A multi-task, multi-stage deep transfer learning model for early prediction of neurodevelopment in very preterm infants

Abstract

Survivors following very premature birth (i.e., ≤ 32 weeks gestational age) remain at high risk for neurodevelopmental impairments. Recent advances in deep learning techniques have made it possible to aid the early diagnosis and prognosis of neurodevelopmental deficits. Deep learning models typically require training on large datasets, and unfortunately, large neuroimaging datasets with clinical outcome annotations are typically limited, especially in neonates. Transfer learning represents an important step to solve the fundamental problem of insufficient training data in deep learning. In this work, we developed a multi-task, multi-stage deep transfer learning framework using the fusion of brain connectome and clinical data for early joint prediction of multiple abnormal neurodevelopmental (cognitive, language and motor) outcomes at 2 years corrected age in very preterm infants. The proposed framework maximizes the value of both available annotated and non-annotated data in model training by performing both supervised and unsupervised learning. We first pre-trained a deep neural network prototype in a supervised fashion using 884 older children and adult subjects, and then re-trained this prototype using 291 neonatal subjects without supervision. Finally, we fine-tuned and validated the pre-trained model using 33 preterm infants. Our proposed model identified very preterm infants at high-risk for cognitive, language, and motor deficits at 2 years corrected age with an area under the receiver operating characteristic curve of 0.86, 0.66 and 0.84, respectively. Employing such a deep learning model, once externally validated, may facilitate risk stratification at term-equivalent age for early identification of long-term neurodevelopmental deficits and targeted early interventions to improve clinical outcomes in very preterm infants.

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1

An overview of proposed multi-task, multi-stage deep transfer learning framework for early joint prediction of neurodevelopmental outcomes at 2-years corrected age in very preterm infants.

Figure 2

Detailed model architectures in source domain (A), intermediate domain (B), and target domain (C). Layer types are color-coded. Purple: input layer; green: fully connected layer; orange: batch normalization layer; yellow: output layer. The number of nodes in each layer are listed below individual layers. Red color dotted boxes highlight the transferred/reused neural network layers between consecutive domains.

Figure 3

Top discriminative functional connectomes explored by our multi-task, multi-stage deep transfer learning model for each of the three neurodevelopmental outcomes. (A) The top 20 functional connections were displayed in the circos plot. (B) Overlapped connections were listed. We found 9 common connections predictive to all three outcomes (red), 12 common connections predictive to both language and motor outcomes (green, red); 14 common connections predictive to both cognitive and language outcomes (blue, red); and 9 common connections predictive to both cognitive and motor outcomes (covered by red).