Creator: Help Dementia Editorial Team
Published: 2026-04-02T08:24:58

Reviewed by the Help Dementia Editorial Team — our editors review every article for accuracy against guidance from the National Institute on Aging, the Alzheimer’s Association, and peer-reviewed sources.

Researchers developed sits at the center of this dementia and brain health question.

Researchers have successfully developed a groundbreaking method to identify future cognitive decline that costs less than $1 per assessment and requires zero additional clinical time—it works by passively extracting data that’s already in your medical records. A team from the Regenstrief Institute, Indiana University, and Purdue University created this approach using machine learning to identify relevant phrases and clinical information from electronic health records, such as patient notes, blood pressure readings, cholesterol values, and medication history. The method is currently in its final year of a five-year clinical trial conducted in Indianapolis and Miami, with results showing promise for early identification of people at risk for dementia. This article explores how this breakthrough came about, the various artificial intelligence approaches researchers are using to predict cognitive decline, and what these advances mean for early detection and intervention.

How Did Researchers Develop a Zero-Minute Assessment for Dementia Risk?
What Machine Learning Approaches Are Researchers Using to Predict Cognitive Decline?
Understanding Subjective Cognitive Decline as an Early Warning Sign
Beyond Traditional Cognitive Tests – What New Detection Methods Are Emerging?
Accuracy, Limitations, and When These Predictions Apply—and When They Don’t
Why Clinical Trials Matter—The Regenstrief Study in Action
Toward a Personalized, Multi-Method Approach to Cognitive Health
Conclusion

How Did Researchers Develop a Zero-Minute Assessment for Dementia Risk?

The key innovation behind this low-cost method is that it requires nothing new from patients or healthcare providers—no additional testing, no extra appointments, no additional cost to the healthcare system. Instead, machine learning algorithms analyze information that already exists in patient electronic health records. A 55-year-old patient visiting their primary care doctor for a routine blood pressure check generates data points that the algorithm can evaluate: the blood pressure reading itself, cholesterol values from recent labs, current medications, and any notes the doctor wrote about cognitive concerns or memory issues. The algorithm learns to recognize patterns in this existing data that correlate with future cognitive decline, selecting the most relevant clinical signals without requiring new assessments.

This approach addresses a critical barrier to early dementia detection: accessibility. Traditional cognitive screening tests like the Montreal Cognitive Assessment or Mini-Cog require time, training to administer, and patient cooperation. The Regenstrief method bypasses these requirements entirely. Because the data already lives in the medical records system, identifying at-risk individuals happens passively—the healthcare system can implement this screening across entire patient populations without changing clinical workflows. The team is currently validating this method in a diverse population across two cities, testing whether the algorithm’s predictions hold up in real-world medical practice with people of different ages, races, and socioeconomic backgrounds.

How Did Researchers Develop a Zero-Minute Assessment for Dementia Risk?

What Machine Learning Approaches Are Researchers Using to Predict Cognitive Decline?

Beyond the Regenstrief method, researchers are developing several complementary artificial intelligence approaches. Dynamic Bayesian Networks, an advanced machine learning technique published in 2025, can forecast subjective cognitive decline—the experience people report of having memory problems—with personalized risk scoring. This approach is valuable because it doesn’t just give a yes-or-no prediction; instead, it calculates an individual’s personal likelihood based on their specific health profile.

Meanwhile, multimodal deep learning models combine brain MRI scans with clinical data using a technique called Longitudinal-to-Cross-sectional transformation, allowing the algorithm to learn patterns from how brains change over time and apply those insights to individual patients. Random Forest models have achieved particularly strong performance, with an Area Under ROC Curve (a measure of diagnostic accuracy) of 0.773 for predicting cognitive decline within twelve months when using clinical variables and MRI volumetric measurements. However, a critical limitation of the MRI-based approaches is that they require expensive imaging infrastructure and advanced training to interpret—they’re not accessible in many clinical settings, particularly in rural or under-resourced areas. In contrast, the Regenstrief method’s reliance on existing EHR data makes it immediately deployable in nearly any healthcare system that has electronic records, regardless of whether that system has advanced imaging capabilities.

Understanding Subjective Cognitive Decline as an Early Warning Sign

For years, people who reported memory problems but performed normally on cognitive tests were often reassured that nothing was wrong. Research has now established that subjective cognitive decline—when someone experiences and reports memory concerns without objective test findings yet—is actually an early indicator of possible future dementia. This shift in understanding is crucial because it means that people’s own perceptions of cognitive changes shouldn’t be dismissed. A 62-year-old woman who tells her doctor “I’m having more trouble remembering names and I’m worried about it” is reporting something clinically meaningful, even if she scores normally on office screening tests today.

This recognition has transformed the landscape of early detection. Rather than waiting for objective cognitive impairment to appear on clinical testing, researchers and clinicians are now identifying people at the “subjective” stage, when intervention may have the greatest potential impact. The Regenstrief algorithm can identify patients reporting memory concerns in their medical records and flag them as needing further evaluation. Similarly, other emerging methods discussed below—EEG and speech analysis—are being studied specifically for their ability to detect decline at these early stages, potentially catching people before significant cognitive changes become apparent.

Understanding Subjective Cognitive Decline as an Early Warning Sign

Beyond Traditional Cognitive Tests – What New Detection Methods Are Emerging?

