부산대학교 도서관

Objective—To evaluate interobserver agreement and diagnostic accuracy of brain MRI in dogs. Design—Evaluation study. Animals—44 dogs. Procedures—5 board-certified veterinary radiologists with variable MRI experience interpreted transverse T2-weighted (T2w), T2w fluid-attenuated inversion recovery (FLAIR), and T1-weighted-FLAIR; transverse, sagittal, and dorsal T2w; and T1-weighted-FLAIR postcontrast brain sequences (1.5T). Several imaging parameters were scored, including the follow- ing: lesion (present or absent), lesion characteristics laxial localization, mass effect, edema, hemorrhage, and cavitationl, contrast enhancement characteristics, and most likely diagnosis (normal, neoplastic, inflammatory, vascular, metabolic or toxic, or other). Magnetic resonance imaging diagnoses were determined initially without patient information and then repeated, providing history and signalment. For all cases and readers, MRI diagnoses were compared with final diagnoses established with results from histologic examination (when availablel or with other pertinent clinical data (CSF analysis, clinical response to treatment, or MRI follow-up). Magnetic resonance scores were compared between examiners with κ statistics. Results—Reading agreement was substantial to almost perfect 10.64 < κ < 0.861 when identifying a brain lesion on MRI; fair to moderate 10.14 < κ < 0.60) when interpreting hemorrhage, edema, and pattern of contrast enhancement; fair to substantial 10.22 < κ < 0.741 for dural tail sign and categorization of margins of enhancement; and moderate to substantial 10.40 < κ <0.78) for axial localization, presence of mass effect, cavitation, intensity, and distribution of enhancement. Interobserver agreement was moderate to substantial for categories of diagnosis 10.56 < κ < 0.69), and agreement with the final diagnosis was substantial regardless of whether patient information was 10.65 < κ < 0.76) or was not 10.65 < κ < 0.681 provided. Conclusions and Clinical Relevance—The present study found that whereas some MRI features such as edema and hemorrhage were interpreted less consistently, radiologists were reasonably constant and accurate when providing diagnoses. [ABSTRACT FROM AUTHOR]