Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A comparative analysis of deep learning models for lung segmentation on X-ray images

Weronika Hryniewska-Guzik, Jakub Bilski, Bartosz Chrostowski, Jakub Drak Sbahi, Przemysław Biecek

TL;DR

This article describes a mobile, artificial intelligence-driven, robotic platform Rico, whose prior usage in similar scenarios, the number of its capabilities, and the experiments it presented should qualify it as a proper arm-less platform for social and assistive circumstances.

Abstract

Robust and highly accurate lung segmentation in X-rays is crucial in medical imaging. This study evaluates deep learning solutions for this task, ranking existing methods and analyzing their performance under diverse image modifications. Out of 61 analyzed papers, only nine offered implementation or pre-trained models, enabling assessment of three prominent methods: Lung VAE, TransResUNet, and CE-Net. The analysis revealed that CE-Net performs best, demonstrating the highest values in dice similarity coefficient and intersection over union metric.

A comparative analysis of deep learning models for lung segmentation on X-ray images

TL;DR

This article describes a mobile, artificial intelligence-driven, robotic platform Rico, whose prior usage in similar scenarios, the number of its capabilities, and the experiments it presented should qualify it as a proper arm-less platform for social and assistive circumstances.

Abstract

Robust and highly accurate lung segmentation in X-rays is crucial in medical imaging. This study evaluates deep learning solutions for this task, ranking existing methods and analyzing their performance under diverse image modifications. Out of 61 analyzed papers, only nine offered implementation or pre-trained models, enabling assessment of three prominent methods: Lung VAE, TransResUNet, and CE-Net. The analysis revealed that CE-Net performs best, demonstrating the highest values in dice similarity coefficient and intersection over union metric.
Paper Structure (5 sections, 3 figures)

This paper contains 5 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Methodology
  3. Related Works
  4. Results and discussion
  5. Conclusions

Figures (3)

  • Figure 1: Types of augmentation on which the ability to adapt to new conditions of segmentation models was tested.
  • Figure 2: Segmentation results (dice similarity coefficient and IoU value) after applying various augmentation methods that have not been performed on the training set before. On the horizontal axis is the dice similarity coefficient value from 0 to 1, and on the vertical axis is the number of samples from 0 to 500.
  • Figure 3: Original X-ray images with segmented lungs (marked in red) compared to lung masks generated by various models using different augmentation methods.