The use of image correlation plots in lung ventilation and perfusion scanning

Experimental modelling was conducted by designing and constructing a phantom consisting of two plastic containers to represent two lungs, with a number of defects resembling emboli. The dimensions and the volume of this phantom were selected by reviewing chest data of patients. Each container was injected with ~100MBq of 99mTc and immediately scanned using a Nuclear Medicine Gamma Camera. Three static views of each phantom were investigated: anterior, lateral right oblique and lateral left oblique. 600,000 counts were used for perfusion and 300,000 counts used for ventilation. The scanned images were registered using a HERMES workstation, and then transferred to Image J software to obtain a joint histogram plot, to assess the spatial relationship of the two datasets.

A joint histogram derived from ventilation and perfusion images was sensitive to the presence of differences between ventilation and perfusion scanning images. A correlation coefficient parameter has been identified which could potentially be an aid to the detection of pulmonary embolism in a mathematical way to support traditional visual inspection of images.

A Monte Carlo model was developed to extend experimental testing by modelling. Computer simulation of the experiment allows variations of defect size and location to be investigated without additional expense and time on equipment

Further simulations are based on different sizes of defects, defect positions and numbers of defects. The results suggest that the method of comparing correlation coefficients to detect a difference between ventilation and perfusion images is sensitive to the defect volume and the number of defects, but not the location of the defect.