Skip to main content
. 2021 Dec 10;11:23823. doi: 10.1038/s41598-021-03204-z

Table 3.

Comparison of ARGO performance using OCR + NLP and NLP alone (internal series).

DATA-FIELD PRECISION (%) RECALL (%) F1-SCORE (%)
OCR + NLP NLP Diff OCR + NLP NLP Diff OCR + NLP NLP Diff
DIAGNOSIS 100.0 100.0 0.0 87.9 90.0 2.1 93.5 94.7 1.2
BIOPSY DATE 100.0 100.0 0.0 97.1 96.2 – 0.8 98.5 98.1 – 0.4
EXAM NUMBER 100.0 100.0 0.0 92.1 96.2 4.2 95.9 96.2 0.3
SPECIMEN TYPE 98.5 99.5 1.0 87.2 92.7 5.5 92.5 96.0 3.5
IHC EXECUTION 100.0 100.0 0.0 95.4 95.8 0.4 97.6 95.8 – 1.8
FISH EXECUTION 100.0 100.0 0.0 93.7 97.5 3.8 96.8 97.9 1.1
BM EXECUTION 100.0 100.0 0.0 92.9 97.1 4.2 96.3 98.7 2.4
KI-67 99.4 99.4 0.0 81.9 89.4 7.5 89.8 94.2 4.4

ARGO: Automatic Record Generator for Onco-hematology, OCR: Optical Character Recognition, NLP: Natural Language Processing, IHC: Immunohistochemistry, FISH: Fluorescent in situ Hybridization, BM: Bone Marrow, Diff: difference.