A recent study published by the Alzheimer’s Association sheds light on a new early detection method that could help save millions of lives. The study examines the use of AI in automated prediction protocols combined with voice scanning methods to accurately determine the progression from Mild Cognitive Impairment (MCI) to Alzheimer’s Disease (AD) within a 6-year window. Here’s what you need to know.
Alzheimer’s & Dementia
Anyone who has lived with or helped a relative with Alzheimer’s & dementia can attest to the difficulties of the situation. Notably, Alzheimer’s Disease is the primary cause of dementia globally. It’s a progressive brain disorder that deteriorates a person’s mental capabilities over time. The process is gradual, with symptoms growing from simple forgetfulness to memory loss, disorientation, self-neglect, and a loss of speech. Notably, patients who suffer from MCI have a much higher chance of becoming infected with Alzheimer’s Disease.
Alzheimer’s Disease Facts
Alzheimer’s Disease is the 6th largest killer of Americans and remains a major concern globally. A 2023 study found that +6.7M Americans suffer from this disease, which can leave patients unable to care for themselves, recognize their family members, or conduct necessary day-to-day tasks.
Source – Wikipedia
Sadly, the research also pointed to increased risks in the future. It predicted 13.8M US citizens suffering from the disease by 2060. These factors have led many researchers to seek solutions and prevention methods to help reduce these risks, potentially saving many lives.
AD Pathology
AD pathology is a term that references how the brain begins to decay under the weight of the disease. Today’s methods require conducting neuropsychological tests (NPTs), which allows healthcare professionals to monitor levels of beta-amyloid peptides (Aβ), hyperphosphorylated tau protein, Reactive astrocytes, microglial activation, and other biomarkers. All of these factors can indicate chronic inflammation and other issues associated with the onset of AD.
Current Limitations of AD Pathology
These methods of determining a patient’s AD risk have some significant drawbacks. For one, they are expensive. Each test can cost thousands, making it only available to those with insurance or access to funding. Additionally, these methods are often invasive. Anytime you have to cut into the skin, the risk of infections and other complications rises.
The Future of AD Detection is Voice-based
Researchers have noticed a direct correlation between speech decline and MCI. As such, there has been a lot of work to document and record these changes for research purposes. Today, thousands of hours of voice recordings from neuropsychological exams are available. Additionally, other data is attached to these recordings. This FHS study is the largest set of recordings.
Framingham Heart Study (FHS)
The Framingham Heart Study (FHS) has been underway since 1948. This long-term study of cardiovascular health researched thousands of men who had strokes or heart attacks. Its long-term approach has enabled the researchers to create 10-30-year risk-prediction models which have proved vital in prevention, creating medication, and now programming AI models.
This collaborative effort has support from the National Heart, Lung, and Blood Institute (NHLBI) and Boston University. and is one of the longest and most successful cardiovascular studies ever conducted. As such, its data has helped countless medical breakthroughs and remains a powerful source of long-term health data.
As part of the study, speech tests were introduced to the data in 2005. Since then, scientists have recorded thousands of speech patterns in patients before and during AD. These recordings provided valuable insight into how the mind decays during AD.
Artificial Intelligence Enhances Diagnosis Methods
The AI revolution has enabled engineers to create more advanced and efficient early warning systems. In this instance, AI Natural Language Processing (NLP) was used to diagnose linguistic features that signal cognitive decline. The algorithm can cross-reference millions of data points to accurately place patients on a dementia scale based on their communications.
Alzheimer’s & Dementia Study
The study applied NLP techniques with machine learning methods to enhance predicted models of MCI and dementia. The team took test interview recordings from n + 166 Framingham Heart Study (FHS) participants to create these models. The models covered a range of options, from demographics to a full spectrum of data.
One main difference from previous attempts was that the team focused primarily on the MCI-to-AD progression rather than normal cognition. It included 90 progressive MCI and 76 stable MCI cases as part of the approach. This data was fed into an advanced AI algorithm that examined key factors.
Input Data
The system used a variation of the logistic regression model that leveraged sub-test scores and TAS as its main prediction indicators. As part of this approach, the team first created a transcribing tool that streamlined the import of previously recorded Framingham Heart Study (FHS) data.
Speech Data
The speech data from the 166 participants included an hour-long interview with each person. This data had been previously recorded and stored digitally, which meant the team needed to create methods to easily convert this data into text and then, AI models.
The Universal Sentence Encoder
The universal Sentence Encoder was designed to reduce the workload of importing FHS data. The system automatically scans FHS data and generates vector embeddings based on specific segments of each patient’s transcript. This approach helps the AI to make more precise predictions while lowering workloads.
AI Model Improvements
The research team used these transcripts to create a powerful AI deep learning algorithm. The system automatically calls random data from the FHS report and samples it. This data was used to create and encode sub-test content. In total, eight specific embeddings were used to gauge AD progression and MCI.
