Position Classifier: Rethinking Position Encoding on Chest X-ray Diseases Identification

Yu Wen Fang, Fang-Yi Su, Jung-Hsien Chiang

Keywords: Position Classifier, Position Embedding, Patch-based, Chest X-ray

TL;DR: We proposed the Position Classifier to encode position information, improving the results of diseases identification.

Abstract: The patch-based method of chest X-ray interpretation often suffers from the loss of information regarding global anatomy structures. We propose using a simple position classifier to train the model to encode the correct position information. Our model can accurately encode the position information with a 99.87% AUC score on patch positioning. With the help of position information, our model can filter out anatomy structures that are commonly misinterpreted as lesions. We believe the proposed method is both effective and easy to implement in common deep learning-based diseases identification framework with only slight modifications.

Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: novel methodological ideas without extensive validation

Primary Subject Area: Detection and Diagnosis

Secondary Subject Area: Application: Radiology

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.