficient Channels for Binary Signal Detection Tasks

Dataset Summary This is the companion dataset and models for the manuscript by Jason Granstedt, Weimin Zhou, and Mark Anastasio titled "Approximating the Hotelling observer with autoencoder-learned efficient channels for binary signal detection tasks." Dataset Structure Files for each of the primary experiments are contained in their own folder. Within each category, there are folders that contain the data, models, and scripts. The data folder contains the data used in the experiments. There are four datasets, two of which employ a lumpy background and two of which employ a numerical breast phantom background. The datasets with a numerical breast phantom background employ the data published by the VICTRE study, located at https://github.com/DIDSR/VICTRE_DM_ROIs/. The models folder contains the trained autoencoder (AE) models used in the paper. Log files for the training process are also included. The scripts folder contains the scripts employed to generate the statistics in the paper. Both MATLAB scripts for generating statistics and the raw data employed to fit the AUC values are included. The included code contains both the python training scripts and shell scripts for sweeping the relevant parameters of the models. The scripts folders contain a subfolder named "VICTRE", which contains the code for the filtered channel observer. This code was taken from https://github.com/DIDSR/VICTRE_MO. Further documentation regarding the scripts can be found at https://github.com/jasonlg/AETSI.