Preterm birth (gestation age, GA < 37 weeks) now accounts for over 80% around the world, and an estimated 14.9 million babies were born prematurely¹. With the advances in neonatal intensive care, mortality has decreased significantly but prevalence of cerebral palsy and cognitive impairments in survivors increased². The third trimester of gestation is a critical development period for the fetus, the brain undergoes a complex but higly programmed sequence of maturation³. Rapid development of neurons, synapses, axons, and myelin sheath occure during this period, and a concomitant growth of thalamocortical axons are burgeoning. Thalamic neurons are commonly affected in preterm infants particularly in newoborns with white matter lesions⁴. Both morphometric and DTI studies have identified diminished volume of thalamus, altered white matter tracts and reduced fractional anisotropy (FA) in the thalamocortical projections in premature infants or children ^{5, 6}. However, the function thalamocortical connectivity in preterm newborns remains unclear. The aim of this study is to examine the development of the functional thalamocortical connectivity in the preterm without visible lessions (preterm) and preterm with punctate white matter lesions (PWMLs), and test the hypothesis that the thalamocortical connectivity is altered in the preterm especially in preterm with punctate white matter lesions (PWMLs).

PWMLs were defined as preterm brain with focal white matter areas of increased signal intensity on T1 and decreased signal intensity on T2 (< 5 mm). Both preterm infants without visible lesion (preterm) and full-term infants (GA, 37 ~ 42weeks) have no overt abnormal signal in the conventional MRI.

Functional images were acquired from a T2*-weighted EPI sequence (TR/TE: 2000ms/30ms, FOV = 220×220mm², matrix size = 64×64, 33 slices, thickness 3.0 mm, resolution: 3.4×3.4×3.0 mm³). After preprocessing (SPM8), the frame-wise displacement (FD) was controlled to be < 0.5 mm in order to further reduce the effect of motion. Volumes with FD > 0.5 mm were removed and subjects with at least 151 continuous volumes were included in our analysis.

Six well recognized networks were selected in a single group ICA framework using GIFT. Specifically, the auditory, cerebellar, default mode network (DMN), salience (SA), sensorimotor (SM) and medial visual network, and more detail descriptions referent to Fig 1. Spatial mean network maps were threshold with Z > 1.5 (Fig 1). Thalamus mask was defined based on the AAL template (197 voxels). Thalamus parcellation was carried out using partial correlation between the mean time series of each network and that of each voxel in the thalamus controlling for time series from other networks. Then, thalamus cluster was labelled with the mean partial correlation showing the greatest. Partial correlations between each defined thalamus cluster and the cortices were reconstructed controlling for other thalamus clusters⁷. Considering different postmenstrual age (PMA), we added PMA as covariate in the two sample T test. We also attempted to examine the longitudinal thalamocortical connections by further subdividing each group with a more narrow PMA range, specifically preterm (<35 weeks, 35 ~ 36 weeks, 37 ~ 39 weeks), full-term (37 ~ 40 weeks, >40 weeks).

The result of thalamus parcellation was shown in Fig 2. The right portion exhibited the histogram of voxel number in each network. Relatively greater number of voxels located in the SM and cerebella in full-term, while voxels in SA was larger in preterm newborns.

In the reconstruction map of thalamocortical projections (Fig 3), the preterm and PWMLs exhibited limited thalamo-SM connections, while the full-term topography showed more network-like profile, and there seemed no difference between the preterm and PWMLs. Of note, the thalamo-SA connectivity was significantly higher in the 2 preterm groups, and the PWMLs showed a more severe and elevated trend. Not very significant thalamo-cerebellar differences between groups were found. In the two sample T test adding the PMA as covariate, thalamo-SM connections presented limited variance between preterm and full-term groups, while thalamo-SA connections still demonstrated significant differences (Fig 4).

In the longitudinal analysis (Fig 5), most thalamocortical projection topography showed obvious development along time. However, in PMA range between 37 and 39 weeks, both preterm groups showed delayed development in the thalamo-SM connections than full-term newborns of corresponding PMA range, and PWMLs manifested a more delayed development than the preterm. Intriguing, the connection in thalamo-SA became gradual decreasing along time in PWMLs.

Preterm birth might affect the thalamocortical connectivity development particularly in the thalamo-SM and thalamo-SA projections. Reduced thalamo-SM and increased thalamo-SA connectivity were found in the preterm newborns, and preterm with punctate white matter lesions (PWMLs) exhibited a more sever trend in the thalamo-SA projection.