BriefGPT.xyz
Feb, 2021
VisualCheXbert:解决放射学报告标签和图像标签之间的差异
VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels
HTML
PDF
Saahil Jain, Akshay Smit, Steven QH Truong, Chanh DT Nguyen, Minh-Thanh Huynh...
TL;DR
使用 BERT 模型直接从放射学报告映射到图像标签的方法,优于现有的医学报告标签方法,与放射科医生标注的胸部 X 光图像的平均 F1 分数为0.12至0.21。
Abstract
Automatic extraction of
medical conditions
from free-text
radiology reports
is critical for supervising
computer vision
models to interpre
→