ACCORDING to the World Health Organization (WHO), an estimated 57 million people are living with dementia globally, with Alzheimer’s accounting for roughly 60 to 70 percent of cases. Nearly 10 million new cases are diagnosed every year, and as populations age, that number is expected to climb sharply in the coming decades.
Despite its growing toll, the disease is often diagnosed only after irreversible brain damage has occurred, limiting the effectiveness of emerging therapies.
Until recently, confirming Alzheimer’s required expensive positron emission tomography (PET) scans or invasive spinal taps to analyze cerebrospinal fluid. Blood-based testing offers a far cheaper, less invasive alternative — and as disease-modifying therapies become available, identifying patients at the earliest stages matters more than ever, since these drugs work best before widespread brain damage sets in.
Now scientists are combining blood-based biomarkers, artificial intelligence and digital health technologies, including artificial intelligence to detect the earliest biological signs of the debilitating disease, potentially transforming diagnosis from a reactive process into a proactive one, focused on prevention and early intervention.
It’s in the blood
Among the most promising advances are blood tests that measure biomarkers such as phosphorylated tau (p-tau217) and beta-amyloid — molecular fingerprints that accumulate in the brain long before symptoms appear. There are three significant studies connected to this technology.
One of the key breakthroughs in Alzheimer’s diagnostics was the identification of phosphorylated tau 217 (p-tau217) as a reliable blood biomarker of the disease. Researchers at Washington University School of Medicine, led by Dr. Randall Bateman and Dr. David Holtzman, were among those who helped pioneer the technology that would later underpin blood-based Alzheimer’s testing.
The research was subsequently validated by an international study published in the Journal of the American Medical Association (JAMA). Led by Gemma Salvadó of Lund University’s Clinical Memory Research Unit in Malmö, Sweden, the study followed 786 participants and found that plasma p-tau217 testing was as accurate as cerebrospinal fluid analysis for diagnosing Alzheimer’s disease. The researchers also reported that the blood test reduced the need for confirmatory PET brain scans by about 80 percent.
Building on those findings, Washington University researchers later developed a method that uses a single blood test to estimate when an individual is likely to begin showing symptoms of Alzheimer’s disease, offering physicians a way to predict disease onset years before clinical signs become apparent.
The technology has since moved into clinical practice through C2N Diagnostics, a Washington University spinout that commercialized the PrecivityAD2 blood test. Based on the work of Bateman, Holtzman and their colleagues, the test is among the leading commercially available assays that measure p-tau217 and other biomarkers to help physicians evaluate patients for Alzheimer’s disease. These tests could eventually become part of routine clinical care, helping physicians identify patients when newly approved drugs are most likely to slow disease progression.
Biomarkers for cancer
The same approach is reshaping oncology through liquid biopsies. Unlike conventional biopsies, which require tissue samples taken directly from tumors, liquid biopsies analyze a blood sample for circulating tumor DNA (ctDNA) and other cancer-derived biomarkers — molecular fragments that can signal cancer growth well before tumors appear on imaging or cause symptoms.
Researchers at institutions such as Johns Hopkins, along with companies including GRAIL and Exact Sciences, believe such tests could eventually detect multiple cancers earlier than current screening methods, improving survival by catching tumors while they remain small and localized. The same biomarker research is applied into cardiovascular disease, inflammatory disorders, metabolic illnesses and other neurodegenerative conditions. By identifying combinations of biomarkers that form disease-specific signatures doctors can make estimations of future health risks and guiding personalized treatment.
Artificial intelligence joins diagnostic arsenal
Artificial intelligence has emerged as a powerful complementary tool. Researchers at Mass General Brigham recently found that AI can detect signs of Alzheimer’s simply by analyzing how a person speaks.
The team developed two machine learning models to evaluate voice recordings from participants in the Longitudinal Early-onset Alzheimer’s Disease Study — 120 people with cognitive impairment and 68 cognitively healthy volunteers, all of whom underwent amyloid PET imaging to confirm whether Alzheimer’s or another disorder was the cause. The more advanced model identified mild cognitive impairment with roughly 99 percent accuracy and distinguished Alzheimer’s-related decline from other causes with up to 90 percent accuracy.
Rather than looking for obvious memory lapses, the AI picked up on subtle linguistic patterns that often escape human notice: patients with early Alzheimer’s tended to omit key details, use fewer proper names and descriptive words, forget dates and times, pause more often, and rely on shorter, less detailed sentences even while speaking longer. These changes reflect disruptions in memory, attention and executive function that begin years before dementia becomes clinically obvious.
The findings carry real weight because early-onset Alzheimer’s is frequently missed — researchers estimate up to 90 percent of cases go undiagnosed in primary care, where symptoms are mistaken for stress, depression, fatigue or normal aging.
Hospitals already use ambient AI systems to transcribe physician-patient conversations for documentation. Researchers envision future versions of these systems simultaneously scanning speech for digital biomarkers of cognitive decline, turning routine visits into screening opportunities without extra exams or equipment. Lead investigator Dr. Neguine Rezaii describes language as a digital biomarker reflecting the health of the brain itself; if validated in larger studies, speech analysis could become an inexpensive, scalable way to flag patients for further neurological testing.
Despite the excitement, experts caution that this science is still evolving. Many proposed blood tests require even larger clinical studies to show they consistently improve outcomes while minimizing false positives and negatives. Physicians will also need evidence-based guidelines for when to order additional testing and how to interpret increasingly complex molecular data.
