AI for early detection and prediction of Alzheimer’s disease
In conjunction with World Alzheimer's Day, GMV reaffirms its dedication to healthcare innovation and research through the application of digital technologies. In cases like that of Alzheimer’s, these technologies are playing a crucial role in uncovering significant insights for improving early detection and prediction of the disease. Since 2016, the multinational company has been engaged in projects such as TARTAGLIA and MOPEAD, both aimed at seeking treatments for this form of dementia. Specifically, within the framework of the TARTAGLIA project, which leverages GMV’s technology to create Spain’s first federated AI data network in the healthcare industry, recent findings in analyzing acoustic properties of communication using AI pave the way for the development of less invasive diagnostic tools, enhancing the ability to detect Alzheimer’s in its early stages or even before symptoms of the disease become apparent.
Alzheimer’s disease stands as the leading cause of dementia, a debilitating condition that, as projected by Alzheimer's Disease International (ADI), will affect over 131.5 million people by 2050. Despite over a century of research, there is still no therapy available to reverse its effects. While early detection and treatment can enhance the chances of delaying its progression, they do not effectively stop its advance.
he World Health Organization has recognized dementia as a top public health priority and is providing funding for research projects aimed at advancing its treatment. In May 2017, the World Health Assembly approved the Global Action Plan on Public Health Response to Dementia 2017-2025. In addition, European Next Generation funds are being allocated to support both healthcare and research projects, as well as the R&D Missions in Artificial Intelligence program as part of Spain's Digital 2025 Agenda and the National Artificial Intelligence Strategy. Notably, among these initiatives, the TARTAGLIA project stands out. This project is centered around the use of artificial intelligence and the development of a federated network encompassing diverse centers focused on identifying early indicators of various medical conditions. One of the lines of work is dedicated to the early detection of Alzheimer's disease, achieved through the analysis of spontaneous speech produced by individuals at various stages of cognitive decline. This research group is led by GMV, with the Ace Alzheimer Center Barcelona serving as responsible for the Alzheimer research group and contributing clinical data, and acceXible, a technology company specialized in spontaneous speech processing.
Meanwhile, in the MOPEAD project, both ACE and GMV have worked on developing an early diagnosis system for the disease involving citizen participation through internet-based social marketing techniques, while also raising awareness in the population about the importance of research in identifying unknown cases.
Artificial intelligence for diagnosis
Neurodegenerative diseases like Alzheimer’s stem from irregularities in certain proteins involved in the cellular cycle. Researchers have identified numerous proteins associated with the disease and linked to inflammation and neurodegenerative processes. Among these, beta-amyloid and Tau proteins are the most widely accepted biomarkers in the scientific community. Disruptions in these protein levels in the patient’s organism appear long before Alzheimer's disease symptoms begin to manifest. Currently, the only way to assess potential protein imbalances is through a lumbar puncture and, in more advanced stages, via MRI scans when the beta-amyloid protein forms brain plaques.
Therefore, within the framework of the TARTAGLIA project, with the goal of discovering less invasive and more cost-effective approaches to detect the disease's onset and broaden the screening possibilities, a collaborative effort was undertaken. This involved a research team from the Ace Alzheimer Center Barcelona, led by Dr. Sergi Valero, alongside technical teams responsible for GMV’s federated network, and specialists from acceXible. They examined eighty-eight acoustic speech factors in dementia patients using artificial intelligence strategies. As Dr. Valero explains, “The results obtained pave the way for enhancing diagnostic tools with non-invasive tests, enabling the creation of a cost-effective and non-invasive physiopathological profile for individuals with dementia. This could facilitate the monitoring of cognitive decline and the identification of at-risk individuals.”
At ACE’s memory clinic, where patients have been diagnosed with mild cognitive impairment, a series of tests were conducted and recorded to examine specific aspects of their speech, often referred to as “spontaneous language”. These assessments included tasks such as describing an image, narrating an experience, or listing a series of animals. Voice recordings were cleaned up and analyzed, with artificial intelligence strategies employed to identify sound characteristics. The analysis of 88 psychoacoustic speech factors (excluding vocabulary and syntax) revealed significant distinctions between patients with a positive amyloid protein marker and those without it. The percentage difference between the two groups was 32+ vs 22.
The findings from this research on predictive models for early detection pave the way for creating less invasive diagnostic tools, enhancing the ability to detect Alzheimer's in its early stages or even before symptoms of the disease become apparent. The added value of this study is its replicability at other centers worldwide, enabling them to delve deeper into understanding the connection between specific speech factors and the presence of the beta-amyloid protein.