Florian A. Huber (1), David Rotzinger (2), Matthias Zadory (3), Felix C. Müller (4), Anna Døssing (5), Veronique Schuppli (1), Lukas Moser (1), John M. Froehlich (3), Hatem Alkadhi (1), Fabio Becce(2)
Affiliation(s):
1. Institute of Diagnostic and Interventional Radiology, University Hospital Zurich and Faculty of Medicine, University of Zurich, Zurich, Switzerland
2. Department of Diagnostic and Interventional Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland
3. KlusLab Research, Zurich, Switzerland
4. Department of Radiology, Herlev and Gentofte Hospital, University of Copenhagen, Copenhagen, Denmark
5. The Parker Institute, Bispebjerg and Frederiksberg Hospital, Frederiksberg, Denmark
Our objective was to assess the diagnostic performance of the first ever clinical photon-counting detector (PCD) computed tomography (CT) scanner with regard to detection and differentiation of calcium hydroxyapatite (HA) vs. calcium pyrophosphate (CPP), and to compare it with routine energy-integrating detector (EID)-CT.
We performed a prospective in vitro phantom study. Therefore, we used custom-made industry-standard cylindric rods of CPP and HA (both 50, 100, 200 mg/ml). We scanned all samples with PCD-CT twice (at 120 kV and 140 kV) and with the standard-of-care EID-CT (dual-source scanner at 80 kV and tin-filtered (Sn) 150 kV). All scans were performed at an identical and clinically realistic dose using a CT dose index of 8 milligray. Ten fixed-size region of interest measurements of 50 voxels each were performed on axial reconstructions per rod and series, respectively. We calculated Dual energy (DE) CT ratios for each material in each scanner. ROC curves and respective AUC values with 95% confidence intervals were calculated to assess differentiability of CPP vs. HA, for each scan setting, respectively. In addition, we assessed limit of detection (LOD, in mg/ml) for both materials, separately per 50-voxels-ROI and per voxel, using linear regression coefficients of attenuation as a function of concentration.
DECT ratios were slightly different between CPP and HA in EID-CT and in PCD-CT scans. AUC values for the performance of differentiation between CPP and HA was significantly higher in PCD-CT scans (AUC: 0.969 and 0.987 vs. 0.874 for PCD-CT at 120kV and 140kV vs. EID-CT; p<0.05). LOD was overall comparable between methods, with the lowest LOD for CPP ranging from 7.69 – 10.84 mg/ml per voxel and from 2.61 – 3.45 mg/ml per ROI, compared to 8.18 – 9.98 mg/ml per voxel and 2.44 – 3.51 mg/ml per ROI for HA across the different scan methods.
In conclusion, we provide preclinical proof of principle that detection and characterization of calcium-based crystals can be performed with PCD-CT at improved performance compared to the best-performing EID-CT, with excellent accuracy. These first results are promising with respect to the potential development of a reliable imaging assessment method for the differentiation and characterization of calcium-based crystal load involved in symptomatic arthropathies. Further studies will be required to demonstrate comparable diagnostic confidence of PCD-CT in clinical settings.