Comprehensive clinical evaluation of novel 4DCT-based lung function imaging methods

Area of Science:

• Medical Imaging
• Radiology
• Computational Biology

Background:

• Traditional 4D CT (computed tomography) ventilation imaging uses Hounsfield Unit (HU) density changes, which have limitations in numerical robustness and do not provide perfusion data.
• Novel image processing techniques allow for the generation of 4D CT-ventilation/perfusion (V/P) lung imaging.

Purpose of the Study:

• To compare the clinical differences between classic HU-based 4D CT ventilation approaches and novel 4D CT V/P imaging methods.
• To evaluate the robustness and clinical applicability of new V/P imaging techniques.

Main Methods:

• Analyzed 4D CT data from 63 lung cancer patients.
• Generated four types of lung function images: 4D CT-ventilation using HU density (4DCT-vent-HU), Mass Conserving Volume Change (4DCT-vent-MCVC), Integrated Jacobian Formulation (4DCT-vent-IJF), and 4D CT-perfusion (4DCT-perf).
• Compared radiologist interpretations, functional contours (Dice Similarity Coefficients), and dose-function metrics between the novel methods and 4DCT-vent-HU.

Main Results:

• Receiver Operating Characteristic (ROC) analysis showed accuracies ranging from 0.55 to 0.73 when comparing novel methods to 4DCT-vent-HU.
• Average Dice Similarity Coefficients (DSC) ranged from 0.41 to 0.44, indicating moderate agreement between methods.
• All novel methods demonstrated significant differences (p < 0.01) in dose-function metrics compared to 4DCT-vent-HU, with 4DCT-vent-MCVC and 4DCT-Perf showing the smallest and largest differences in functional mean lung dose (fMLD), respectively.

Conclusions:

• This study provides the first comprehensive comparison of novel 4D CT V/P imaging methods against classical formulations.
• Significant differences exist between the various 4D CT-based functional imaging methods.
• Further research is necessary to determine which 4D CT functional imaging methods offer the most reliable clinical results.