scjjb
commented
12 months ago @ljhOfGithub From looking at the slide it's just an issue of have a small amount of usable tissue. Smaller patches may be able to pick it up, but I've mostly found benefit in cases like this from lowering sthresh (can try somewhere between 1-5), lowering a_t (around 20), and increasing close (around 100). You just have to play around with parameters until you get what you want.
When I use command:python3 create_patches_fp.py --source xx --save_dir xx --patch_size 256 --seg --patch --stitch, I got the following result: {'seg_params': {'seg_level': -1, 'sthresh': 8, 'mthresh': 7, 'close': 4, 'use_otsu': False, 'keep_ids': 'none', 'exclude_ids': 'none'}, 'filter_params': {'a_t': 100, 'a_h': 16, 'max_n_holes': 8}, 'patch_params': {'use_padding': True, 'contour_fn': 'four_pt'}, 'vis_params': {'vis_level': -1, 'line_thickness': 250}}
progress: 0.00, 0/1 processing TCGA-A7-A0CD-01Z-00-DX2.609CED8D-5947-4753-A75B-73A8343B47EC.svs Creating patches for: TCGA-A7-A0CD-01Z-00-DX2.609CED8D-5947-4753-A75B-73A8343B47EC ... Total number of contours to process: 0 segmentation took 0.305422306060791 seconds patching took 6.532669067382812e-05 seconds stitching took -1 seconds average segmentation time in s per slide: 0.305422306060791 average patching time in s per slide: 6.532669067382812e-05 average stiching time in s per slide: -1.0 And I can't get the correct patches.How can it work by modifying the parameters? The .svs file can be gotten:https://portal.gdc.cancer.gov/files/1c4ec716-8a06-4bf9-b113-4bd0c8134c3f.