Mri findings, looking behaviour and affect recognition in very preterm children: A pilot study

Children born very preterm often exhibit atypical gaze behaviors, affect recognition difficulties and are at risk for cerebral white matter damage. This study explored links between these sequalae. In 24 12-year-old children born very preterm, ventricle size using Evans and posterior ventricle indices

Children born very preterm have increased risk for adverse neurodevelopment including visual perception [1] and social functioning [2].Such difficulties include poorer recognition of emotional facial expressions (i.e.affect recognition) [3,4], less positive play interactions with peers, and less synchronized interactions with parents [5].As a group, preterm born individuals also display more autistic traits [6], are more often alone [2] and socially withdrawn [7] compared to term born peers.
Social functioning is complex, and several abilities are coordinated and timed to achieve fluency in social interactions.Minor deficits in one of those abilities may not be obvious, but can have a large impact on social interplay.Proficiency in recognizing affect, tied to visual perception, is crucial for social interaction [8].Accurate affect recognition relies on social attention such as gaze scanning.Typically developing children attend more to the eyes than the mouth during scanning of faces [9].Preterm-born children aged 1 to 10 years exhibit shorter attention spans on faces [10] and less eye focus than term-born peers [11].Children aged 8 to 18 with bipolar disorder show atypical face scanning, correlating with deficits in emotion labeling [12].In school-aged term-born children, deficient affect recognition correlates with lower social status and negative peer experiences [8].Additionally, in typically developing individuals (ages [13][14][15][16][17][18][19][20][21], eye-directed gaze relates to faster face-sensitive brain responses [13].Consequently, studies of individuals with autism spectrum disorder demonstrate less focus to the eyes than mouth when viewing faces, compared to neurotypical and developmentally disabled groups [14].Further, autistic children with better socio-emotional skills focuse more on eyes [15].Less eye focus suggests reduced social interest, linked to impaired interaction. Very preterm infants are born during a phase of intensive structural brain development.Due to immaturity, the brain is vulnerable to complications, often affecting the periventricular white matter and the visual pathways [16].Perinatally, multiple factors including disturbed vascular perfusion can cause white matter injury.During the first weeks after preterm birth, a lack of blood regulating abilities of the vascular system can lead to intraventricular hemorrhage (IVH), or bleeding into the surrounding parenchyma.These complications are risk factors for suboptimal neurodevelopment [17].The maturing oligodendrocytes are also vulnerable to conditions outside of the womb, leading to white matter alterations [18].Both white matter injury and IVH may affect the periventricular area.Ventricular enlargement, especially wider posterior horns of the lateral ventricles, is prevalent.These white matter regions also include the visual pathways [19].Thus, multiple factors affect brain development in very preterm-born infants [16][17][18][19], resulting in smaller brain volumes.A meta-analysis, covering very preterm born individuals from early childhood to adolescence, found that white and grey matter was reduced by 0.5 SD at a group level when compared to full-term peers [19].The volume of subcortical structures, including the corpus callosum, is similarly affected [19].Enlarged ventricles and thinner posterior corpus callosum persist into adulthood after very preterm birth [20].Several studies show that quantitative measures of ventricle enlargement have high inter-observer reliability [18,21].
A recent study [18] of young adults born very preterm, or with very low birth weight, evaluated the use of Evans index and a combined index of the frontal and occipital horns ratio to appraise ventricular volume.These indices' predictive value for ventricular enlargement compared to volumetry was highly significant.Gestational age and intracranial hemorrhage were significant risk factors for ventricular enlargement, suggesting that the parenchymal atrophy might result either from white matter dysmaturation or damage after hemorrhage [18].By using a posterior ventricle index (ratio of the maximum width of the occipital horns of the ventricles and the maximal internal diameter of the skull) a recent study demonstrated that posterior ventricular enlargement was associated with reduced visual acuity and contrast sensitivity in a group of 12-year-old children born very preterm.The authors suggested that this was due to disturbances of the posterior visual pathways related to diffuse white matter lesions [21].
We wanted to explore Evans and posterior ventricle indices further, by connecting them to functional measures of affect recognition in a group of 12-year-old preterm-born children.We reasoned that a posterior ventricle index is relevant as it is associated with white matter volume including the posterior visual pathways.
Possible white matter alterations affecting networks for processing of emotional faces should be reflected in behavior and strategies during active affect perception.The subtest Affect Recognition from the Developmental neuropsychological assessment battery (NEPSY-II) measures how well a child can match emotional expressions between photos.With an eye-tracker, the gaze scanning strategy (i.e.attention to the eyes versus the mouth, number of gaze shifts to the eyes versus to the mouth) and processing speed (i.e.time to give an answer) can be explored.These measures, analysed in relation to brain imaging, can shed light on the behavioral effect of white matter alterations.
