I would like to ask 2 questions: 1. How reliable is autism diagnosis if it is based on MRI ? 2. What are the reasons for a sharp increase in people diagnosed with autism?
1. Without getting into all the details, an MRI can be performed several ways, and this will be able to look at the gross anatomy with some detail, and there may be specific structures within the brain's anatomy that would be consistent with some forms of autism, but not all. There are at least 3 known brain morphologies that autistics may have. However, MRI might not be the best way to do this neuroimaging, but rather PET scans which can be useful for showing conductivity and connectivity (the voltages and wires), and this is where autistic brains typically differ from neurotypicals, and it can be quite different. Most of what we know about the autistic brain, within the context of anatomy, is from postmortem studies where researchers can study the brain at the cellular level. It's there that we see the cellular anatomy of the neurons can have too little connections (looking more like a stick) and areas where there might be too many connections (looking like a bush), neither is normal. The densities and layering of neurons are different. The cerebellum structures of autistics are quite telling, as the internal anatomy is markedly different from neurotypicals. Concentrations of glial cells are higher in autistics. The bottom line here is that we need to look at the brain at more the cellular level versus gross anatomy, and currently we are unable to image the brain in that way and say,
"Oh, that's an autistic brain." Google Scholar will have multiple studies on this topic. We are not there yet with diagnosing autism with neuroimaging alone, although it is logical that it will, someday, have its place.
The same thing with genetics, if we can get to a point where we can do an "autism profile" using DNA, this could be helpful, but right now, there are over 100 known DNA markers for autism, another 1000 or so epigenetic markers for autism, and over 2000 RNA markers for autism. So, this gives you an idea of what a daunting task a potential "autism profile" might include and the work that goes into it. Maybe if it can be done quickly with a supercomputer, then maybe, but right now, it can take years to get results back.
2. As others have suggested, awareness of the condition and better screening have contributed to an increase. However, epigenetic risk factors include, (1) more people having children in their 30's and 40's, and less in their 20's (old sperm and egg), (2) older mothers also mean a population with elevated estrogen levels, higher insulin levels, lower thyroid hormone, and longer exposures to environmental toxins, all of which, independently, increase risks. The intrauterine environment appears to have a significant impact. Keep in mind, from an obstetric perspective a pregnant woman who is 35+ is considered "geriatric" and is associated with increased health risks to her and her baby. I know people these days are just starting families in their 30s, but this is NOT how we were designed, as a species. (3) The US is in the midst of an obesity and diabetes epidemic, again, elevated insulin, estrogen, and inflammatory mediators playing key roles here. (4) We, in neonatology, are saving more and more extremely premature infants, currently down to 22-23 weeks' gestation. 20 years ago, this would have never happened. These are basically fetuses. When a pregnancy is interrupted, the babies are exposed to all sorts of environmental insults and stimuli that they never would be in the womb. Inflammation, infections, pain, sensory issues, hormonal alterations, blood flow alterations, and all of this is happening as the brain is developing, and this can lead to abnormal neural pathways and ultimately the whole brain development. Autism in this population is disproportionately high. Within this context, the premature infant model may be considered an "injury model" of autism etiology.