Data repository for the upcoming publication "Deep learning enables pathologist-like scoring of NASH models". Corresponding github: https://github.com/FabianHeinemann/Deep_learning_for_liver_NAS_and_fibrosis_scoring Content (uploaded after acceptance): - **Image tiles for training / validation** of 4 CNN models to classify: - Ballooning (higher resolution tiles) - Steatosis (higher resolution tiles) - Inflammation (higher resolution tiles) - Fibrosis (lower resolution tiles) - **Image tiles to calibrate thresholds** to map continuous scores to discrete scores according to the NAS (higher resolution tiles) and fibrosis score (lower resolution tiles). The zip archives containing these images were too large to be stored at osf (~300 GB). They can be obtained upon reasonable request from: fabian.heinemann@boehringer-ingelheim.com. - **2 x tables with pathologist ground truth NAS and fibrosis scores** (train and test). - **4 x Thresholds as json for mapping of continuous scores to pathologist scores** for the four features. - **4 x Weights of trained** CNN models.