Research | Funded Projects

EARLY NEURAL CONNECTIVITY DEVELOPMENT: A PRENATAL FUNCTIONAL MAGNETIC RESONANCE IMAGING STUDY

Study Design: Cross-sectional study on neural network features at different gestational ages (from 20th to 40th gestational week) by means of resting state functional MRI. Eventual subsequent qualitative/quantitative comparison with prematures of the same postmenstrual age. Pregnant women requiring fetal MRI for CMV/Toxoplasma infection or other maternal risk factors such as gestational hypertension, suspicion of brain abnormalities at ultrasound evaluation and preterm newborn will be enrolled. After obtaining mother's informed consent, each fetus will be investigated by means of a 1.5T MR scanner (Intera, Philips). We estimate that about 100 fetuses and 50 preterms will be enrolled in the 2-year study period. Among them about the half will present excessive motion artefacts to perform fMRI analysis (personal experience). The total MRI study duration will be approximately 30 min, which includes anatomical and functional imaging. Clinical data of newborns and findings at 1 year follow-up will be recorded. Fetuses and preterms that will present neurological involvement will be excluded from fMRI analysis. A resting state fMRI scan entailed 320 continuous functional volumes (TR=1500msec, TE=60msec, flip angle=90°, 15-18 slices, matrix=128×128, voxel size=3×3×4mm; FOV 300×300mm, acquisition-time=8min). For spatial normalization and localization, high-resolution fast-spin-echo T2 weighted anatomical images are acquired (TR=7419msec, TE=80msec, flip angle= 90°, ETL=165, voxel size=5×1×1mm, matrix 340×342mm, FOV=340×340mm).Resting-state scans are preprocessed using both AFNI_2010_10_19_1028; http://afni.nimh.nih.gov/afni and FSL 4.1.6; http://www.fmrib.ox.ac.uk. In a preliminary study, we dealt with motion correction by following a series of steps that included both automatic and manual processing of images (i.e. manual tracing of masks to reduce maternal/fetal body interference, manual reorientation of images, removal of volumes with the larger-scale movements, motion correction using Fourier interpolation). Subsequent preprocessing include spatial smoothing, mean-based intensity normalization, linear and quadratic detrending. A high-pass filter (>0.005 Hz) is used to reduce very-low frequency artefact such as scanner draft and a low-pass filter (<0.1 Hz) to remove any components in the high-frequency spectrum. Finally, we perform spatial normalization via estimation of a linear transformation from the individual functional space to a standard brain space using each individual's T2 high-resolution anatomic image. For spatial normalization we use an age-specific atlas of preterm newborns according to gestational age (Kuklisova).Temporal-concatenation group Independent Component Analysis (ICA) is used in the final sample to generate group-level components of the dataset using MELODIC (FSL). The number of dimensions is estimated using the Laplace approximation to the Bayesian evidence of the model order. ICA maps are thresholded at z>2.30.