Abstract #4629

Early prediction of neurodevelopmental deficits in very preterm infants using a multi-task deep transfer learning model

Lili He^1,2,3, Hailong Li^1,3, Jinghua Wang⁴, Ming Chen^1,5, Jonathan R. Dillman^4,6, and Nehal Parikh^1,2

¹The Perinatal Institute and Section of Neonatology, Perinatal and Pulmonary Biology, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States, ²Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH, United States, ³Imaging Research Center, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States, ⁴Department of Radiology, University of Cincinnati College of Medicine, Cincinnati, OH, United States, ⁵Department of Electronic Engineering and Computing Science, University of Cincinnati, Cincinnati, OH, United States, ⁶Department of Radiology, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States

We proposed a deep transfer learning model using the fusion of clinical and brain functional connectome data obtained at term for early neurodevelopmental deficits prediction at two years corrected age in very preterm infants. The proposed model was first trained in an unsupervised fashion using 884 subjects from publicly available ABIDE repository, then fine-tuned and cross-validated on 33 very preterm infants. Our model achieved an AUC of 0.77, 0.63 and 0.74 on the risk stratification of cognitive, language and motor deficits, respectively. Our findings demonstrated the feasibility of using deep transfer learning on connectome data for abnormal neurodevelopment prediction.

This abstract and the presentation materials are available to members only; a login is required.

Join Here