The human brain is the most complex object known in the Universe. At least for exactly whom is able to know – and the only known entities known to be that able are humans, at least as the current common knowledge asserts. It is a complex entangled web of organic tissue composed of an incredible diversity of cells connected in numerous ways, communicating with each other with still more complex processes. It is quite a remarkable machine.
But the human brain, like all other tissues in the human body, decays with age. The older brain does not have the performance of the younger one. Despite this, this astonishing organ displays another characteristic that may compensate for the effects of age: it is plastic, this meaning that it can change and explore in on itself different network configurations and change how their networks are connected. At least up to a point. On the counterpart to that capacity it is the recognition that plasticity is not always synonym of better performance; it just that the right set of combinations and circumstances afford increasing performance with a highly plastic brain.
With this said I present here another remarkable research paper found and reviewed by the MIT Technology Review weekly paper round-up. It is accordingly about the determination of the human brain age using the current state-of-the-art developments within deep learning research and techniques. It is fascinating read, just the review, but the review itself definitely encourages, open the appetite to read the full paper. Interestingly one outcome of the paper is how deep learning techniques beats a traditional statistical method known as Gaussian process regression in determining brain age, by a wide margin, with increased speed and minimal requirements.
Predicting Brain Age with Deep Learning from Raw Imaging Data Results in a Reliable and Heritable Biomarker
Machine learning analysis of neuroimaging data can accurately predict chronological age in healthy people and deviations from healthy brain ageing have been associated with cognitive impairment and disease. Here we sought to further establish the credentials of ‘brainpredicted age’ as a biomarker of individual differences in the brain ageing process, using a predictive modelling approach based on deep learning, and specifically convolutional neural networks (CNN), and applied to both pre-processed and raw T1-weighted MRI data (…)
From the MIT critical review we read:
At the same time, the team compared the deep learning approach to the conventional method of determining brain age. This requires extensive image processing to identify, among other things, white matter and gray matter in the brain followed by a statistical analysis called Gaussian process regression.
The results make for interesting reading. Both deep learning and Gaussian process regression accurately determine the chronological age of patients when given preprocessed data to analyze. Both methods do this with an error of less than five years either way.
However, deep learning shows its clear superiority when analyzing raw MRI data, where it performs just as well, giving the correct age with a mean error of 4.66 years. By contrast, the standard method of Gaussian process regression performs poorly in this test, giving a rough age with a mean error of almost 12 years.
What’s more, the deep-learning analysis takes just a few seconds compared with the 24 hours of pre-processing required for the standard method. The only data processing required for the deep-learning machine is to ensure the consistency of the image orientation and the voxel dimensions between images.
Of note is also the confirmation of links between early brain age with conditions such as diabetes, schizophrenia and traumatic brain injury. So this is a promising result for clinicians of brain health, providing these professionals with one more diagnostic tool that may help them reach a better decision for treatment prescription.
Featured and Body text images: Predicting brain age with deep learning from raw imaging data results in a reliable and heritable biomarker