Diagnostic accuracy and agreement between whole-body diffusion MRI and bone scintigraphy in detecting bone metastases

Abstract

This study was done to determine the diagnostic value of whole-body magnetic resonance using diffusion-weighted imaging with background suppression (WB-DWIBS) for detecting bone metastases compared with whole-body bone scintigraphy (WB-BS). Twenty-three patients with solid tumours underwent both WB-DWIBS imaging and WBBS. A nuclear medicine specialist interpreted WB-BS images and two blinded radiologists, first independently and then jointly, interpreted the WB-DWIBS images by completing a reading grid categorising the skeletal segments. Cohen's k statistic was used to determine interobserver agreement in reading the WB-DWIBS images and the agreement between WB-BS and WB-DWIBS. Sensitivity and specificity were calculated per patient and per lesion. Interobserver agreement in reading the WBDWIBS images was substantial or good, with κ=0.68. Analysis of agreement between the nuclear physician's and the radiologists' readings provided κ=0.87 [95% confidence interval (CI)=0.76-0.98)] Per-lesion analysis gave a sensitivity of 80% (95% CI=75-85) and a specificity of 98.2% (95% CI=96.5-99.8). We found a good level of interobserver agreement for the WB-DWIBS images and an excellent level of agreement in the subjective judgement of presence or absence of disease between WB-BS and WB-DWIBS after consensual double reading. WB-DWIBS has the same specificity as WB-BS in detecting bone metastases. The anatomical sites exhibiting the highest level of disagreement between WB-DWIBS and WB-BS are the pelvis, the coccyx, and the sternum, all sites at which detection with WB-BS has the greatest limitations.