Abstract
1499
Objectives The recently introduced xSPECT Bone (xB) incorporates tissue boundary delineation from a CT for bone SPECT imaging with Tc99m DPD or MDP. We compare confidence in image interpretation between Flash3d (F3D, OSEM3D) and a representative prototype of xB (previously: HD-Bone).
Methods We show confidence levels from 6 physicians reading 76 anonymized scans reconstructed with either F3d or xB, displayed as either F3d, xB, F3d/CT, or xB/CT, in fully randomized sessions. All SPECT images were cross registered and included 3d resolution recovery, attenuation and scatter correction. The readers assessed 3 VOI's placed on 3 bones. The VOI’s were pre-selected by a physician, unknown to the readers and included bones with malignant or benign lesions. A five point scale was used to rate if a lesion was present in the VOI, and then if the lesion was benign or malignant. We used a chi2 analysis to assess between-modality correlation and frequency analysis z score to compare reader confidence.
Results Lesion detection: Correlations between the confidence levels with xB and F3d were high with Chi2 values of 514 without CT and 607 with CT. The increase in confidence level with xB as compared to F3d was 1.33x higher without CT (Z=3.75) and 1.30x higher with CT (Z=3.43). Lesion characterization: Correlations between the confidence levels with xB and F3d were also high with Chi2 values of 99 without CT and 207 with CT. The increase in confidence level with xB as compared to F3d was 3.07x without CT (Z=11.12) and 1.46x with CT (Z=4.37). p<0.001 for all.
Conclusions Frequency (Z score) analysis shows that the confidence level to detect and characterize lesions is significantly higher with xB than with F3d both with and without display of the fused CT scan, despite the correlation between the xB and F3d scores.