Prediction of Motor Function in Very Preterm Infants Using Connectome Features and Local Synthetic Instances

Abstract

Abstract. We propose a method to identify preterm infants at highest risk of adverse motor function (identified at 18 months of age) using connectome features from a diffusion tensor image (DTI) acquired shortly after birth. For each full-brain DTI, a connectome is constructed and network features are extracted. After further reducing the dimensionality of the feature vector via PCA, SVM is used to discriminate between normal and abnormal motor scores. We further introduce a novel method to produce realistic synthetic training data in order to reduce the effects of class imbalance. Our method is tested on a dataset of 168 DTIs of 115 very preterm infants, scanned between 27 and 45 weeks post-menstrual age. We show that using our synthesized training data can consistently improve classification accuracy while setting a baseline for this challenging prediction problem. This work presents the first image analysis approach to predicting impairment in motor function in preterm-born infants.

Diagnostic accuracy of early magnetic resonance imaging to determine motor outcomes in infants born preterm: a systematic review and meta‐analysis

“…S1, online supporting information). Twenty‐six studies met the inclusion criteria and five further studies were identified by manual review of references of included papers . Demographic and clinical data extracted from the 31 studies included in this review are presented in Table .…”

“…Seventeen studies from nine distinct cohorts acquired early diffusion MRI data . Sensitivity and specificity of diffusion MRI findings to determine later motor outcomes and/or CP in individual studies are presented in Table .…”

“…Diffusion biomarkers, regions evaluated, and associations with motor outcomes are presented in Table SIII (online supporting information). Three diffusion MRI studies reported diagnostic accuracy of their methods, although the data were not suitable to be combined in a meta‐analysis . Asymmetry of the PLIC on visual inspection of diffusion images, detected cases of hemiplegic CP with a sensitivity of 86% and a specificity of 100% .…”

A comparison of network definitions for detecting sex differences in brain connectivity using Support Vector Machines

“…Hundreds of studies have employed this definition in brain connectomics with the obvious limitation of “diluting” the connectome. The second approach restricted connectivity based on only the endpoints of each fiber such that a single fiber could connect at most two regions and that fiber is counted only once in the final connectome (two-region or endpoint fibers; Figure 3C) [17, 18, 19, 20]. Here, we hypothesized that using two-region (endpoint) fibers would be more informative by avoiding the “overcounting” of connections and focusing on real axonal pathways in the connectome.…”

Characterising brain network topologies: A dynamic analysis approach using heat kernels

“…As an example application, we used heat kernel features for classifying between preterm infants with normal and adverse motor function. MRI-based features such as white matter injury (WMI), intraventricular haemorrhaging (IVH) or diffusion MRI measures of white matter tract integrity from infants scanned near term-equivalent age have been shown to be associated with developmental outcome [69,49,70]. Brown et al classified preterm infants by motor score and showed that a combination of standard global network measures from diffusion tractography, WMI, IVH and GA achieved sensitivity, specificity and accuracy scores of 66, 79, 72.3%, respectively [70].…”