The great topmost sheet of the mass, that where hardly a light had twinkled or moved, becomes now a sparkling field of rhythmic flashing points with trains of traveling sparks hurrying hither and thither. The brain is waking and with it the mind is returning. It is as if the Milky Way entered upon some cosmic dance. Swiftly the head mass becomes an enchanted loom where millions of flashing shuttles weave a dissolving pattern, always a meaningful pattern though never an abiding one; a shifting harmony of subpatterns.
Few scientists write like that anymore. Science is the poorer for it. Sherrington was able to dash that off when the most complicated device to provide an analogy was a 1801 Jacquard loom capable of weaving complicated patterns on the basis of punched cards. So, Sherrington thought of a loom, Freud of a hydraulic system of pipes, Pavlov of a telephone exchange, and more recently everyone and their uncle think of the brain as a computer.
As we progress, the capacity to image the brain, and image it in motion, improves rapidly. Now we can look at fMRI outputs, and deduce the fine detail of connectivity in the brain. A problem arises which the authors of a recent paper identify as a source of bias: even at rest, these outputs differ according to race.
Here is an interesting study which looks at the patterns detectable in brain activity, and which identifies significant racial differences.
Evidence For Bias Of Genetic Ancestry In Resting State Functional MRI. April 2019
Conference: IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019), 8-11 April 2019, Venice, Italy Volume: 16. Andre Altmann and Janaina Mourão-Miranda
Resting state functional magnetic resonance imaging (rs-fMRI) is a popular imaging modality for mapping the functional connectivity of the brain. Rs-fMRI is, just like other neuro-imaging modalities, subject to a series of technical and subject level biases that change the inferred connectivity pat-tern. In this work we predicted genetic ancestry from rs-fMRI connectivity data at very high performance (area under the ROC curve of 0.93). Thereby, we demonstrated that genetic ancestry is encoded in the functional connectivity pattern of the brain at rest. Consequently, genetic ancestry constitutes a bias that should be accounted for in the analysis of rs-fMRI data.
Resting state is what the subject does when lying in a scanner without any task to attend to. Of course, resting state can be very busy: thinking of nothing is impossible, so subjects may be worrying about claustrophobia, enumerating their daily tasks and duties, mulling over a recent problem, or idly considering the sensual advantages of polymorphous perversity. I digress, and no digression is ever fully restful.
Although age and sex and health status can affect functional MRI readings “there is one additional subject level characteristic that is known to affect head and brain morphology but is rarely considered as a confound in rs-fMRI studies or brain imaging studies in general: genetic ancestry.”
They took 1003 subjects from the Human Connectome Project with the relevant fMRI data, and for 950 of those for they also had genomic data. They ended up with 764 European, 138 African, and 39 Asian subjects. To increase discriminative power they compared the Europeans against the rest (using various cut-off assumptions).
They use signal detection theory, which entered psychology in the 1960s, and made “Receiver Operating Characteristics” and “Areas Under the ROC curve” familiar to researchers.
It is a tiny matter, but the same analysis of functional MRI which classifies race at 93% accuracy classifies sex at 98% accuracy. Both these categories can be detected on brain waves in the form of minute blood flows.
So, the functional connectivity of the brain differs among racial groups defined by their genomes, and this difference can be picked up with a high level of accuracy (93%).
The exact origin of these apparent connectivity differences between continental ancestries remains elusive at the moment. However, we hypothesize that the observed differences are not based on true neuronal differences but that they originate from differences in head and brain morphology as reported in [8, 9]. These morphological differences may be carried forward through the standard rs-fMRI processing pipeline and affect the inferred functional connectivity. In addition, rs-fMRI connectivity is based on correlations between blood-oxygen-level dependent (BOLD) signal time series at rest. Thus, it is conceivable that genetic differences contributing to blood circulation, perfusion and elasticity of the vascular system may modify BOLD dynamics. This is exemplified by reports identifying ethnicity as independent risk factors for cardio blood oxygenation level dependent vascular disease and intracranial artery tortuosity. In addition, brain hemodynamic responses are known to be heritable traits.
So, the authors think that it is racial difference in skull and brain shape which may account for these differences, which make it easy to detect the subject’s race from the connectivity of the brain, rather than any neuronal differences.
Perhaps they are keeping this for later work, but subjects in the Human Connectome Project have completed intelligence tests, and will also have a measure of MRI derived brain size.
Previous 2015 work on the Human Connectome Project shows that brain networks are associated with intelligence.
Previous 2017 work on the Human Connectome Project shows that brain networks are associated with intelligence.
Previous 2018 work on Human Connectome Project likewise.
So, it would be possible to see if the differences in connectivity related in any way to intelligence in the three groups studied: Europeans, Africans and Asians. If there were no differences this would tend to disprove assumptions about brain size and brain organization as a source of racial difference in intelligence.
This paper has a very large sample, employs standard measures and has appropriate statistics, cautiously interpreted. All these aspects are reassuring. It is a niggle, but the authors use the notion of “bias” in a particular way, because they have identified human differences which are nothing to do with bias. These differences were detected by very precise measures of brain activity, thus revealing something, not obscuring it. By analogy, their mapping of the sea floor has shown previously unrecognized under-water contours and ships can now navigate more safely.
Like any other study, this needs to be replicated, but this is a significant finding which stands until refuted by a study of similar sample size and uniformly applied measures.