Impact of diffusion-weighted imaging on agreement between radiologists and non-radiologist in musculoskeletal tumor imaging using magnetic resonance.

Background: Diffusion-weighted imaging (DWI) is widely used in neuroradiology or abdominal imaging but not yet implemented in the diagnosis of musculoskeletal tumors. Aim: This study aimed to evaluate how including diffusion imaging in the MRI protocol for patients with musculoskeletal tumors affects the agreement between radiologists and non-radiologist. Methods: Thirty-nine patients with musculoskeletal tumors (Ewing sarcoma, osteosarcoma, and benign tumors) consulted at our institution were included. Three raters with different experience levels evaluated examinations blinded to all clinical data. The final diagnosis was determined by consensus. MRI examinations were split into 1) conventional sequences and 2) conventional sequences combined with DWI. We evaluated the presence or absence of diffusion restriction, solid nature, necrosis, deep localization, and diameter >4 cm as known radiological markers of malignancy. Agreement between raters was evaluated using Gwet's AC1 coefficients and interpreted according to Landis and Koch. Results: The lowest agreement was for diffusion restriction in both groups of raters. Agreement among all raters ranged from 0.51 to 0.945, indicating moderate to almost perfect agreement, and 0.772-0.965 among only radiologists indicating substantial to almost perfect agreement. Conclusion: The agreement in evaluating diffusion-weighted MRI sequences was lower than that for conventional MRI sequences, both among radiologists and non-radiologist and among radiologists alone. This indicates that assessing diffusion imaging is more challenging, and experience may impact the agreement.