Accuracy and interobserver variation in the interpretation of computed tomography in solitary brain lesions

The clinical data and computed tomographic findings of 64 patients with solitary supratentorial brain lesions were presented to two panels of six experienced clinicians. The diagnoses predicted by these clinicians were compared with each other (interobserver variation) and with the definite diagnosis, which in almost all cases was based on histologic examination of the involved tissue (validity of predicted diagnosis). The interobserver agreement was only moderate. The predicted diagnoses agreed with the definite diagnoses in only 57% of cases. A high number of errors were made in distinguishing between high-grade and low-grade glioma and between high-grade glioma and cerebral metastasis, and in the detection of primary cerebral lymphoma. Possible implications of these findings for clinical practice are discussed.