Purpose: In this work we developed a novel adaptive dual-energy (aDE) algorithm, which accounts for differences in beam attenuation within imaging plane allowing for enhanced material cancellation and noise removal compared to conventional log-subtraction dual-energy (DE) algorithm.
Methods: Pre-calibrated weighting factors were obtained with a step phantom, which consists of overlapping slabs of solid water and bone with different thicknesses. Bone and solid water thickness were varied in [0-6]cm and [0-30]cm range respectively. Thicknesses varied in orthogonal directions forming 49 regions with overlapping materials of different thicknesses. Optimal material cancellation weighting factors for each region was determined by varying weighting factors to achieve zero contrast-to-noise (CNR) between regions with overlapping materials and corresponding target material (bone or soft tissue). The optimal weighting factors were fitted to average intensities of each region in low (LE) and high energy (HE) images. Image noise was suppressed with anti-correlated noise reduction (ACNR) algorithm. ACNR weighting factors were determined by maximizing signal-to-noise-ratio of each overlapping region, and then were fitted with corresponding LE and HE region intensities. To test the DE algorithm, HE and LE images of Rando phantom were fed as inputs to the optimal fitting functions to create patient-specific weighting factors maps. Soft tissue and bone only DE images of Rando phantom were obtained with different DE algorithms and CNR for several selected regions of interest within the Rando phantom were compared for different algorithms.
Results: CNR values for aDE algorithm were higher compared to simple log-subtraction algorithm. Implementation of aACNR algorithm results in further reduction of image noise and improvements in CNR.
Conclusion: A novel DE algorithm was developed, which has improved material cancellation and noise reduction compared to the conventional DE technique. This method can be easily implemented in a clinical environment for real-time DE image generation.