Related Experiment Video
Updated: Jun 13, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.1K
Automatic Quantification of Abnormal Lung Parenchymal Attenuation on Chest Computed Tomography Images Using
Alysson R S Carvalho1,2,3,4, Alan Guimarães3, Rodrigo Basilio2
1Department of Radiology and Imaging Diagnosis, Hospital Universitário Polydoro Ernani de São Thiago, Universidade Federal de Santa Catarina, Florianópolis.
Journal of Thoracic Imaging
|September 11, 2024
Summary
Convolutional neural networks (CNNs) show strong correlation with lung densitometry for detecting chest CT abnormalities. CNNs offer improved characterization of lung areas, especially normal ones, but require significant computing power.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Medicine
- Pulmonary Disease Diagnosis
Background:
- Chest computed tomography (CT) is crucial for diagnosing lung abnormalities.
- Texture-based analysis using convolutional neural networks (CNNs) offers a novel approach to image interpretation.
- Lung densitometry is a traditional method for quantifying lung attenuation.
Purpose of the Study:
- To compare the efficacy of CNN-based texture analysis against traditional lung densitometry in detecting chest CT abnormalities.
- To evaluate the ability of both methods to characterize different lung tissue types (low, normal, and high attenuation areas).
Main Methods:
- A U-NET model was employed for lung segmentation.
- An ensemble of seven CNN architectures was trained to classify low-attenuation areas (LAAs), normal-attenuation areas (NAAs), and high-attenuation areas (HAAs).
- CNN and densitometry severity indices were calculated and compared across 812 CT scans from normal subjects and patients with emphysema or interstitial lung disease (ILD).
Main Results:
- CNN-derived and densitometry-derived severity indices demonstrated a strong correlation (ρ=0.90) and increased with disease severity.
- CNN severity indices were lower for emphysema but higher for moderate to severe ILD cases compared to densitometry.
- CNNs provided higher estimations for normal attenuation areas, suggesting potentially more accurate characterization.
Conclusions:
- CNN-based texture analysis closely aligns with lung densitometry in assessing lung abnormalities on CT scans.
- CNNs offer improved estimation of normal lung areas and better differentiation of similar abnormalities.
- The implementation of CNNs for this application necessitates substantial computing resources.
