Data-driven severity classification based on natural language extracted labels and computer vision-based cardiothoracic ratio calculation.

<p>A: Violin plot of the cardiothoracic ratio (CTR) based on the computer vision (CV) segmentation model. Each adjective/adverb relating to cardiomegaly was normalised (i.e. ‘mild<i>ly</i>’ to ‘mild’). Also showing reports without cardiomegaly mentions. Comparison of CV-CTR across...

Full description

Saved in:

Bibliographic Details
Main Author:	Tianhao Zhu (4166059) (author)
Other Authors:	Kexin Xu (757891) (author), Wonchan Son (21395383) (author), Kristofer Linton-Reid (15139762) (author), Marc Boubnovski-Martell (21395386) (author), Matt Grech-Sollars (21395389) (author), Antoine D. Lain (21395392) (author), Joram M. Posma (1696315) (author)
Published:	2025
Subjects:	Cell Biology Genetics Biotechnology Science Policy Space Science Environmental Sciences not elsewhere classified Biological Sciences not elsewhere classified Information Systems not elsewhere classified text radiology reports natural language processing missing metadata annotations large language model corrected several incorrect generated report ’ diagnostic imaging tests wrongly labelled views based report generation system identified 0 publicly available mimic div >< p certified radiologists assessed assisted diagnosis system 100 cxr images report generation assisted diagnosis diagnostic tool unlabelled views assigned views work found view labelling sex mentions severity classification relaxed agreement relaxed ). nlp ). human radiologists four board f </ dataset composed cxr dataset chest x cardiothoracic assessment cardiomegaly detection 827 free 81 ),
Tags:	Add Tag No Tags, Be the first to tag this record!

Data-driven severity classification based on natural language extracted labels and computer vision-based cardiothoracic ratio calculation.

Similar Items