Clin Radiol. 2026 Apr 28;99:107371. doi: 10.1016/j.crad.2026.107371. Online ahead of print.
ABSTRACT
AIM: When radiologists encounter an unfamiliar musculoskeletal neoplasm on conventional and advanced MR images, they now have an option to search the internet for MRI image examples. The purpose of this project was to systematically evaluate the quality of images returned by a widely used general search engine.
MATERIALS AND METHODS: Systematic internet searches were conducted for 16 benign and malignant musculoskeletal neoplasms, focusing on T1-weighted (T1W), T2-weighted (T2W), contrast-enhanced, diffusion-weighted (DWI), and out-of-phase MR images. The top five images from each search were evaluated for image quality and clinical relevance using a 5-point scoring key.
RESULTS: General internet engine returned correct sequence among the top five results for 88% of lesions when searched for T1W images, 100% for T2W images, 100% for contrast-enhanced images, 63% for DWI, and 25% for out-of-phase images. Fleiss Kappa statistic demonstrated substantial agreement (Kappa = 0.72) for binary “useful” vs “not useful” image designation, and moderate agreement (Kappa = 0.48) for all five categories.
CONCLUSION: The general internet search engine returned useful results when searching for conventional MRI sequences but performed sub-optimally when searching for advanced MR image examples.
PMID:42275688 | DOI:10.1016/j.crad.2026.107371
