Unencrypted login | home

Program Information

Study of Deformable Registration Based 4DCT Ventilation Imaging Methods

J McClelland

J McClelland1*, Y Suh2, S Ahmad2, D Hawkes1, (1) Centre for Medical Image Computing, University College London, (2) Guy's and St. Thomas' NHS Trust

TH-E-218-4 Thursday 1:00:00 PM - 2:50:00 PM Room: 218

Purpose: In recent years there has been considerable interest in methods of imaging the local ventilation using deformable registration of 4DCT images. This work investigates three different methods for calculating these ventilation images, and examines the effect that the registrations and other parameters can have on the resulting images.

Methods: The end-inhalation and end-exhalation 4DCT images were aligned using B-spline deformable image registration. The registrations were run several times, with different constraint term weights.
Ventilation images were generated using three methods based on those that have appeared in the literature. One of these measures changes in the density of the lung parenchyma from the HU values in the aligned CT images. One measures local volume change from the Jacobian determinant of the registration results. The third combines these methods to potentially account for changes in the amount of blood in the lungs (i.e. change in mass) over the breathing cycle.
The results from the three methods were compared to each other when using different registration results and amounts of smoothing applied to the CT images.

Results: Ventilation images were calculated for 10 patient datasets. All registrations used produced visually good results and had low TREs (mean TRE 1.0, 1.0, and 1.2 mm).
There were large discrepancies between the different ventilation measures (more so than would be expected due to changes in mass). The registration used and the amount of smoothing could dramatically affect the ventilation results. Artefacts in the 4DCT images could also have a large impact on the ventilation results.

Conclusions: The ventilation measures are very sensitive to the registration result and the amount of smoothing used, and there is marked disagreement between the different measures. More work is required before these or similar ventilation imaging methods can be considered reliable, robust, and accurate.

Contact Email