Images supporting 'Fully automated platelet differential interference contrast image analysis via deep learning'

This dataset supports the publication 'Fully automated platelet differential interference contrast image analysis via deep learning' submitted to the journal Scientific Reports. All data was gathered or generated at the University of Reading from 2020 to 2021. This dataset consists of i) 120 training images used to train a convolutional neural network (CNN) to automate platelet analyses, ii) 12 test images independent from the training images to test the CNN performance, and iii) a total of 225 8-bit images (representative of 3 independent experiments, with five fields of view captured per condition) of platelets spread over three different substrates (CRP-XL, fibrinogen, and vWF) and in the presence or absence of inhibitors (dasatinib, ibrutinib and PRT-060318) or an agonist (thrombin). These inhibitors and agonist are known to impact platelet morphology, and were used to assess the CNN�s performance on morphological extremes. The original 16-bit images with dimensions 2424x2424 were rescaled and converted to 970x970 8-bit images to reduce the file size using ImageJ. All images within this dataset are the 8-bit images.