Consequently, this exploratory pilot study aimed to investigate the relation between gaze scanning strategies and ability to recognize emotional faces with magnetic resonance imaging (MRI) indicators of possible white matter injury in 12-year-old children born very preterm.Specifically, we wanted to explore if 1. gaze scanning patterns were related to the ability to correctly distinguish similar facial expressions of emotions.
2. gaze scanning patterns and ability to recognize facial emotional expressions were related to lateral ventricle indices, and the area of the posterior corpus callosum.
We also investigated associations with general intelligence as this might affect both affect-recognition ability and visual scanning patterns.
The study comprised 24 children born before 32 gestational weeks in Uppsala county, Sweden, a subset of the LOVIS project cohort [21,22].These children completed the NEPSY-II Affect recognition subtest (mean age=12 years ±4.2 months) and MRI scanning (mean age=13 years ±8.5months).
Ethical approval was obtained (2016/400 and 2019-03473) and the parents provided written informed consent.
The Affect recognition subtest consists of facial photos of children expressing either happiness, sadness, anger, fear, disgust or a neutral expression.The items were presented on a stand according to standardized administration guidelines (Fig. 1A).The child was asked to find two children expressing the same emotion among 3-5 alternatives.The test took 5-10 min, and a Tobii X2-60 eye tracker (Tobii Technology AB, Danderyd, Sweden) recorded the gaze.Number of correctly matched emotional expressions was noted for subtest items 9 through 25 (Affect recognition score), as these items tap visual perceptual affect recognition only.Also, the time from the Affect recognition item presentation to the given answer as indicated by pointing was measured (looking time to answer).
For the gaze scanning pattern analysis, two equally sized areas of interest (AOIs) were placed on each of the photos of emotional faces (Fig. 1), one AOI covering the eyes and the other AOI covering the mouth.All AOIs were kept constant in size (Fig. 1B).
We calculated two ratios: Time ratio -between the time spent looking at the eyes and the mouth AOIs; and Gaze shift ratio -between the number of gaze shifts to the eye and to the mouth AOIs (ratio=eye/ eye+mouth).An eye-to-mouth ratio less than 0.5, either in looking time or gaze shift frequency, indicated more attention to the mouth than to the eyes.
General intelligence was assessed by Full-Scale IQ, using the Wechsler Intelligence Scales for Children, Fifth Edition.
MR imaging was performed using a 3T scanner (Achieva 3 TX, Philips Medical Systems, Best, The Netherlands), without sedation.The imaging protocol included an axial T2 weighted sequence, an axial 3D T1 weighted sequence, a sagittal 3D Flair sequence and a Diffusion weighted imaging sequence [21].
Two radiologists made independent evaluations of the MR images, blinded to the subjects' neonatal characteristics.Metric measurements were used to evaluate lateral ventricle width, corpus callosum area and periventricular white matter thickness.Evans index was calculated as the ratio between the maximum width of the frontal horns and the maximal internal diameter of the skull at the same slice employed in axial MRI (Fig. 2A).Similarly, the maximum width of the occipital horns was related to the maximal internal diameter of the skull (posterior ventricle index, Fig. 2B).Higher Evans and posterior ventricle indices represented larger ventricle sizes.To assess loss of white matter, the mean of the bilaterally thinest white matter part next to the trigone of the lateral ventricles was calculated in mm (Fig. 2C).The size of the corpus callosum was assessed by measurement of the surface area in mm 2 on a mid-sagittal slice (Fig. 2D).The area of posterior corpus callosum was assessed by measuring the surface area of the posterior half when divided into two parts (Fig. 2D).The mean of the MR measures was computed from the two investigators' results, and agreement was assessed with Intraclass Correlation Coefficient and Bland-Altman plot.For more details on MR procedure see Karimi et al. [21].
Study group characteristics are found in Table 1.
Normal data distribution (Shapiro-Wilk test) allowed Pearson correlations between children's task performance and white matter thickness.The findings revealed that a lower posterior ventricle index correlated with reduced looking at the mouth and shorter looking time to answer.There was also a trend towards better Affect recognition score (Fig. 3).A lower Evans index correlated with better affect recognition and more eye focus.A larger area of posterior corpus callosum linked to better affect recognition and higher intelligence.However, intelligence wasn't associated with looking behavior.The bilaterally thinnest white matter measures did not correlate with any functional variables.
To manage familywise multiple comparisons, we applied the Holm-Bonferroni method.This only affected significance levels between the Evans index and both Affect recognition score and Gaze shift ratio to statistical trends.
Taken together, results rendered varying answers for the research questions.First, more focus on the eyes was not related to better Affect recognition score.However, both looking behavior and Affect recognition score were correlated to the MRI findings.Specifically, the larger the ventricles and the smaller the area of the posterior corpus callosum, the more visual focus was put on the mouth, as compared to the eyes, of the emotional faces and the less correct judgements were made.The posterior ventricle index was associated to most functional measures, consistent with the localization of the visual pathways.We found associations between the posterior ventricle index and looking time to answer, looking ratios and a trend to Affect recognition score.
