Objective To evaluate the diagnostic accuracy of classical measurements for basilar invagination (BI) of type B at MRI. Methods This study used head MRIs from 31 participants with BI type B and 96 controls. The radiological criterion for BI was the odontoid process invagination using the obex as reference. It based on the independent prospective reading of two neuroradiologists. Concordance between the two neuroradiologists was analysed through the KAPPA index, and the discrepancy was resolved in a consensus meeting. A third examiner measured in two occasions (double blind) the distance of the odontoid apex to Chamberlain's line (DOCL) and McGregor's line (DOMG), clivus canal angle (CCA), Welcker's basal angle (WBA), and Boogaard's angle (BOA). Intra-examiner reproducibility of the measurements was evaluated with the intraclass correlation coefficient and the diagnostic accuracy by ROC curve. All analyses were at 95% confidence interval. Results Agreement between the two neuroradiologists was statistically relevant (KAPPA = .91; P = .0001). The intra-examiner reproducibilities were .98 (DOCL), .97 (DOMG), .96 (CCA), .94 (WBA), and .95 (BOA) (P < .05). The areas under the ROC curve were .963 (DOCL), .940 (DOMG), .880 (CCA), .867 (WBA), and .951 (BOA) (P < .05). The cut-off criteria were ≥ 7 mm (DOCL), ≥ 8 mm (DOMG), ≤ 145° (CCA), ≥ 142° (WBA), and ≥ 136° (BOA). The diagnostic accuracies were .904 (DOCL), .870 (DOMG), .844 (CCA), .810 (WBA), and .899 (BOA). Conclusion The DOCL and BOA presented the highest diagnostic accuracy for BI type B.