Diagnostic accuracy of MR, CT, and ECT in the differentiation of neoplastic from nonneoplastic spine lesions.

ABSTRACT

AIM: To provide guidance for appropriate imaging examinations for diagnosing spinal tumors or tumor-like lesions. METHODS: A total of 121 patients with suspected spinal tumors were included this retrospective study. Each patient underwent ≥2 imaging examinations, including computerized tomography (CT), magnetic resonance (MR), and/or emission computed tomography (ECT). All patients were diagnosed by pathology after core needle or surgical biopsies. The results were compared with those of pathological examinations using paired chi-squared tests, and compared with each other. Statistical indicators that tested the consistency of the results included McNemar's and kappa coefficients, as well as receiver operating characteristic curves. RESULTS: The differences among MR, CT, ECT, and pathology were not significant. The kappa coefficient of MR, CT, and ECT was 46.1%, 36.0%, and 55.9%, respectively. The area under the curve of ECT, MR, and CT scans was 0.809, 0.705, and 0.704, respectively; and the differences among them were significant (P < .05). Post hoc multiple comparisons showed no significant differences among imaging examinations in terms of sensitivity, specificity, misdiagnosis rate, and coincidence rate (P > .05). However, significant differences were noted in the kappa coefficient and missed diagnosis rate (P < .05). CONCLUSIONS: Although ECT was the most accurate imaging method, its high cost and large radiation dosage limit its widespread application. Furthermore, MR verified spinal tumors more effectively; however, CT excluded them more efficiently. In summary, when all factors are considered, MR is still the optimal modality for the diagnosis of spinal tumors, especially during the initial screening.