The leading cancer mortality globally is Lung Cancer. A key objective for increasing lung cancer survival is discovering the illness early, allowing for the most effective treatment choices. Lung cancer develops from lesions in the bronchial epithelium of the lung mucosa. These bronchial lesions can progress to squamous cell lung cancer and assist in forecasting other lung cancers’ development. As a result, approaches for early diagnosis of bronchial lesions are critical for improving lung cancer patient treatment. Using bronchoscopy to image the airway epithelium during a regular airway exam is a noninvasive technique for clinicians to look for such lesions.

Autofluorescence bronchoscopy is one of the most sensitive advanced bronchoscopic video procedures available today. It can efficiently distinguish growing bronchial lesions from the normal epithelium. Unfortunately, the current standard requires human inspection of an incoming AFB video stream, which is time-consuming and error-prone. While some research has looked toward computer-based lesion analysis approaches for AFB video frames, all of these studies have one or more limitations as follows: 

Researchers believe this is the first time someone has used AFB video for automated real-time segmentation of bronchial lesions. Furthermore, their proposed efficient stage-wise feature pyramid (ESFP) encoder on Mixtransformer (MiT) with SOTA performances on public datasets demonstrates a significant capacity for medical picture segmentation.

Their architecture is depicted in the figure below. It employs the Mix Transformer (MiT) encoder as the backbone and an efficient stage-wise feature pyramid (ESFP) decoder to create segmentation outputs. 

GitHub has the official implementation of this paper.

Content Writing Consultant Intern at Marktechpost.

Free-2 Min AI Newsletter Join Our AI Community EmailEnter your email address Subscribe

Marktechpost is a California based AI News Platform providing easy-to-consume, byte size updates in machine learning, deep learning, and data science research

© 2021 Marktechpost LLC. All Rights Reserved. Made with ❤️ in California

Learn the AI best practices from 120+ experts at the TransformX Conference