search

Jiangbo-Shi / SGMF

A Structure-aware Hierarchical Graph-based Multiple Instance Learning Framework for pT Staging in Histopathological Image (TMI 2023)
7 stars 1 forks source link

Data preprocessing #3

Open syy-create opened 1 week ago

syy-create commented 1 week ago

hi， @Jiangbo-Shi May I ask for a brief explanation of the data preprocessing method for tissue graph and patch graph? I'm not sure what the correct processing method is. Best wishes!

Jiangbo-Shi commented 1 week ago

Patch graph is built based on Patch-GCN. The node of the tissue graph is segmented using the SLIC method (skimage.segmentation.slic package) on the resized image. The edge is constructed using the nearest neighbor method. To find the corresponding spatial relations between patches at different magnifications, you can rescale the coordinate into the same magnification.

syy-create commented 1 week ago

Hi, @Jiangbo-Shi Thank you for your reply. Excuse me again, I would like to confirm whether the patch features were extracted using ResNet50 or other methods in the feature extraction section. Can you also provide example feature files for these patches and tissues in the script file. --data_folder_s gastric_pt_staging_20x \ --data_folder_l gastric_pt_staging_10x \ --tg_file tissue_graph_files_slic \ Best wishes!