Single-channel EEG (electroencephalogram) analysis represents an intriguing alternative to traditional testing. A 2025 study found that analyzing EEG patterns from a single electrode could detect cognitive decline with a sensitivity of 0.90 (meaning it correctly identifies 90 percent of people who are actually experiencing decline) and a specificity of 0.57 (meaning it correctly identifies 57 percent of people who are not experiencing decline). The method uses dimensionality reduction techniques to identify EEG features that correlate with MMSE cognitive scores. The advantage of EEG is that it’s portable and relatively inexpensive compared to MRI; the limitation is that this particular approach has lower specificity, meaning a positive result would need confirmation with another test rather than being definitive on its own.

Speech-based detection represents another frontier. A systematic review from 2025 found that artificial intelligence models analyzing speech patterns show performance comparable to clinical cognitive assessments for detecting cognitive decline. What makes this approach compelling is its potential for integration into everyday technology—theoretically, speech analysis could occur during routine phone calls with healthcare providers without requiring specialized equipment. However, the field is still in early stages, and questions remain about how to implement this in diverse populations, whether different accents or languages affect accuracy, and how to maintain privacy when analyzing speech. Each emerging method has different tradeoffs: EEG requires a clinic visit but is quick and inexpensive; speech analysis is non-intrusive but requires further validation; MRI is highly accurate but expensive and not widely accessible.

Accuracy, Limitations, and When These Predictions Apply—and When They Don’t

An important caveat about all cognitive decline prediction methods: they estimate probability, not destiny. A machine learning model showing that someone has a 65 percent risk of cognitive decline in the next five years means exactly that—there is substantial probability, but also a substantial probability (35 percent) that cognitive decline will not occur. Additionally, these models are only as good as the data they’re trained on. If a model is developed using data primarily from European-American populations, its accuracy may differ when applied to other populations. The Regenstrief team is specifically addressing this by testing their method in diverse populations across multiple cities.

Another limitation applies particularly to imaging-based approaches: detecting brain changes on an MRI doesn’t always mean cognitive decline will become noticeable or problematic. Some people can tolerate significant brain changes with minimal cognitive effects, while others are highly sensitive to smaller changes. This is why researchers are increasingly using multimodal approaches—combining brain imaging with clinical data and other biomarkers—to improve accuracy. A person with brain atrophy on MRI plus cognitive complaints plus certain patterns in their medical records gets a more reliable risk assessment than any single indicator alone. Finally, it’s crucial to understand that these prediction methods identify people at increased risk; they do not yet predict who will develop Alzheimer’s disease specifically versus other causes of cognitive decline or normal aging.

Accuracy, Limitations, and When These Predictions Apply—and When They Don't

Why Clinical Trials Matter—The Regenstrief Study in Action

The five-year clinical trial being conducted by the Regenstrief Institute in Indianapolis and Miami is in its final year, which means researchers should be releasing comprehensive results soon. These trials matter because laboratory validation is different from real-world validation. A model that works perfectly on a research dataset must prove itself when deployed in actual medical systems with actual patients who have varied backgrounds and health conditions.

The Regenstrief trial is specifically testing whether the low-cost algorithm improves outcomes when implemented in busy clinical practices. This type of validation is essential before widespread adoption. Healthcare systems need to know whether the algorithm improves detection, whether it leads to meaningful interventions, whether patients and clinicians find it useful, and whether implementation creates unintended consequences. Because the method is completely automated and low-cost, there’s minimal downside to implementing it broadly, but the trial results will establish the baseline evidence that justifies the approach and guides how it should be integrated into clinical workflows.

Toward a Personalized, Multi-Method Approach to Cognitive Health

The future of cognitive decline detection likely involves integrating multiple approaches rather than relying on a single method. A comprehensive assessment might combine the Regenstrief EHR-based algorithm (to identify people with subtle risk markers), direct questioning about subjective cognitive complaints, modern imaging when clinically indicated, and emerging biomarkers. This multi-method approach reflects a broader shift in dementia science: moving away from waiting for significant cognitive impairment to develop, and instead identifying people with evidence of brain change, subjective concerns, or clinical risk factors years before they notice cognitive problems in daily life.

What makes current research particularly promising is the convergence of these approaches. Every method—whether EHR-based, image-based, EEG, or speech-based—is finding ways to identify at-risk people earlier, with better accuracy, and at lower cost. As these methods continue to be refined and validated, the opportunity increases for earlier intervention, whether through cognitive training, lifestyle modifications, treatment of vascular risk factors, or pharmaceutical interventions as they develop.

Conclusion

Researchers have developed multiple evidence-based approaches to identifying future cognitive decline, with the most immediately implementable being the Regenstrief Institute’s machine learning method that analyzes existing medical record data in less than a dollar and zero additional clinical time. Complementary approaches using machine learning, EEG analysis, speech analysis, and advanced imaging offer different tradeoffs of cost, accessibility, and accuracy. Recognition that subjective cognitive decline itself is meaningful—not something to dismiss—has expanded who should be considered for early assessment and monitoring.

The recognition of cognitive decline before it becomes severe enough to disrupt daily life creates an opportunity for earlier intervention when preventive approaches may have the greatest impact. If you or a family member are concerned about memory changes, discussing these concerns with a healthcare provider is important. The methods discussed in this article are still being refined and validated, but the trajectory is clear: cognitive decline prediction is becoming more accessible, more accurate, and more integrated into routine healthcare.