Keywords: vision transformers, Swin transformer, medical object detection
Abstract: Current state-of-the-art detection algorithms operating on 2D natural images utilize the relation modeling capability of vision transformers to increase detection performance. However, the feasibility of adapting vision transformers for the 3D medical object detection task remains largely unexplored. To this end, we attempt to leverage vision transformers for organs-at-risk detection and propose a novel feature extraction backbone, dubbed SwinFPN, which exploits the concept of shifted window-based self-attention. We combine SwinFPN with Retina U-Net's head networks and report superior detection performances. Code for SwinFPN will be available in our medical vision transformer library https://github.com/bwittmann/transoar.
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.