Feed-forward neural networks using cerebral MR spectroscopy and DTI might predict neurodevelopmental outcome in preterm neonates

Abstract

Objectives: We aimed to evaluate the ability of feed-forward neural networks (fNNs) to predict the neurodevelopmental outcome (NDO) of very preterm neonates (VPIs) at 12 months corrected age by using biomarkers of cerebral MR proton spectroscopy (¹H-MRS) and diffusion tensor imaging (DTI) at term-equivalent age (TEA).

Methods: In this prospective study, 300 VPIs born before 32 gestational weeks received an MRI scan at TEA between September 2013 and December 2017. Due to missing or poor-quality spectroscopy data and missing neurodevelopmental tests, 173 VPIs were excluded. Data sets consisting of 103 and 115 VPIs were considered for prediction of motor and cognitive developmental delay, respectively. Five metabolite ratios and two DTI characteristics in six different areas of the brain were evaluated. A feature selection algorithm was developed for receiving a subset of characteristics prevalent for the VPIs with a developmental delay. Finally, the predictors were constructed employing multiple fNNs and fourfold cross-validation.

Results: By employing the constructed fNN predictors, we were able to predict cognitive delays of VPIs with 85.7% sensitivity, 100% specificity, 100% positive predictive value (PPV) and 99.1% negative predictive value (NPV). For the prediction of motor delay, we achieved a sensitivity of 76.9%, a specificity of 98.9%, a PPV of 90.9% and an NPV of 96.7%.

Conclusion: FNNs might be able to predict motor and cognitive development of VPIs at 12 months corrected age when employing biomarkers of cerebral ¹H-MRS and DTI quantified at TEA.

Key points: • A feed-forward neuronal network is a promising tool for outcome prediction in premature infants. • Cerebral proton magnetic resonance spectroscopy and diffusion tensor imaging can be used for the construction of early prognostic biomarkers. • Premature infants that would most benefit from early intervention services can be spotted at the time of optimal neuroplasticity.

Keywords: Diffusion tensor imaging; Magnetic resonance spectroscopy; Neural networks, computer; Neurodevelopmental disorders/diagnosis; Premature infants.

Fig. 1

Flowchart of total study population with excluded and included infants

Fig. 2

Regions of interest (ROIs) in all regions evaluated, positioned in MD (a), FA (b) and MRS localiser (c). Corresponding spectrum in frontal white matter on the right side (d). Frontal white matter right (FWMR) and left (FWML) in blue, central white matter right (CWMR) and left (CWML) in green and parietal white matter right (PWMR) and left (PWML) in orange

Fig. 3

The network diagram represents the NNs employed in this study. We used single-hidden-layer feed-forward neural nets, as they are considered well suited for classification tasks [26]. NNs of this style consist of three layers: the input layer, the hidden layer and the output layer, where each unit of the input and hidden layer connects to each unit of the subsequent layer. By these connections, every unit of the hidden and output layer is a linear combination of all units of the preceding layer, followed by a nonlinear transfer function. The input units x_i correspond to the variables used for the class prediction, in this study metabolite ratios or DTI characteristics. The hidden units can be thought of as new derived variables that are not directly observable in the data. The output units represent the probabilities for the input characteristics to belong to a certain class

Fig. 4

The illustration of the second step of the proposed predictor