Simultaneous dual-radionuclide imaging: are the images trustworthy?

The routine use of a single radionuclide for patient imaging in nuclear medicine can potentially be significantly enhanced by studies employing two tracers to examine two different processes in a single organ. Such studies most frequently employ simultaneous imaging of both radionuclides in two different energy windows, yielding perfectly registered image pairs. This approach has been in use for over a decade, for example in parathyroid disease [1], and more recently in liver disease [2], cardiac disease [3-7], and studies of brain perfusion [8]. In addition, simultaneous transmission/emission imaging has been described in single-photon emission tomography (SPET) [9, 10], with one radionuclide used for the transmission study and a second for the emission study. Indeed, such systems are now commercially available. There is thus currently considerable interest in dual-radionuclide imaging. This interest will likely continue to increase because of the already demonstrated utility of "rest/stress" studies, particularly in the heart and brain, and the need for accurate attenuation correction in SPET studies. A major problem with all dual-radionuclide imaging is the "crosstalk" between the two radionuclides. Such crosstalk frequently occurs, because scattered radiation from the higher energy radionuclide is detected in the lower energy window, and because the lower energy radionuclide may have higher energy emissions which are detected in the higher energy window. Depending on the energies of emissions from the two radionuclides, the positions of the energy windows relative to these emissions, and the energy resolution of the imaging system, the crosstalk contribution may be of either lower or higher spatial frequency in character. For example, in the case of 2°iT1 and 99mTc, the crosstalk from 99mTc into the 20~T1 window is primarily from multiply scattered photons and Pb x-rays from the collimator, and is thus primarily low frequency in character, while the crosstalk from 2°IT1 into the 99mTc window is from both unscattered and scattered photons, and thus contains some higher frequency information [10]. On the other hand, in the case of 99mTc-99m and 123I, much of the crosstalk is from unscattered photons, and thus may contain significant high spatial frequency information [11]. Crosstalk has been demonstrated to be a major problem in many dual-radionuclide imaging studies. For example, Ivanovic and co-workers [12] used Monte Carlo simulations to demonstrate that the crosstalk component may be as high as 50% in SPET brain imaging. Weinstein and co-workers [7] and Kiat and co-workers [13] demonstrated similar magnitude effects in myocardial perfusion SPET. Kiat et al. concluded that the "simultaneous dual-radionuclide protocol should not be used until satisfactory techniques for correction of Tc-99m crosstalk are developed". There is only one published study indicating minimal effects of crosstalk [14]. Of importance, this study used phantoms with large defects. An accompanying editorial [15] warned of the limited utility of these findings in human SPET studies. Crosstalk can be conceptualized as an object-dependent addition to the "direct" radionuclide signal in each window. For example, Bailey and co-workers [9] modeled the imaging process as linear and shift-invariant. They described the crosstalk from the higher energy radionucl ide into the lower energy window as the convolution of the higher energy radionuclide's distribution with a crosstalk function (which they referred to as the point spread function for the higher energy radionuclide as seen in the lower energy window). In practice, since the true distribution for the higher energy radionuclide was unknown, an empirically modeled scatter function was convolved with the higher energy window image, and the resulting "crosstalk image" subtracted from the lower energy window image. As an alternative, Frey and coworkers [10] modeled the crosstalk as a contribution to the other window proportional to the direct window image, and empirically obtained proportionality constants (which they referred to as correction factors). In essence, a fraction of each direct window image (determined by the appropriate proportionality constant) was subtracted from the other window, although the exact implementation was more analytically correct than this oversimplified statement. Knesaurek [16] extended Bailey's approach to dualradionuclide 2°lT1/99mTc cardiac SPET. He showed significant improvement in quantification of the distribution of 2°1T1 in a cardiac phantom as a result of crosstalk correction. Of importance, Knesaurek's goal was only to "remove" the crosstalk contribution to each image, resulting in images which were still degraded by direct blur and noise. Presumably, these images could then be