Maybe artificial intelligence will help us diagnose Alzheimer’s soon and he will be able to do it by analyzing human writing. IBM (International Business Technology Corporation) and Pfizer (American Multinational Pharmaceutical Corporation) have developed a model of artificial intelligence that can detect early signs of the disease in humans by studying language patterns in their vocabulary.
So far, many types of models have been developed by other researchers to detect cognitive disorders, including Alzheimer’s symptoms. The scientists based these models on data types such as brain scans and clinical trials. But, this latter study is particularly important because it draws on historical information from many generations of Fremingham’s cardiological studies. These studies have been conducted on more than 14,000 people of three different generations since 1948. Researchers think that if the AI model’s ability to detect any particular trend in such data could be used to study larger and more diverse populations, they would be able to predict the development of Alzheimer’s in humans a few years before the symptoms of the disease become so noticeable. Become possible by traditional methods. This method of diagnosis does not require invasive tests or scans. The results of a Pfizer-funded and then IBM-funded study were published in EClinicalMedicine on Thursday last week.
“The new AI model helps practitioners see how unnoticed changes in humans can be detected before they are diagnosed clinically. In fact, these unobtrusive processes may be a warning that other significant changes in the human body are expected in the future and are more accurate in predicting them.” “Research is needed,” said Ejei Royuru, IBM, vice president of health and life sciences research.
To improve the models created, the researchers looked at digital transcripts of handwritten responses from Fremening Cardiologists, who were asked to verbally describe a photograph of a woman washing dishes and two children with biscuits on her back. “This model does not perceive human handwriting when observing descriptions,” said Roda Au, director of neurophysiology at Fremingham Research and a professor at Boston University. (His team was responsible for creating transcripts of Mozambique, but on the other hand did not participate in the study). Still, even without the physical handwriting, IBM explains that their main AI model was able to detect and detect linguistic features in the text, which is often an early sign of the development of cognitive deficits. These include spelling mistakes, word repetition, and the use of simple phrases instead of complex grammatical sentences. “These signs are consistent with doctors’ perceptions of how Alzheimer’s can affect a person’s speech,” Royuru said.
The main model was able to predict with 70% accuracy which of the Fremingham participants under the age of 85 developed Alzheimer’s. Of course, this result was also obtained based on the study of historical data.
“The AI model used data from the earliest group of Framingham Cardiologists, most of whom are white people of non-Hispanic descent. This, in part, prevents us from imagining how these results change in more diverse groups in the United States and elsewhere.” Notes au. “It’s still unclear how this model will work in large populations: EClinicalMedicine’s study found data on only 40 people who eventually developed the disease and another 40 people who did not develop Alzheimer’s symptoms,” said Jekaterina Novikova, director of the Laboratory of Mechanical Engineering. Novikova, who did not participate in the new study, also questions the ability of IBM’s AI model to predict Alzheimer’s development with equal accuracy before diagnosis at different points in time.
Nevertheless, he and Au believe that the paper will make a major contribution to the development of this field and that the discovery of Alzheimer’s AI model with the help of new research may attract more attention and resources. “Personally, I like the fact that this paper is one of the few studies based on the analysis of large-scale, diverse and long-term real-time data collection,” Novikova said.
“If the new model could perceive the manuscript, researchers would be able to study it with greater precision,” explains Au. With this skill it would be possible to find additional information, e.g.