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

TL;DR

This paper compares various deep learning models for lung segmentation in X-ray images, highlighting CE-Net as the top performer based on key accuracy metrics.

Contribution

It provides a comprehensive evaluation of existing deep learning methods for lung segmentation, including implementation and performance analysis.

Findings

CE-Net achieves the highest dice similarity coefficient.

TransResUNet and Lung VAE are also evaluated but perform less well.

Only nine out of 61 papers provided usable models or code.

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.