How Alzheimer's disease is diagnosed has been a moving target for a number of years. But the gold standard has been and remains -- wait for it -- the autopsy. But some of us would like to know more before then.
Until 2011 or so, doctors didn't usually pronounce Alzheimer's disease until it was already presenting as actual dementia. Since then, it's become clearer that the disease may begin as much as 20 years before dementia and 15 years before the first symptoms. We call that nominal five years between when symptoms first appear until dementia "mild cognitive impairment" (MCI).
So how is Alzheimer's disease usually diagnosed? And, if the disease is progressing silently for 15 years, can it be diagnosed prior to MCI?
I'm not a doctor, and I don't even play one here. But my read of available information is that a doctor will assess cognitive issues mainly through interviews with relatives combined with cognitive testing. (We discuss this in Beating the Dementia Monster.) Prior to a formal diagnosis, this evidence must be augmented by biomarker evidence. Originally, this might be a brain MRI showing typical atrophy of the hippocampus and expansion of the ventricles. This is how I was assessed. More recently, biomarkers may be different types of PET scans of the brain, blood tests, and analyses of cerebrospinal fluid. But even with these in place, our understanding of the disease is so fluid that we may still have difficulty declaring that a case of MCI or dementia is caused by Alzheimer's disease rather than any of the 30 or more other known causes of dementia.
A goal of some of these diagnostic tools is to identify the disease during that initial 15 years before the first symptoms appear. Some hope that blood tests will reliably do that. But it looks to me like the jury is still out. So there is some subjectivity to the process for the medical professional. What can be done to refine the process and improve its accuracy and our confidence in the diagnosis?
A company called Darmiyan has just received FDA approval of some artificial intelligence (AI) software to do that. The software is called BrainSee, and it uses AI to analyze the same information a radiologist will look at. But it should make a better educated guess. In their clinical trials, they stacked up diagnoses made by live medical professionals looking at MRIs and cognitive test results with those of their AI robot. The result? In summary, the claim is that the clinicians got it right 71% of the time, while BrainSee got it right 83% of the time.
As we said in Beating the Dementia Monster, all brains atrophy with age. But Alzheimer's disease accelerates the atrophy, and the atrophy may occur in different patterns. The claim of the developers is that the AI robot can see specific patterns in a way that's better than a human clinician. Since it's AI, it's up to the robot to decide how to assess the MRI and cognitive testing evidence, so it sounds to me like they really don't understand the details of how the robot gets better results.
Or that's how I read it.
Darmiyan says, "In a side-by-side comparison between BrainSee and hippocampal volume to predict progression to AD-dementia in 409 [Alzheimer's MCI patients], hippocampal volume performed with 65.0 percent sensitivity, 78.2 percent specificity and 71.6 percent balanced accuracy; BrainSee performed with 84.3 percent sensitivity, 81.6 percent specificity and 83.0 percent balanced accuracy."
A lot has changed since my initial diagnosis in 2015. I've changed due to my implementation of the Dementia Toolkit (meaning my brain has changed -- for the better). But the science and diagnostic tools have also changed. Considering my robust brain health, a common question I get is, "Are you sure you have Alzheimer's disease?" I understand the question, but all I know is what the clinicians told me using the evidence they collected in 2015. I wonder what the BrainSee robot would say, given the same information.
No comments:
Post a Comment