Analysis of errors made on in utero MR studies of the foetal brain in the MERIDIAN study.

OBJECTIVES: In utero magnetic resonance (iuMR) imaging to diagnose foetal brain abnormalities has been established and is supported by meta-analyses of retrospective and prospective studies. In this paper we describe and classify the iuMR errors made in the largest diagnostic accuracy study to date (MERIDIAN). We also correlate the error rates and types with the prior experience of the reporting radiologists in order to inform how to provide a national programme with the best diagnostic accuracy achievable. METHODS: The MERIDIAN cohort of 570 foetus formed the basis of this study and included 40 cases with a confirmed diagnostic error, compared with the Outcome Reference Diagnosis. Analysis included the potential clinical effect of the error and classification of error type through an Expert Neuroradiological Panel re-reporting the study. Assessments were made regarding radiologists experience prior to MERIDIAN. RESULTS: The overall confirmed error rate for iuMR was 7·0% and it was considered that there would have been an adverse effect on prognostic information in 22/40 cases if the iuMR had informed counselling. The experienced central reporter made statistically significant fewer errors than the less experienced non-central reporters (3·8% v 11·0%) and the central reporter made fewer clinically significant errors. Furthermore, the type of cognitive errors differed between central and non-central reporters. CONCLUSIONS: Although iuMR imaging improves the diagnostic accuracy of detecting foetal brain abnormalities there remains a substantial error rate, which can have major clinical significance. We have shown that error rates are lower for more experienced reporting radiologists with fewer potential deleterious clinical implications. We discuss the implications of these findings in terms of providing a uniform national service. KEY POINTS: • Overall confirmed error rate for iuMR diagnosing foetal brain abnormalities was 7·0%. • IuMR reports had an adverse effect on counselling in 55% of error cases. • Error rates are consistently lower for more experienced radiologists. • Collaboration between radiologists, dual reporting, overseeing scan and formal training can reduce errors.

MRI during cochlear implant assessment: Should we image the whole brain?

OBJECTIVES: Magnetic resonance imaging (MRI) is a standard part of a cochlear implant assessment in most centres. While there is ample literature on the temporal bone-specific imaging that is required, the role of whole brain imaging has not been as fully studied. We present the first report of the incidence of associated brain abnormalities in the whole cochlear implant population, including adults and consider their significance. METHODS: We retrospectively reviewed 51 (12 adults and 39 children) sequential cases since we added whole brain MRI sequences to our cochlear implant assessment protocol. We reviewed the scans for abnormalities of the cochlea and cochlear nerve and a neuroradiologist reviewed the images of the whole brain sequences for further abnormalities. RESULTS: We identified abnormalities on the whole brain sequences in 21 (41%) of these patients, 5 of 12 adults (42%) and 16 of 39 children (41%). Thirty-six (71%) patients subsequently had at least one implant inserted, 13 with abnormalities on whole brain MRI (36%) and 23 without. Of the 15 patients who did not undergo subsequent implantation, 8 had positive findings on their whole brain MRI sequence (53%). There was no statistical difference in the probability of finding an abnormality on the whole brain MRI between those who did and those who did not go on to have an implant (P = 0.35). There were abnormalities within the inner ear in five patients. DISCUSSION: The abnormalities detected on the whole brain images are heterogenous and of wide ranging clinical significance ranging from truly incidental findings to abnormalities that are so severe that they may predict a very poor prognosis such that an implant may contribute little.