How well do radiologists diagnose intracerebral tumour histology on CT? Findings from a prospective multicentre study

The management of patients with intracranial tumours relies on accurate diagnosis of tumour type. To assess the accuracy with which tumour histology could be diagnosed from brain CT we reviewed data from a prospective, population-based study from three Scottish neuroscience centres. Between October 1997 and April 1999 all patients from the Edinburgh, Aberdeen and Dundee Neuroscience Centres with a CT-diagnosis of a solitary, supratentorial intra-cerebral tumour were recruited. General and neuroradiologists were asked to give their best guess tumour diagnosis. Biopsy confirmed neuropathological tumour type was available for comparison with best guess CT-diagnosis in 221 of 324 patients. Histological diagnosis was either malignant glioma, low grade glioma or metastasis in 199 cases. The accuracy of CT lesional. diagnosis for these three categories was 60% (95% confidence interval 54-67%), 85% (80-89%) and 82% (77-88%), respectively. The diagnostic accuracy of an intra-cerebral tumour was 0.81 with a positive predictive value of 0.93. There was no significant difference between histological diagnostic accuracy of neuroradiologists and general radiologists. In 10% of patients the initial CT scan was reported as negative for intracranial tumour, with 62% of these scans having been carried out without contrast. Based on CT alone radiologists are good at identifying an intra-cerebral tumour, but not so good at distinguishing between different tumour types. The implications of the findings for patient management are discussed.