Purpose: Dual-energy CT (DECT) is arguably the most accurate energy mapping technique in CT-based attenuation correction (CTAC) implemented on hybrid PET/CT systems. However, this approach is not attractive for clinical use owing to increased patient dose. The authors propose a novel energy mapping approach referred to as virtual DECT (VDECT) taking advantage of the DECT formulation but using CT data acquired at a single energy (kV(P)). For this purpose, the CT image acquired at one energy is used to generate the CT image at a second energy using calculated kV(P) conversion curves derived from phantom studies.
Methods: The attenuation map (μ-map) at 511 keV was generated for the XCAT phantom and clinical studies using the bilinear, DECT, and VDECT techniques. The generated μ-maps at 511 keV are compared to the reference derived from the XCAT phantom serving as ground truth. PET data generated from a predefined activity map for the XCAT phantom were then corrected for attenuation using μ-maps generated using the different energy mapping approaches. In addition, the generated μ-maps using the above described methods for a cylindrical polyethylene phantom containing different concentrations of K(2)HPO(4) in water were compared to actual attenuation coefficients. Likewise, CT images of five clinical whole-body studies were used to generate μ-maps using the various energy-mapping approaches were compared with μ-maps acquired at 511 keV using (68)Ge/(68)Ga rod sources for the clinical studies.
Results: The results of phantom studies demonstrate that the proposed method is more accurate than the bilinear technique. All three μ-maps yielded almost similar results for soft and lung tissues whereas for bone tissues, the DECT and the VDECT methods produced a much smaller mean relative difference (3.0% and 2.8%, respectively) than the bilinear approach (11.8%). Likewise, the comparison of PET images corrected for attenuation using the various methods showed that the proposed method provides better accuracy (6.5%) than the bilinear method (13.4%). Clinical studies further demonstrated that, compared to the bilinear method, the VDECT approach has better agreement for bony structures with the DECT technique (1.5% versus 8.9%) and transmission scanning (8.8% versus 17.7%).
Conclusions: It was concluded that the proposed method outperforms the bilinear method especially in bony structures. Further evaluation using a large clinical PET/CT database is underway to evaluate the potential of the technique in a clinical setting.