Monday, September 14, 2015

Identifying poor outcomes in autism as early as possible

A recent study on autism [1] was featured in the popular journal Science [2] and presented functional brain imaging patterns associated with autism subjects who have poor language outcome. Early identification of autistic subjects with poor outcomes is an important research avenue with the potential to assist in informing patient care by steering treatment decisions, observing disease progression and monitoring therapeutic response. It is encouraging to see such considerable potential from functional magnetic resonance imaging (MRI) of the brain and it should also be noted that MRI is capable of acquiring a wide variety of different physiological measurements. I was fortunate to get a short letter/comment published with the journal in online format discussing the potential to further characterize autism and to potentially identify those subjects with poorer outcomes using pH sensitive MRI [3] and machine learning technologies. I introduce the reader to this topic here:


It is known that patients with autism often exhibit abnormally reduced cerebral blood flow [4-5]. Since oxygen is delivered through the blood stream, this cerebral blood flow reduction may result in an inadequate oxygen supply to the affected region(s) of the brain. This effect may be more pronounced in autism subjects with poor outcome as inadequate blood flow leads to inadequate oxygen delivery which is critical to healthy brain function. Furthermore, cells rely on adenosine triphosphate (ATP) for standard metabolic activity. ATP is produced through standard aerobic processes which rely heavily on a supply of oxygen. When that oxygen supply is greatly reduced, affected cells must rely on anaerobic processes to produce ATP which yields lactic acid as a byproduct. This may cause tissue acidosis detectable by MRI (through chemical exchange saturation transfer imaging focused on amide proton transfer - an emerging imaging method which is pH sensitive). Assessment of tissue pH may also shed light on an important stage in autistic brain development. MRI is capable of providing many different physiological measurements that may be relevant to identifying poor outcomes in autistic subjects such as pH sensitive MRI, functional MRI, structural MRI and blood perfusion MRI. Multivariate pattern analysis technologies such as machine learning have considerable potential towards combining the predictive capacities of a variety of measurements in order to predict poor outcomes in autistic subjects more accurately than any of the individual MRI based imaging methods could accomplish alone. It is hoped that a combination of new biomarkers for autism characterization (such as pH sensitive MRI) combined with pre-existing measurements using machine learning technologies will be useful in the assessment of autism spectrum disorders.

References: 
[1] M V Lombardo, et al., Neuron 86(2):567-577 (2015). 
[2] P J Hines, Catching it early. Science 348(6233):409 (2015) 
[3] J Zhou, et al., Methods Mol Biol 711:227-237 (2011). 
[4] T Ohnishi, et al., Brain 123(9):1838-1844 (2000). 
[5] W H Yang, et al., Chin Med J (Engl) 124(9):1362-6 (2011).

Letter/comment from which this article is based:
J. Levman, E. Takahashi, “Identifying poor outcomes in autism as early as possible,” Science DOI:10.1126/science.348.6233.409-f Comment (2015).