Automatic polar map projection for cardiac PET images

Objectives Determining the long axis of left ventricular is an essential procedure for polar map projection in cardiac PET studies. It usually relies on manual processing which is subjective and time-consuming. In this study, an automatic method was explored to generate polar map objectively.

Methods According to cardiac PET images, the region of left ventricle was firstly derived by threshold and morphological processing. Then the left ventricle was fitted to an ellipsoid using least square algorithm and determined its long axis. The left ventricle was rotated to making the long axis parallel with the X axis. Secondly, the rotation correction was performed to eliminate the error between 3D and 2D mode. In this step, the least square method was used to estimate the corrected long axis based on the center of rotated image in each transverse slice. After that, the angle between the new long axis and X axis would be the smallest. Finally, the polar map projection can be automatically performed by using coordinate mapping algorithm. This proposed method was evaluated by 6 sets of cardiac PET images. The results were compared with manual results which were obtained by experienced physicians. The performance of the rotation correction was also evaluated by comparing the results of step 1 and step 2.

Results The results of the performance evaluation were listed in the following table. The angles between the corrected long axis and X axis in the plane of X-Y and X-Z were significantly reduced (86.2% and 67.4%) compared with the estimates without rotation correction. While comparing with the manual results, the average angle in the plane of X-Y was increased by 6.44%, the average angle in the plane of X-Z was reduced by 9.48%.

Conclusions The proposed method has the potential to provide an effective and convenient solution for automatic polar map projection of cardiac PET images.