Logistic Regression Models
New AI models trained from the quantitative data associated with each sub-test content were created. The content was scored and used as individual inputs for modeling. These models were then broken down and tested for effectiveness resulting in a (TAS) transcript average score for each. These scores and TAS enabled AI to accurately monitor MCI progression.
The primary model included a host of valuable data on participants. Specifically, it included demographics, APOE carrier status, health factors, and text features imported from the FHS study. This model scored the highest, showing nearly 80% accuracy in determining AD onset within six years.
The team also tested a model that simply took into account demographic data on participants. There was also a text-only model that leveraged the FHS text inputs as its primary review source. Notably, the demographic AI model predicted AD with 68.8% accuracy. Impressively, the most accurate model, which included all the relevant data, achieved 79% accuracy.
Testing the Models
Each Model’s performance was tested using a group k-fold cross-validation approach that examined key factors including accuracy, precision, sensitivity, and specificity. Also, cross-validation and feature selection tests were conducted.
Results
The results of the study are promising. The new AI detection model had a 79% accuracy in determining the onset of AD within 6 years. The team found that the more data inputs added the more accurately the AI could determine a person’s risk. As such, speech data, age, sex, and education level are all factors that play a role in a person’s risk of exposure to AD.
Benefits of the Alzheimer’s Disease Study
There are some serious benefits that this study could provide the average person. For one, AI-based prediction methods are far more affordable and accessible to the public. The cost of today’s current tests within the for-profit healthcare system continues to rise do to the need to use laboratories and samples. The new AI system eliminates these needs, greatly reducing costs for all participants.
Additionally, integrating an automated detection system can help doctors and healthcare professionals more accurately and quickly prescribe treatments and medications when possible. These improved responses could make a life-and-death difference in many instances.
Non-Invasive
This AI method for detecting AD is non-invasive. There is no need for doctors to take tissue samples to determine your AD progression, reducing the risk of infection or other complications.
Companies that May Benefit from this Study
Several companies could improve their products and services by integrating this new automated AI AD detection method. These companies specialize in AD prevention and prediction methods. As such, they could better prescribe and diagnose patients, leading to faster response times. Here are two companies that are positioned to integrate this tech today.
1. Biogen 
Biogen is a leading biotech firm based out of Massachusetts, USA. The company has a variety of products designed to prevent, treat, and help assist those suffering from neurological and autoimmune diseases. Integrating an automated AD detection model such as the one created by the researchers would significantly improve the company’s response time.
Biogen stock is considered a stronghold due to its history and positioning in the market. The company has seen slow and steady gains over the last four years and currently has a market cap of $33.16B. Analysts predict the manufacturer will see further revenue increases due to its pioneering efforts and increased demand for early prevention methods.
2. Eisai 
Tokyo-based Eisai remains a global force in the medical research and development sector. The company specializes in the import and export of pharmaceuticals and has collaborated with other leading brands, including Biogen, on an Alzheimer’s Disease drug pact. The pact enables these firms to share data freely with the hopes of making a significant breakthrough in the diagnosis and treatment of this ailment.
Eisai is a leader in the market and has helped research and develop some very successful pharmaceuticals. The company has strong support from the investment community and has been featured in the Argus U.S. ESG model portfolio. This stock recommendation is based on environment, governance, and social criteria. For these reasons, ESALY is a wise addition to any portfolio.
Future of AI Healthcare
This study is one of many that have captured the attention of those suffering from Alzheimer’s. Soon, there may be automated and very affordable at-home diagnostic tools that would enable anyone, from anywhere, to monitor their health risk. These devices wouldn’t need to be restricted to simply Alzheimer’s prevention. Here are some other cool AI medical enhancements.
AI Stroke Detection
AI systems are now in place to help detect strokes. For emergency personnel, determining if a person is suffering from a stroke or other ailment is often the first step in saving their lives. Notably, a team of Biomedical Engineers from RMIT University created a new tool that leverages AI to identify a CVA in seconds. The software uses a smartphone camera to view and determine via their facial features if a stroke occurred.
AI-Assisted Speech
AI-assisted speech is another area of healthcare that has seen significant improvement. These systems use large language models to help those who are unable to pronounce words due to accidents or impediments create sentences in real-time. This technology is already helping thousands of people who suffer from communication problems overcome adversity and live more fulfilling lives.
AI-Assisted Movement
AI systems have also helped improve mobility for millions of people. Researchers are now using AI to help improve exoskeleton performance. These devices, which used to require a remote control for the user to operate, can now determine your next move and automatically help you to climb stairs, steep slopes, and other terrain that would usually be impossible for other walking assistance systems.
AI Helping Beat Alzheimer’s Disease
AI tech can help those who suffer from Alzheimer’s Disease keep track of their fading memories and live better lives. This latest development demonstrates how AI prevention and diagnosis could be the first steps towards finally beating this ailment. As such, this research provides a new level of confidence while opening the door for a healthier and more cognitively active population in the future.
Click for more Cool AI Developments.