Accuracy of artificial intelligence-based simulation for assessing lung vessels and volume using unenhanced computed...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Thoracic Surgery

Background:

• Preoperative three-dimensional (3D) image simulations using enhanced computed tomography (ECT) are valuable for anatomical lung resections.
• Contrast agents required for ECT pose a barrier for certain patients.

Purpose of the Study:

• To evaluate the accuracy of an artificial intelligence-based 3D simulation using unenhanced computed tomography (UECT) data.
• To compare UECT-based simulations with ECT-based simulations for anatomical lung resections.

Main Methods:

• 18 lung cancer patients undergoing anatomical lung resections were included.
• Artificial intelligence software was used to create 3D bronchovascular tree images from both UECT and ECT data.
• Pulmonary vessel identification accuracy and lung segment volumes were compared between UECT and ECT.

Main Results:

• UECT identified 96.6% of artery branches and 82.1% of vein branches, closely matching ECT's 98.9% and 85.7%.
• Correlation coefficients for branch detection (0.9783) and artery-oriented segment volumes (R=0.8330 right, R=0.8082 left) were significant.
• UECT showed high concordance with ECT in visualizing bronchovascular structures.

Conclusions:

• Artificial intelligence-based 3D simulations using UECT are comparable to ECT.
• This UECT technique offers a promising alternative for preoperative planning in anatomical lung resections, avoiding contrast agents.