The VAMPIRE challenge: A multi-institutional validation study of CT ventilation imaging

Area of Science:

• Medical Imaging
• Radiology
• Pulmonary Imaging

Background:

• CT ventilation imaging (CTVI) is crucial for lung cancer radiation therapy.
• Standardized validation tools are needed for CTVI algorithms.

Purpose of the Study:

• Introduce the VAMPIRE Dataset and Challenge for CTVI validation.
• Compare relative ventilation distributions from different CTVI algorithms and established modalities.

Main Methods:

• Developed the VAMPIRE Dataset with 50 4DCT and reference ventilation image pairs.
• Conducted the VAMPIRE Challenge with 7 research groups using deformable image registration (DIR) methods.
• Evaluated CTVI algorithms against reference ventilation images using Spearman correlation and Dice similarity coefficients.

Main Results:

• High inter-algorithm and inter-subject variability observed in CTVI performance.
• Algorithm rankings shifted based on the chosen evaluation metric.
• The top-performing algorithm used biomechanical modeling with a hybrid metric, achieving moderate correlations.

Conclusions:

• CTVI performance varies significantly with algorithm, reference modality, subject, and metric.
• Lack of a gold standard and observed variability necessitate further validation for clinical translation.
• The VAMPIRE Challenge provides insights to guide future CTVI validation efforts.