Deep texture representation analysis of histopathological images from TCGA whole slide images: intermediate data

This is a generated intermediate data and expected result of deep texture representation (DTR) analysis from 79,110 image patches of hematoxylin eosin-stained histological samples of various human cancers. The dataset consists of image patches sized 256 x 256 pixel (1.0 μm/pixel) from 5 solid cancer types were downloaded from Daisuke Komura, & Shumpei Ishikawa. (2021). Histology images from uniform tumor regions in TCGA Whole Slide Images [Data set]. in August 2022 and stored in Patches.zip. Filename: Patches/[cancer_type]/5/[TCGA Barcode]/[region]-[number]-[pixel resolution in original WSI image].jpg [TCGA Barcode] TCGA-XX-XXXX represents patient ID. (Please refer to https://docs.gdc.cancer.gov/Encyclopedia/pages/TCGA_Barcode/ for detail.) DTRs of these images were calculated by convolutional neural network (VGG16) and stored in TCGA_dtr_imglist_df.pkl. Supervised learning model of tumor classification based on DTR were created with Support Vector Machine, the result were stored in TCGA_svm_model.pkl An example of whole slide image (WSI) in NDPI format and annotated regions of interest in NDPA format suitable to be included in DTR analysis are included in Example_WSI.zip. For viewing in QuPath, please select the Example_WSI.qpdata