Prediction Of Age And Intelligence Based On Elastic NET

With the help of magnetic resonance imaging technology,such as Magnetic resonance imaging(MRI),diffusion tensor imaging(DTI),and functional magnetic resonance imaging(fMRI),we can obtain the structural and functional information of human brain noninvasively.Machine learning techniques allows us to acquire an"empirical model" based on the magnetic resonance images and cognitive indexes of known samples,and then quantitatively evaluate the cognitive parameters of the unknown samples based on the magnetic resonance images of unknown samples.That means it is possible to assess the individual physical and cognitive indicators objectively and accurately if we use the neuroimaging technology and machine learning technology reasonably.Elastic net(E-Net)is an effective regression analysis method.By imposing a sparsity requirement on the data space,E-Net can directly obtain predictive models that are easy to interpret without the need of feature selection beforehand.In this study,we performed individuals' brain age estimation and intelligence quotient estimation using E-Net based on fMRI data and DTI data.Main works are as follows:(1)We performed brain age estimation using the resting state fMRI(RS-fMRI).Based on the resting-state functional connectivities(RSFCs)between 160 regions of interest(ROIs)evaluated on the RS-fMRI data of 63 adult subjects aged 18 to 45 years,we builded the brain age prediction model using E-Net.The result of leave one out cross validation shows that the correlation between the estimated and chronological age was 0.78.(2)We performed brain age estimation using RS-fMRI data.Based on the fractional anisotropy,mean diffusivity,axial diffusivity and radial diffusivity extracted from the DTI data of 111 adults aged 18 to 55 years,we builded the brain age prediction model using E-Net.The result of leave one out cross validation shows that the correlation between the estimated and chronological age was 0.87.(3)We predicted intelligence quotient(IQ)estimation using with the same method.Based on the resting-state functional connectivities(RSFCs)between 160 regions of interest(ROIs)evaluated on the RS-fMRI data of 76 adult subjects aged 18 to 55 years,we builded the IQ prediction model using E-Net.The result of leave one out cross validation shows that the correlation between the estimated and chronological IQ was 0.64.The significances of this research are:1)we found that the functional and structural alterations of adults' brains are sufficiently specific to decode individuals'ages.These results highlighted the necessity of careful considerations of the influences of these changes in related studies.2)Based on MRI we preliminary explored the IQ estimation,and we verified the feasibility of predicting IQ based on RS-fMRI.