Related Experiment Video
Updated: Nov 1, 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
14.4K
Automatic lung segmentation in COVID-19 patients: Impact on quantitative computed tomography analysis
L Berta1, F Rizzetto2, C De Mattia1
1Department of Medical Physics, ASST Grande Ospedale Metropolitano Niguarda, Piazza Ospedale Maggiore 3, 20162 Milan, Italy.
Summary
Automatic lung segmentation for CT imaging shows variable accuracy. While CNNs performed best, no method fully replaced manual correction, impacting quantitative analysis metrics. Individual metric evaluation is crucial.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence in Medicine
Background:
- Accurate lung segmentation is vital for quantitative analysis of CT scans.
- Automatic segmentation methods aim to improve efficiency but require validation.
Purpose of the Study:
- To evaluate the impact of lung segmentation accuracy in automatic pipelines for quantitative CT analysis.
- To compare different automatic lung segmentation algorithms.
Main Methods:
- Four automatic lung segmentation platforms (CNN, region-growing, atlas-based) were tested on CT images from 55 severe COVID-19 patients.
- Radiologists provided qualitative scores (QS), and manual reference segmentations (RS) were created.
- Quantitative metrics (QM) and Dice Index (DI) were calculated, comparing automatic segmentations to RS.
Main Results:
- Convolutional Neural Network (CNN) algorithms showed higher qualitative scores and lower quantitative metric differences (ΔQMs).
- However, only 45% of CNN segmentations required minimal correction, and 31% were accurate enough for reference segmentation (RS) without manual input.
- Significant differences in QMs were observed between automatic segmentations and RS, with the histogram 90th percentile being the most unstable metric.
Conclusions:
- No tested automatic lung segmentation algorithm provided fully reliable results for quantitative CT analysis.
- Segmentation accuracy significantly impacts various quantitative metrics.
- Individual evaluation of each quantitative metric is necessary based on the specific analysis goals.
