Enhanced dynamic FDG-PET tumor detection with constrained temporal filtering

Residual FDG activity in normal tissues, such as blood vessels and the liver, as well as the spilled in background activity can impair the detection of small or modestly tracer-avid tumors in FDG-PET cancer imaging. In order to maximize tumor visualization but minimize background and artifacts, an efficient new method, adapted from the constrained temporal filtering processing widely used for signal detection, is extended to the application of dynamic FDG-PET processing. Comparing with the well-known Patlak analysis and spectral analysis (SA), the proposed method can objectively remove the partial volume effect superimposed onto the TAC as well as the residual blood activities and result in the pixel-by-pixel estimations of the influx constant as the output of the filter. Since the constrained temporal filter is designed to preserve the power of tumor signal, therefore, it is likely to offer a desirable noise canceling result, while make no or minimum distortion on tumor signatures. In contrast, Patlak analysis may fail to classify tumor and normal structures when the tumor is severely interfered by background activity. In this case the tumor present approximately similar time activity curve as that of normal tissue at the later stage of acquisition. The signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) are served as the figure of merit for performance evaluation. Our digital phantom study demonstrates that the proposed method outperforms Patlak and SA methods. With microPET two mouse studies are also dynamically acquired on the 6th day and the 12th day, respectively, after the cancer cell implantation. The proposed method is applied to process the earlier acquired images and the resulting findings are confirmed with the later acquisitions. The results show that the new method can enhance the tumor-to-background ratio at an early stage and is promising for improving lesion detectability.