Stephen T. Brunner BS, Presenter: Nothing to Disclose

Yinsheng Li, Abstract Co-Author: Nothing to Disclose

Guang-Hong Chen PhD, Abstract Co-Author: Research funded, General Electric Company Research funded, Siemens AG Research funded, Varian Medical Systems, Inc Research funded, Hologic, Inc

The purpose of this study was to investigate the feasibility of using prior image constrained compressed sensing for radiation dose efficiency improvement in photon-counting CT.

A numerical phantom (24 cm diameter) is simulated with water background and inserts of air, bone, breast, and adipose (densities 1.0×ρadipose, 1.3×ρadipose, 1.4×ρadipose, 1.5×ρadipose). The inserts have diameters: 2, 4, 8, 12, 16, 20, and 30 mm. A 140 kVp x-ray spectrum is generated using Spektr (300,000 total photons). The transmitted photons are collected in 6 energy bins: [20, 40] keV, [40, 60] keV, [60, 80] keV, [80, 100] keV, [100, 120] keV, [120, 140] keV. Poisson noise is added to the transmitted photons in each energy bin. FBP is used to reconstruct each energy bin. Then, the 6 energy bin images are averaged to form the low-noise prior image. PICCS reconstructions are generated using the prior image to constrain the reconstruction of each energy bin. In each energy bin image, contrast-to-noise ratio is quantified between contrast insert and background. The average CNR and the standard deviation of the CNRs across contrast sizes are calculated. Accuracy is quantified by percent difference between the contrast insert and the no noise ground truth.

For 1.5×ρadipose, the maximum CNR was obtained in the [60, 80] keV image with CNRFBP=6.39±.36 and CNRPICCS=13.40±.80. The FBP accuracy was 0.79%±0.79% vs. 0.53%±0.43% for PICCS. For 1.4×ρadipose, the maximum CNR was achieved in the [60, 80] keV image with CNRFBP=4.75±.53 vs. CNRPICCS=12.25±3.83. The FBP accuracy was 5.41%±6.20% vs. 2.45%±2.05% for PICCS. For 1.3×ρadipose, the maximum CNR was achieved in the [60, 80] keV image with CNRFBP=3.20±.62 and CNRPICCS=6.04±.89. The FBP accuracy was 4.38%±3.58% vs. 4.62%±8.27% for PICCS. For 1.0×ρadipose, the maximum CNR was achieved in the [40, 60] keV image with CNRFBP=1.92±.34 vs. CNRPICCS=3.73±.29. The FBP accuracy was 2.13%±1.96% vs. 1.96%±2.7% for PICCS.

The application of PICCS in photon-counting CT can maintain accuracy and improve radiation dose efficiency by utilizing a low noise prior image formed from the average of multiple energy bin images.

Methods of improving image quality in photon-counting CT are important to study in order to assess the potential clinical benefit of photon-counting CT technology.

Brunner, S, Li, Y, Chen, G, Radiation Dose Efficiency Improvement in Photon-counting CT Using Prior Image Constrained Compressed Sensing.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12034978.html