Automatic three-dimensional matching of CT-SPECT and CT-CT to localize lung damage after radiotherapy

The aim of this study was to develop a fast and clinically robust automatic method to register SPECT and CT scans of the lungs.

CT and SPECT scans were acquired in the supine position from 20 patients with healthy lungs. After partial irradiation of the lungs by radiotherapy, the scans were repeated. Two matching methods were compared: a conventional method with external skin markers and a new method using chamfer matching of the lung contours. In the latter method, a unique value for the SPECT threshold, needed for segmentation of the SPECT lungs, was determined by iteratively applying the chamfer matching algorithm.

The new technique for CT-SPECT matching could be implemented in a fully automatic manner and required less than 2 min. No large systematic shifts or rotations were present between the matches obtained with the marker method and the lung contour method for healthy or partially irradiated lungs. For healthy lungs, the number of ventilation SPECT counts outside the CT-defined lung was taken as a measure for a good match. This number of outside counts was slightly lower for the new method than for the conventional method, which indicates that the accuracy of the new method is at least comparable to the conventional method. For ventilation, a systematic difference between the results of the matching methods, a small translation in the anterior --> posterior direction, could be attributed to an inconsistency of the marker positions (2 mm). For perfusion, a somewhat larger anterior --> posterior shift was found, which was attributed to the gravity force. CT-CT correlation on the lung contours using chamfer matching was tested with the same dataset. For accurate matching, the CT slices encompassing the diaphragm had to be deleted.

The new method based on lung contour matching is a fast, automatic procedure and allows accurate clinical follow-up.