Histogram matching for the generation of ventilation-perfusion difference images in SPECT lung scanning: a phantom...

Diagnosis of acute pulmonary embolism (PE) is commonly done by acquiring SPECT scans of lung ventilation and of lung perfusion. The two image sets are compared, to identify regions which are ventilated but not perfused ("mismatched defects"). This paper describes the application of histogram matching to the calculation of a ventilation/perfusion difference image, and an investigation of the feasibility of the technique using phantom data.

An empty balloon was inserted into the lung compartment of an anthropomorphic torso phantom. The lungs were filled with polystyrene beads and with water containing 0.20 kBq/ml of (99m)Tc. Two scans were acquired to mimic a matched ventilation/perfusion pair. Then, 30 ml of water containing 0.01 kBq/ml of (99m)Tc was injected into the balloon and the phantom was rescanned. This was repeated four more times, adding 30 ml each time. Each perfusion scan thus had a mismatched defect of a different size. A CT scan was also performed after each perfusion scan, to verify the size and location of the balloon. Histogram matching was applied to each perfusion scan, which was then subtracted from the ventilation scan, yielding a difference image in which voxels with positive values identified mismatched defects. For each scan, a volume of interest (VOI) was automatically generated on the defect and was also copied across to the contralateral side to determine target to background ratios.

All mismatched defects were clearly visible in the difference images, including the smallest, which corresponded in size to a small subsegmental defect. Voxel values for the mismatched defects ranged from 17 to 26, compared with contralateral regions, which had voxel values of 0 or 1.

Histogram matching provides a simple, automatic data-driven method for scaling ventilation and perfusion studies without user intervention.