Ke Li MS, Presenter: Nothing to Disclose

Nicholas Benjamin Bevins PhD, Abstract Co-Author: Nothing to Disclose

Joseph Zambelli PhD, 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

Although it has been shown that x-ray differential phase contrast (DPC-CT) provides superior contrast of certain materials, a clinically relevant question is determining where DPC-CT can provide improved detectability over absorption CT, as both the signal and noise properties of the two modalities are radically different. The purpose of this work is to present our human observer studies on the performance of DPC-CT and absorption CT for a range of imaging tasks under the constraint of equivalent CNR.

Noise images used in the study were collected from a DPC-CT system. Raw projections were taken from 360 views with 1° angular interval, and were reconstructed to form absorption CT and DPC-CT images. Digital objects analogous to clinically relevant tasks (detection of large uniform masses, discrimination between a vessel and a nodule, and discrimination between lobulated or uniform lesions) were added to the experimental data. For this study, the CNR in the two modalities were matched. For each task and each modality, 200 images were generated. Four physicists rated each image using a 5-point scale. Responses were analyzed using ROCKIT1.1B2.

Operating under the constraint of equal CNR, for the detection task analogous to a uniform mass, Az=1 for absorption CT and Az=0.83 for DPC-CT (p<0.001). For discriminating an object analogous to a vessel from an object analogous to a nodule, Az=0.99 for absorption CT and Az=0.86 for DPC-CT (p<0.05). For discriminating an object analogous to a lobulated lesion from an object analogous to a uniform mass, Az=0.88 for absorption CT and Az=1 for DPC-CT (p<0.001). The ROC results show that when the CNR are matched, detectability in absorption CT and DPC-CT can be very different depending on the spatial frequency content of the imaging task.

For equal CNR, this study indicates that superiority of DPC-CT to absorption CT depends on the imaging task. However, at high spatial resolutions the CNR of DPC-CT may greatly exceed absorption CT because of differences in contrast formation and noise scaling laws. The evaluation of DPC-CT's clinical potential must take not only its superior contrast but also the unique noise texture into account.

Determining dominating factors influencing DPC-CT's low contrast detectability plays an essential role in moving it toward the clinic.

Li, K, Bevins, N, Zambelli, J, Chen, G, Human Observer Performance Studies of X-ray Differential Phase Contrast CT.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12027827.html