Workflow-centred open-source fully automated lung volumetry in chest CT

Area of Science:

• Medical Imaging
• Radiology
• Pulmonology

Background:

• Accurate measurement of total lung capacity (TLC) is crucial for diagnosing and managing respiratory diseases.
• Current methods for TLC extraction from computed tomography (CT) images often require manual intervention, limiting efficiency.
• There is a need for automated, robust, and open-source solutions for TLC quantification in clinical practice.

Purpose of the Study:

• To develop a fully automated, open-source algorithm for extracting total lung capacity (TLC) from CT images.
• To validate the accuracy of the automated TLC measurements against pulmonary function testing (PFT).
• To demonstrate the integration of the automated TLC extraction method into a clinical Picture Archiving and Communication System (PACS) workflow.

Main Methods:

• An open-source region-growing algorithm was developed for automated lung segmentation and TLC calculation from CT data.
• The algorithm's performance was evaluated on 288 retrospective patient CT scans.
• Validation involved correlating CT-derived TLC (TLC_CT) with PFT-derived TLC (TLC_PFT) in a subgroup of patients.

Main Results:

• The automated algorithm achieved excellent segmentation in over 95% of cases (13/288 patients showed poor results).
• A strong positive correlation was found between TLC_CT and TLC_PFT (r=0.87, p<0.001).
• Measurements demonstrated high agreement between different CT reconstruction kernels (ICC=0.99), with rapid calculation times (average 5 seconds).

Conclusions:

• A robust, open-source, and fully automated method for TLC extraction from CT images has been successfully developed.
• The algorithm provides accurate lung volume quantification with high success rates and seamless integration into clinical PACS environments.
• This automated approach enhances diagnostic efficiency and supports radiologists by providing rapid, reliable TLC measurements.