Landmark Detection using Transformer Toward Robot-assisted Nasal Airway Intubation

TL;DR

This paper introduces a transformer-based method for accurate landmark detection in robot-assisted nasal intubation, improving localization of nostrils and glottis with a deformable DeTR and semantic alignment.

Contribution

It proposes a novel transformer-based landmark detection approach with semantic alignment, tailored for small object detection in medical robotic procedures.

Findings

Achieved high detection accuracy on a public glottis dataset.

Successfully annotated and detected nostrils using the proposed method.

Code is publicly available for reproducibility.

Abstract

Robot-assisted airway intubation application needs high accuracy in locating targets and organs. Two vital landmarks, nostrils and glottis, can be detected during the intubation to accommodate the stages of nasal intubation. Automated landmark detection can provide accurate localization and quantitative evaluation. The Detection Transformer (DeTR) leads object detectors to a new paradigm with long-range dependence. However, current DeTR requires long iterations to converge, and does not perform well in detecting small objects. This paper proposes a transformer-based landmark detection solution with deformable DeTR and the semantic-aligned-matching module for detecting landmarks in robot-assisted intubation. The semantics aligner can effectively align the semantics of object queries and image features in the same embedding space using the most discriminative features. To evaluate the performance of our solution, we utilize a publicly accessible glottis dataset and automatically annotate a nostril detection dataset. The experimental results demonstrate our competitive performance in detection accuracy. Our code is publicly accessible.