The enormous computing power of machines and their action “without prejudice” could open new avenues both on the prevention and on the therapeutic front
In the largely unexplored field of Alzheimer’s disease, clinicians and artificial intelligence experts have joined forces to seek new openings that help to better understand the causes of this form of dementia and consequently find possible solutions both in terms of prevention and treatment. “There are still large black holes in the interpretation of both etiology and pathogenesis. For this reason, studies based on artificial intelligence can represent small and big steps forward, as regards biological data and those related to risk factors “, explains Professor Marco Trabucchi president of the Italian Association of Psychogeriatrics and scientific director of the Geriatrics Research Group of Brescia.
The publication
Internationally, several groups of researchers are carrying out studies in this field to understand if it is possible to identify the onset of Alzheimer’s using AI. Researchers of the University of Chieti-Pescara, of the University of California-Irvine have recently published a work on the subject on the Journal of Alzheimer’s Disease (HERE the text). The study used a huge international database which collects information on thousands of patients with Alzheimer’s dementia and a machine learning model developed by a team of young Romans from the computer company ASC27. Coordinated by Professor Stefano Sensi director of the Dnisc, Department of Neuroscience, Imaging and Clinical Sciences of the University of Chieti and from the Cast, the Center for Studies and Advanced Technologies, the study focused onweight analyzes that have factors present inside and outside the brain in producing the transition that leads from an initial and potentially treatable condition such as mild cognitive impairment (Mild Cognitive Impairment or Mci) to dementia. And this means that, properly identified, it could be possible to intervene in advance on the risk factors that determine the disease and change its course.
“Benign forgetfulness”
«We tried to use machine learning tools to understand if they can to develop protocols, obviously still experimental, aimed at identifying first those factors that affect the onset and progression of the disease. In the context of dementia there are subjects who are suffering from cognitive deficits with minimal impact on common activities of daily life. In a large part of the cases, fortunately, the situation remains like this for years and it is what it once was called “Benign forgetfulness” and today Mild Cognitive Impairment or MCI. But unfortunately 20 percent of these MCI subjects progress to Alzheimer’s and it is essential to identify in time what it is because it makes them evolve because it is these individuals who could benefit from therapies and prophylaxis »says Professor Sensi.
Big data
He thought about providing the “big data” necessary for exploration Adni (Alzheimer’s Disease Neuroimaging Initiative) which collects data including MRI (magnetic resonance) and PET (positron emission tomography) images, genetics, cognitive testing, cerebrospinal fluid and blood biomarkers as predictors of disease. “This is a huge database that is made available to all by US federally funded Alzheimer’s research centers (by Nih, the National Institutes of Health basically). Each patient is examined extremely rigorously with detailed protocols common to each center that provides the data. In addition to gathering all the classic information on the pathology, Adni offers a whole series of apparently unrelated parameters, that is, parameters that involve everything that happens outside the nervous system ».
Machines have no “prejudices”
To extract useful information from this mare magnum, the group coordinated by Professor Sensi has developed an algorithm which then went on to fathom hundreds of data contained in Adni. «The goal was to try to understand which of these factors had more weight to train the machine in identifying among MCI subjects who was destined to start dementia. The machine knows nothing, luckily for him, and therefore he is completely “unbiased”, that is, devoid of prejudices and theoretical constructs: it takes numbers and manages numbers and therefore also offers us ideas to go beyond the dominant theories, uses other parameters to classify the subjects.
The diagnosis
«And in fact in the training phase we have seen that to arrive at an almost correct diagnosis, that is, with an accuracy between 85 and 97%, the machine did use classic diagnostic parameters but surprisingly showed some associations between extracerebral factors such as the levels of some bile acids and other metabolites and the possibility of development of neurodegenerative processes. This is only partly surprising and actually in line with a number of new evidence pointing to one Gut-Brain connection (the gut-brain axis;
Here an article explaining what it is
ed). In other words, peripheral and affecting alterations of the gastrointestinal system and its microbiome are capable of producing changes in the functioning and well-being of the brain ».
First steps
Of course it is still stammering, first steps far from concrete applications. «The result of our study is that practically we have identified alternative routes or better, possible disease mechanisms that are a little out of the box, that is, of what are precisely the current theoretical schemes and the goodness of this approach is precisely to explore unknown or little explored territories », adds Sensi. “The scenarios are very fascinating. The almost infinite computational power of machines allows large volumes of data to be computed in statistical terms and to produce unexpected inferences and associations. TO
we finally have the opportunity to generate highly innovative hypotheses and to implement a healthy “thinking out of the box” which is always a harbinger of productive epistemological turning points ».
Future scenarios
What could the future scenarios be? «Studies based on artificial intelligence can represent small and big steps forward, as regards biological data and those related to risk factors », echoes Professor Trabucchi. «On the former, the group of Professor Sensi has contributed significantly, as regards the data obtained through neuropsychological detection, magnetic resonance, CSF and blood data. Also with regard to the role of risk factors, the contribution of artificial intelligence could be decisive; in fact, if obtained with traditional epidemiological methods, they have shown as a whole to represent 45% of the risk. However, we do not know their reciprocal interactions and therefore their effects in reality, while artificial intelligence could give us valuable answers, especially for set up preventive interventions, which can be adopted both at the population level and in the single individual with MCI: it is never too late to intervene by modifying one or more risk factors, which could facilitate the transition to dementia ».
January 27, 2022 (change January 27, 2022 | 13:31)
© REPRODUCTION RESERVED
#Alzheimers #big #data #hunting #signs #disease