Neonatal brain MRI: how reliable is the radiologist’s eye?

Abstract

This is a serious limitation of which we need to be aware, as it can lead to contradictory conclusions in the challenging context of term-equivalent age brain MRI in premature infants. These results highlight the need for a semiautomatic tool to help in objectively analyzing MRI signal intensity in the neonatal brain.

The challenge of cerebral magnetic resonance imaging in neonates: A new method using mathematical morphology for the segmentation of structures including diffuse excessive high signal intensities.

“…Obs.1. As shown in [33], deciding visually whether hyperintensities are present or not is highly subjective, and the precise delineation of the corresponding regions is even more difficult. This is confirmed by the low values of Dice coefficient, and high values of HD and AVD.…”

“…[6,24,39]. A hypothesis to explain this controversy may be the lack of robustness and reproducibility in the assessment itself of the presence of DEHSI [33]. A reproducible automatic (or semiautomatic) segmentation of DEHSI may benefit from the segmentation of newborn brain MR images into different brain tissues, i.e., cortical gray matter (CoGM), basal ganglia and thalami (BGT), white matter (WM), ventricles (Vent), and cerebrospinal fluid (CSF).…”

“…The "gold standard" for neonatal brain segmentation [42] is manual delineation, which must follow precise guidelines (e.g., the protocol proposed in [18]) to be reproducible, and which relies on anatomical landmarks when possible. Yet, due to the challenging issues mentioned above, one may have different manual segmentation results at different times, and disagreement among experts can be significant [33]. This non-reproducible issue is especially important in the evaluation of DEHSI.…”

“…The manual corrections allow the user to correct the segmentation process on the fly if needed, and to further improve the accuracy of the segmentation, according to each user's need. Indeed, as demonstrated in [33], variability among radiologists could be very significant. As demonstrated in Sections 3 and 4, these manual interactions and corrections are easy to apply.…”

Sequelae of Premature Birth in Young Adults

“…Perinatal WMI covers a spectrum ranging from its most severe form, cystic periventricular leukomalacia (PVL), to more diffuse forms caused by disturbed pre-oligodendrocyte maturation and axon myelination [9][10][11]. Punctate white matter lesions (PWML), ventriculomegaly due to white matter dysmaturation and diffuse excessive high signal intensity (DEHSI) on FLAIRweighted images have been proposed as correlates of noncystic WMI in prematurely born infants [10][11][12]; however, DEHSI has been shown to be a highly subjective imaging feature, dependent on the time of scanning and without reliable associations to outcomes of neurodevelopment [12,13]. The pathophysiology underlying long-term structural brain alterations is under extensive investigation on a microstructural and macrostructural level using advanced neuroimaging methods [2,[14][15][16][17].…”