Deep learning models for radiography body-part classification and chest radiograph projection/orientation classification: a multi-institutional study

1

Deep learning models for chest radiographs have advanced to estimate cardiac function, respiratory function, and biological age.
2

Quality control in labeling is crucial as errors in DICOM metadata can compromise deep learning model reliability.
3

Two deep learning models were developed: Xp-Bodypart-Checker for body part classification and CXp-Projection-Rotation-Checker for projection and rotation detection.
4

The study utilized a large-scale, multi-institutional dataset to enhance the generalizability of the models across diverse clinical settings.
5

Radiographs were classified and verified by board-certified radiologists to ensure accuracy in the model training process.

European Radiology

by Yasuhito Mitsuyama, Hirotaka Takita, Shannon L. Walston, Ko Watanabe, Shoya Ishimaru, Yukio Miki, Daiju Ueda
October 22, 2025

1