There were associations between the Evans index and Affect recognition score, suggesting relevance for functions dependent on prefrontal  areas such as decision making [23].The area of posterior corpus callosum, relevant for, among others, the visual pathways, was also relevant for the Affect recognition score.General intelligence was only related to the posterior corpus callosum.As expected, the ventricle indices were intercorrelated and larger Evans index was associated with smaller area of posterior corpus callosum.
There was no association between looking pattern and Affect recognition score in this preterm group.In a separate analysis (not included here) of full-term born children (n = 11) performing the same task more focus on the eyes was associated to better Affect recognition score (r = 0.7, p=.017 and r = 0.6, p= .044,for gaze shifts and looking time at the eye area, respectively).However, this group had no MRI assessment.Thus, it is possible that other factors might confound or override the association in the preterm group.
In children born very preterm, it is relevant to measure the posterior parts of the lateral ventricles, as these are adjacent to areas commonly affected by preterm brain injury [24].The posterior ventricle index reflects white matter thickness at the level of the lateral ventricle trigone, and is expected to be sensitive to these particular injuries.Our results show, that for the amount of time devoted to the area of the eyes, this was the most relevant measure.Number of gaze shifts into the eye area and the Affect recognition score (more global function of correctly identifying emotions), were related to the Evans index, but the links were reduced to trends when correcting for multiple comparisons.
The corpus callosum area measured in a midsagittal plane is an indication of a reduction in white matter of the commissural association fibers.As the posterior part of the structure contains, among others, visual connections, it was measured separately.It was related to both Affect recognition score (a visual task) and General intelligence scores.
The Evans and posterior ventricle indices do not directly measure the emotion processing network, where the dorsomedial prefrontal cortex is of special relevance for inferring emotional states [25].Nonetheless, they reflect white matter thickness and our results show that these basic measures capture aspects relevant for looking behavior and identification of emotional expressions in faces.
Thus, a looking strategy including more focus (in looking time and number of gaze shifts) directed towards the eyes, and less time to answer were related to thicker white matter in this group of children born very preterm.These findings can be directly linked to social functioning.Most human activities involve social interactions, which depend on correct information of emotional states of the other person.This information is key to respond appropriately during the interaction.Individuals with difficulties to interpret the facial expressions may miss crucial cues, leading to inadequate responses and heightened social distress.In line with this thought, a smaller posterior corpus callosum was previously associated with increased distress in social situations among very preterm-born adults [26].This in turn effects peer relations and social functioning [27].
Affect recognition is important in theory of mind, i.e. we impute the emotions of others in order to infer mental states [28].In everyday situations, we register how others feel to infer why they feel that way, and how they are going to act next.Frontal brain regions are associated with theory of mind and mental state representation, which is congruent with our findings that the Evans index is associated to proficiency in affect recognition.
This study used a combined method including clinical test scores, registration of gaze strategies and MRI data.In contrast to several previous studies, we disentangled visual perception effects from possible language processing effects, as the children were instructed to perceptually match the similar emotions and not asked to name them.We also measured the looking time to answer as a slower processing time can reveal difficulties as well as effect fluency in interactions negatively.Further studies should focus on the feasibility to give instructions on which areas of a face to attend to, and if this can enhance affect recognition proficiency.
Study limitations include a small sample size and potential power issues.While the brain imaging measures used are crude, the evaluations were performed on slices with standardized angulations, to prevent systematical bias.
The present study expands on previous findings, highlighting the neural basis for development of eye/mouth gaze scanning patterns and performance in affect recognition at 12 years of age in children born very preterm.Although no correlations between looking behavior and Affect recognition score were found, we can state that the posterior ventricle index is a simple measure that have direct links to gaze behavior.This suggests a clinical applicability of the posterior ventricle index.Ventricle enlargement is reflected on a functional level in looking strategies in tasks of social visual perception, and weaker in scores for correctly identified emotions in this group of children born preterm with known probability for white matter damage.

Declaration of competing interest
None.

Fig. 1 .
Fig. 1. [A] Experimental set-up.The Tobii Eye tracker placed under the NEPSY-II Affect recognition subtest.[B] An example of the pictures illustrating how the AOIs were placed.

Fig. 2 .
Fig. 2. MRI measurements.[A] Evans Index= t/T.[B] Posterior ventricle index = y/Y.[C] The bilaterally thinnest white matter part next to the trigone of the lateral ventricles.[D] Area of the corpus callosum (blue and red) and the posterior the corpus callosum (blue).

Fig. 3 .
Fig. 3. Significant correlations between looking patterns (looking time and gaze shifts at/into AOI ratios), affect recognition scores and ventricle indices and area of the posterior corpus callosum.A lower eye-to-mouth ratio indicated more attention to the mouth than to the eyes.*** p<.001, **p<.01,*p<.05, y p<.1.

Table 1
Characteristics of the study group (n = 24).