| case_barcode,patient_filename,has_report_row,why_excluded,evaluation_role,pdf_pages,page_images,pdf_text_chars,pdf_text_words,embedded,note | |
| TCGA-EC-A24G,TCGA-EC-A24G.5DE142C5-7A0D-4C5E-8EF6-2809B10378C4,true,text_len_2_below_threshold,visual_only,1,1,436,17,false,"The one genuine visual-rescue case. Its single page is a fully completed UTERUS TISSUE CHECKLIST printed sideways, so the scan-time OCR captured only the horizontal handwritten annotations and the barcode. The corpus row is the two characters '1.'. The checklist is entirely legible to a human: adenocarcinoma, poorly differentiated, invasion of cervix, 0/15 nodes, venous invasion absent, margins uninvolved, extent of myometrial invasion not specified. Page is rotated roughly 90 degrees and is rendered as-is; see #32." | |
| TCGA-AJ-A3BH,TCGA-AJ-A3BH,false,no_report_row_in_corpus,corpus_missing_text_recoverable,5,5,8229,747,false,"Not a visual-only case. The corpus has no row for it, but the GDC PDF carries a clean five-page text layer of ordinary pathology prose. A text extractor reading the PDF directly recovers it in full." | |
| TCGA-BK-A139,TCGA-BK-A139,false,no_report_row_in_corpus,corpus_missing_text_recoverable,2,2,4580,412,false,"Not a visual-only case, and possibly not a checklist case at all: the specimen is a pelvic exenteration for recurrent disease, not the hysterectomy the CAP protocol applies to. The GDC PDF carries a clean two-page text layer." | |