DFC-Igloo: A dynamic functional connectome learning framework for identifying neurodevelopmental biomarkers in very preterm infants

Abstract

Background and objective: Very preterm infants are susceptible to neurodevelopmental impairments, necessitating early detection of prognostic biomarkers for timely intervention. The study aims to explore possible functional biomarkers for very preterm infants at born that relate to their future cognitive and motor development using resting-state fMRI. Prior studies are limited by the sample size and suffer from efficient functional connectome (FC) construction algorithms that can handle the noisy data contained in neonatal time series, leading to equivocal findings. Therefore, we first propose an enhanced functional connectome construction algorithm as a prerequisite step. We then apply the new FC construction algorithm to our large prospective very preterm cohort to explore multi-level neurodevelopmental biomarkers.

Methods: There exists an intrinsic relationship between the structural connectome (SC) and FC, with a notable coupling between the two. This observation implies a putative property of graph signal smoothness on the SC as well. Yet, this property has not been fully exploited for constructing intrinsic dFC. In this study, we proposed an advanced dynamic FC (dFC) learning model, dFC-Igloo, which leveraged SC information to iteratively refine dFC estimations by applying graph signal smoothness to both FC and SC. The model was evaluated on artificial small-world graphs and simulated graph signals.

Results: The proposed model achieved the best and most robust recovery of the ground truth graph across different noise levels and simulated SC pairs from the simulation. The model was further applied to a cohort of very preterm infants from five Neonatal Intensive Care Units, where an enhanced dFC was obtained for each infant. Based on the improved dFC, we identified neurodevelopmental biomarkers for neonates across connectome-wide, regional, and subnetwork scales.

Conclusion: The identified markers correlate with cognitive and motor developmental outcomes, offering insights into early brain development and potential neurodevelopmental challenges.

Keywords: Biomarker detection; Dynamic graph learning; Functional connectome; Very preterm infants.

Figure 1.

Overview of the proposed dynamic functional connectome iterative guided learning via structure connectome (dFC-Igloo) framework. (A) The rs-fMRI data is preprocessed using the developing Human Connectome Project (dHCP) atlas to extract blood–oxygen-level-dependent (BOLD) signals from regions of interest (ROI) for each subject. A sliding window (SW)-approach is used to divide the entire time sequences into temporal intervals. (B) Using the dHCP DTI pipeline, DT tractography is extracted. Fractional anisotropy (FA)-weighted structural connectome SC is obtained as prior knowledge to guide the learning of the FC. (C) The proposed dFC-Igloo framework is a graph learning process that iteratively performs construction and denoising. (D). A close-up of the proposed iterative learning process. In the construction step, an intermediate FC is constructed from the given graph BOLD signals using a graph Laplacian learning. In the denoising step, BOLD signals are denoised, enforcing signal smoothness on the intermediate FC and SC. The iteration stops when the learned FC converges. BOLD signals from all time windows are utilized individually to construct FC, resulting in the formation of a dFC for each subject.

Figure 2.

Visualization of the ground truth FCs constructed from ER, BA, and WS models and reconstructed FCs using our proposed approach and competing methods. (A) the ground truth FC, (B) FC recovered from RBF kernel method (C) FC recovered from graph learning (D) FC recovered from dFC-Igloo framework with medium coupling SC (E) FC recovered from dFC-Igloo framework with high coupling SC. The proposed framework achieved better reconstruction of FC compared with graph learning and RBF kernel-based methods.

Figure 3.

Recovery of the ground truth FC with different noise levels. We compare the recovery performance (values are presented in mean and shade represents standard derivations) over different noise levels of our proposed framework and competing methods with SNR from 1 to 20. The proposed method achieves better performance (recovery accuracy) across noise levels.

Figure 4.

Recovery of the ground truth FC with different coupling SCs. We compared the recovery performance over 10 pairs SC of our proposed framework. Our dFC-Igloo framework observes gradually increased recovery performance as the coupling between the SC and FC becomes stronger.

Figure 5.

Three dimensional structural brain renderings illustrating the locations of the identified ROI level biomarkers for cognitive abilities and motor skills associated with the brain FC, respectively.

Figure 6.

Subnetwork interaction maps between low- and high-risk infants calculated from the averaged network connectivity. The color bar represents the connectivity between subnetworks. For clarification, we also plotted the subnetwork differences for both cognitive and motor outcomes (the increased connectivity in the low-risk group is marked in red and decreased connectivity is marked in blue). MEM=memory network; SAL=salience network; AUD=auditory network; SEM=sematic network; VIS=visual network; SMO=sensorimotor network; ATT=attention network; ECN=executive control network; FPN= frontoparietal network.

Figure 7.

Between subnetworks interaction differences between low- and high-risk cognitive and motor deficits infants. The positive values represent increased inter network interactions in low-risk infants while negative values represent decreased inter network interactions.

Figure 8.

The predicted Bayley-III scores versus the ground truth scores from multivariable regression with nine cognitive regions and six motor regions.