Convolutional Neural Network-Based Regression for Quantification of Brain Characteristics Using MRI
Preterm birth is connected to impairments and altered brain growth. Compared to their term born peers, preterm infants have a higher risk of behavioral and cognitive problems since most part of their brain development is in extra-uterine conditions. This paper presents different deep learning approaches with the objective of quantifying the volumes of 8 brain tissues and 5 other image-based descriptors that quantify the state of brain development. Two datasets were used: one with 86 MR brain images of patients around 30 weeks PMA and the other with 153 patients around 40 weeks PMA. Two approaches were evaluated: (1) using the full image as 3D input and (2) using multiple image slices as 3D input, both achieving promising results. A second study, using a dataset of MR brain images of rats, was also performed to assess the performance of this method with other brains. A 2D approach was used to estimate the volumes of 3 rat brain tissues.
KeywordsPreterm infants Rat brain Brain quantification Deep learning Convolutional neural networks Regression Magnetic resonance imaging
This work was supported by COMPETE: POCI-01-0145-FEDER-007043 and FCT – Fundação para a Ciência e Tecnologia within the Project Scope: UID/CEC/00319/2013. We gratefully acknowledge the support of the NVIDIA Corporation with their donation of a Quadro P6000 board used in this research.
- 4.Petrasek, T., et al.: A rat model of Alzheimer’s disease based on Abeta42 and pro-oxidative substances exhibits cognitive deficit and alterations in glutamatergic and cholinergic neurotransmitter systems. Front. Aging Neurosci. 8(83), 1–12 (2016)Google Scholar
- 7.Moeskops, P., et al.: Prediction of cognitive and motor outcome of preterm infants based on automatic quantitative descriptors from neonatal MR brain images. Sci. Rep. 7(2163) (2017)Google Scholar
- 8.Kaggle’s competition: Second Annual Data Science Bowl. https://www.kaggle.com/c/second-annual-data-science-bowl. Accessed 26 Nov 2018
- 9.Liao, F., Chen, X., Hu, X., Song, S.: Estimation of the volume of the left ventricle from MRI images using deep neural networks. arXiv:1702.03833v1 (2017)
- 10.Dubost, F., et al.: 3D regression neural network for the quantification of enlarged perivascular spaces in brain MRI. arXiv:1802.05914v1 (2018)
- 11.de Vos, B.D., Lessmann, N., de Jong, P.A., Viergever, M.A., Isgum, I.: Direct coronary artery calcium scoring in low-dose chest CT using deep learning analysis. In: 103rd Annual Meeting Radiological Society of North America (2017)Google Scholar
- 13.Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556v6 (2015)
- 14.Theano Development Team: Theano: A Python framework for fast computation of mathematical expressions. ArXiv e-prints, vol. abs/1605.02688 (2016)Google Scholar
- 15.He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for Image Recognition. arXiv:151203385v1 (2015)
- 16.Magalhães, R., et al.: The dynamics of stress: a longitudinal MRI study of rat brain structure and connectome. Mol. Psychiatry, 1–9 (2017)Google Scholar