• Imaging
• Neonatology
• Neurodevelopment
• Neurodisability

MRI is a neuroimaging technique which is increasingly being used in extremely preterm infants cared for on a level 3 neonatal intensive care unit. The technique was initially introduced in the 1980s in newborn infants with severe brain injury and was considered to be very useful for the prediction of especially motor outcome. Many centres have started to perform MRI routinely at discharge or term equivalent age (TEA) and have been able to show that subtle white matter abnormalities are better and more reliably depicted with MRI than cranial ultrasound (cUS). A landmark paper by Woodward et al1 of 167 infants born before 30  weeks’ gestation concluded that abnormal findings—in particular moderate-to-severe white matter injury (WMI)—on brain MRI at TEA are significantly better at predicting adverse neurodevelopmental outcome at 2 years of age than abnormal findings on cUS. In more recent years, the MRI technique has become more sophisticated, with new sequences providing additional information. Furthermore, sequential MRI provides insight into growth and development of the brain and neuronal networks of the extremely preterm infant, using 3D-volumetric imaging, diffusion tensor imaging and functional MRI.2

MRI is considerably more expensive than cUS, necessitates the transport of infants to the scanner, often requires sedation of the patient and cannot be repeated easily. To obtain high-quality preterm MRI, age-appropriate sequences with thin slices must be used. Without sedation, there is a risk of obtaining images that display movement artefacts, making interpretation more difficult and being inferior to good sequential cUS. A single MRI at TEA will not identify a minor haemorrhage or smaller cystic lesions that may have resolved between birth and TEA (a period sometimes exceeding 3 months).

Whether every extremely preterm infant should have at least one MRI and when is the best time to perform this MRI are questions which are often raised. In an accompanying article, Plaisier and colleagues address these questions, comparing sequential cUS with a single early MRI.3 This is different from most previous studies, comparing sequential cUS with an MRI performed at discharge or TEA. Only a few studies have performed sequential MRIs, and two of these studies compared cUS data with MRI findings. In the study by Debillon and colleagues, early MRI findings predicted late MRI findings in 98% of patients (95% CI 89.5 to 99.9) compared with only 68% for early cUS (95% CI 52.1 to 79.2).4 In a study by Miller and colleagues using early and TEA-MRI, WMI remained the same in 41 newborns (87%), and intraventricular haemorrhage (IVH) was the same grade at both time points in 44 (90%).5 These studies would suggest that in most, though not all, infants an early MRI provides sufficient information and that there is no need for a second MRI at discharge or TEA.

Performing an MRI early has the advantage of obtaining detailed information early and diagnosing some of the milder lesions which may have resolved by TEA. In our centre, we performed sequential MRIs in 112 preterm infants with punctate white matter lesions, and noted an evolution of the lesions between the two MRIs often with a change in the appearance of the lesions and an apparent decrease in lesion load between the early and TEA-MRI.6 Similar to the data of Plaisier and colleagues, we were not able to perform an early MRI in some of the sickest infants, where the MRI might have been most informative. On the other hand, performing an MRI at TEA will additionally show injury with an onset beyond 30 weeks postmenstrual age (PMA), myelination of the posterior limb of the internal capsule and the size of the extracerebral space indicative of brain growth. Using advanced MRI techniques, adverse effects of the neonatal course will not yet be apparent at 30 weeks PMA, but are more likely to be visible and measurable at TEA.

The strength of the accompanying study is that the authors are well known for their expertise in performing cUS as well as performing good quality MRI, allowing a fair comparison between the two techniques. Although cUS was also performed beyond the early MRI, we are not told whether these later cUS scans provided additional information, for example, about progressive ventricular dilatation, either posthaemorrhagic or ex vacuo following WMI. We wondered why the number of infants with WMI was so small and why there was a discrepancy between the number of infants with cystic periventricular leukomalacia, which was more often diagnosed with cUS (n=3) than with MRI (n=1); even though the authors say that none of the discrepancies found between cUS and MRI was due to the onset of brain lesions after the MRI was performed, it is important to realise that it is not uncommon to see the development of severe WMI or perforator stroke beyond 30 weeks PMA, for example, following necrotising enterocolitis or a viral illness.7 Repeating cUS and, whenever possible, an MRI following any clinical deterioration is therefore strongly recommended.

A major limitation of their study is that of the 307 eligible infants, 126 (41%) did not get an MRI due to death or transfer to a level 2 hospital before 30 weeks PMA. Of the remaining 181 infants, only 122 (67%) were stable enough to have an early MRI and these were the infants with a less complicated neonatal course, compared with the 59 who were scanned at a later point in time. It would be of interest to know how many of the MRIs were difficult to assess because of movement artefact, which is, in our experience, a much more common problem with the early scan compared with the TEA-MRI. The discrepancy between diagnosing cranial sinus venous thrombosis (CSVT) on cUS and MRI is not in agreement with their recent data, where CSVT was confirmed with MRI in seven out of nine infants.8 Although not clearly stated, it seems that cUS rather than MRI was taken as the gold standard for diagnosing IVH in the absence of histological data, while in previous studies MRI was used as the gold standard. They suggest that cUS is more reliable for diagnosing a small IVH, which was not seen on the early MRI. However, it is also possible that they overdiagnosed IVH with cUS. Susceptibility weighted imaging (SWI) would have been able to clarify this issue, but this sequence was only introduced in the latter part of their study. They do not mention the diagnosis of punctate white matter lesions in the tables, even though it is well illustrated in their figure 2. This diagnosis was commonly found by Kersbergen and colleagues as well as being reported in the studies from Miller and colleagues and both groups found it to be most marked on an early MRI.5 ,6

As expected, the main discrepancy found between the two neuroimaging techniques was related to small (folial) cerebellar haemorrhages which were best detected on MRI. This in agreement with the second paper by Parodi and colleagues, comparing sequential cUS, using the anterior and mastoid window, with SWI-MRI.9 As has previously been shown by others, small cerebellar lesions (<5 mm) are seldom identified using cUS.10 Parodi and colleagues enrolled infants with a gestational age up to 34 weeks, even though it is well known that the most immature infants are most at risk of developing cerebellar lesions. Even so, the incidence of all cerebellar haemorrhages in their cohort is 20%, compared with only 6.8% in the cohort of Plaisier and colleagues, who studied a more immature cohort. The lower incidence found in the cohort of Plaisier and colleagues is likely to be due to the introduction of SWI only in the latter part of their study. It would therefore have been of interest and of additional value to have a comparison between cUS with conventional MRI as well as SWI in the study by Parodi et al.9

As the best time to perform an MRI in preterm infants is still uncertain, additional data like those from Plaisier and colleagues are extremely valuable. They suggest that an MRI at 30 weeks PMA will help in the decision-making process to redirect care. We however doubt this, as these infants may either be too unstable to be transferred to the MRI unit as well illustrated in their study, or the brain lesion is likely to be so severe that cUS will provide sufficient information together with neurophysiology and neurological examinations to make the appropriate decision. Performing an MRI early and at TEA will however help to determine brain growth over time and evaluate the effect of neuroprotective strategies. As both studies have elegantly shown, the quality of cUS has improved over the years and the two neuroimaging techniques are still complimentary. We look forward to the correlation between the neuroimaging findings and long-term outcome in both studies, which will further establish the additional value of an early MRI and the effects of cerebellar lesions for the future of infants born preterm.