A bootstrapping residuals approach to determine the error in quantitative functional lung imaging

Abstract

Purpose: To implement and validate an algorithm to determine the statistical errors in self-gated non-contrast-enhanced functional lung imaging. Methods: A bootstrapping residuals approach to determine the error in quantitative functional lung imaging is proposed. Precision and accuracy of the median error over the lungs, as well as reproducibility of the approach were investigated in 7 volunteers. The algorithm was additionally applied to data acquired in a patient with cystic fibrosis. Results: The obtained bootstrapping error maps appear comparable to the error maps determined from repeated measurements, and median absolute error values for both methods show comparable median errors when reducing the number of averages. In a volunteer in whom 10 consecutive measurements were carried out, the median functional parameters were ventilation = 0.22 mL gas/mL lung tissue, perfusion amplitude = 0.028, perfusion timing = -82 ms, whereas precision and accuracy of the median error were below 3.2 × 10-3 mL gas/mL lung for ventilation tissue, 4.4 × 10−4 for perfusion amplitude, and 11 ms for perfusion timing. In the measurement of the patient, low errors in areas with reduced ventilation support the assessment as real defects. Conclusion: Using a bootstrapping residuals method, the error of functional lung MRI could be determined without the need for repeated measurements. The error values can be determined reproducibly and can be used as a future means of quality control for functional lung MRI.