Diagnosing dementia is not a single test but a layered process that typically begins with brief cognitive screening, moves through brain imaging, and increasingly relies on biological markers measured in spinal fluid or blood. No one tool gives a definitive answer on its own. Instead, clinicians piece together results from interviews, pen-and-paper tests, MRI scans, and sometimes specialized PET imaging or fluid biomarkers to determine whether a person has dementia and, if so, which type. The process has changed dramatically in recent years, with blood tests now approaching the accuracy of far more invasive methods.
Cognitive Screening Tests
The first step in most evaluations is a short office-based test of thinking ability. Two tools dominate clinical practice: the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA). Both take roughly ten minutes and probe memory, attention, language, and orientation. The MoCA generally outperforms the MMSE at catching early or subtle impairment. In a head-to-head comparison adjusted for education level, the MoCA achieved about 90% sensitivity and 87% specificity, while the MMSE managed roughly 78% and 77% respectively. Overall diagnostic accuracy for the MoCA reached nearly 88%, compared with about 71% for the MMSE when education-adjusted cutoffs were used.1PubMed Central. Accuracy of the Mini-Mental State Examination and Montreal Cognitive Assessment in Detecting Cognitive Impairment in Older Adults: A Comparative Study Adjusted for Educational Level A study of genetic frontotemporal dementia found a similar pattern, with the MoCA showing stronger ability to distinguish people who were still presymptomatic from those already showing early signs of decline.2PubMed Central. Montreal Cognitive Assessment vs the Mini-Mental State Examination as a Screening Tool for Patients With Genetic Frontotemporal Dementia
Neither test is perfect. Both can be influenced by education, cultural background, and language fluency, and a low score does not automatically mean dementia. Other screening tools exist as well: the Addenbrooke’s Cognitive Examination (ACE-III) and the Rowland Universal Dementia Assessment Scale (RUDAS) both showed high accuracy in one comparative study, with the ACE-III performing slightly better overall than the MoCA for Alzheimer’s screening.3Dementia and Geriatric Cognitive Disorders. Comparative Diagnostic Accuracy of the ACE-III, MIS, MMSE, MoCA, and RUDAS for Screening of Alzheimer Disease The RUDAS was designed specifically to minimize cultural and language bias, which makes it useful in diverse populations where the MMSE or MoCA might underperform. A screening score flags concern. It is not a diagnosis.
The Informant Interview and Clinical History
Screening tests capture a snapshot of how someone performs in a clinic. But dementia is defined by decline from a previous level, and that trajectory is often best described by someone who knows the person well. Tools like the AD8, a brief eight-question informant interview, ask a family member or close friend whether they have noticed changes in judgment, interest in hobbies, ability to handle finances, and other everyday cognitive tasks. AD8 scores correlate strongly with formal neuropsychological testing and with clinical staging scales used to rate dementia severity.4PubMed. Validity and reliability of the AD8 informant interview in dementia
The clinical history also includes a thorough medical review. Clinicians look for the timeline of symptoms, whether decline was sudden or gradual, whether personality or behavior changed before memory did, and whether other conditions could explain what is happening. This matters because a number of treatable conditions mimic dementia. Depression is the most common culprit, but medication side effects, alcohol or drug misuse, thyroid problems, vitamin B-12 deficiency, and even certain brain lesions can produce cognitive symptoms that improve with treatment.5PubMed Central. Reversible dementias Standard guidelines recommend blood work and brain imaging for everyone presenting with cognitive complaints, in part to rule these out before concluding that a neurodegenerative disease is responsible.
Structural Brain Imaging
An MRI scan is the workhorse of dementia imaging. It does not directly detect Alzheimer’s disease or other dementias, but it reveals patterns of brain shrinkage that point toward specific diagnoses. In Alzheimer’s, the hippocampus and surrounding structures in the medial temporal lobe tend to shrink early on. Radiologists rate this shrinkage using visual scales, the most widely used being the medial temporal lobe atrophy (MTA) scale. A cutoff score of 2 or above on that scale shows high accuracy in distinguishing Alzheimer’s patients from healthy controls.6Egyptian Journal of Radiology and Nuclear Medicine. Entorhinal cortex atrophy score versus temporal lobe atrophy score in diagnosis and grading of Alzheimer’s disease The MTA scale has been validated as a reliable marker even in heterogeneous clinical populations, meaning it holds up when patients have a mix of underlying conditions.7PubMed Central. Validity and reliability of the medial temporal lobe atrophy scale in a memory clinic population
One important catch: what counts as “abnormal” atrophy on MRI depends on age, sex, and education. In a large population study of 75-year-olds, the threshold for abnormal MTA differed between men and women and between people with different educational backgrounds. Women with less education had a lower cutoff for abnormality than men or highly educated women.8PubMed Central. Medial temporal lobe atrophy ratings in a large 75-year-old population-based cohort: gender-corrected and education-corrected normative data This means a scan needs to be interpreted in context, not just compared against a single universal number.
MRI also helps distinguish between dementia types. Frontotemporal dementia tends to show shrinkage in the frontal and temporal lobes rather than the hippocampus. Vascular dementia often shows white matter damage or evidence of past strokes. CT scans can serve a similar purpose when MRI is not available or not tolerated, though MRI provides much finer detail.
Metabolic and Molecular PET Scans
While MRI shows structure, PET scans show function and molecular pathology. There are several types, each answering a different question.
FDG-PET measures how actively different brain regions are using glucose. Each major type of dementia produces a recognizable pattern of reduced metabolism. Alzheimer’s disease typically shows decreased activity in the parietal and temporal lobes, while frontotemporal dementia shows reduced metabolism in the frontal regions instead. One multicenter study developed standardized patterns that correctly classified about 95% of Alzheimer’s cases, 92% of dementia with Lewy bodies, and 94% of frontotemporal dementia.9Journal of Nuclear Medicine. Multicenter Standardized 18F-FDG PET Diagnosis of Mild Cognitive Impairment, Alzheimer’s Disease, and Other Dementias FDG-PET is especially useful when the clinical picture is ambiguous, because these metabolic signatures tend to diverge even before structural changes are visible on MRI.10PubMed. Brain [F-18]FDG PET for Clinical Dementia Workup: Differential Diagnosis of Alzheimer’s Disease and Other Types of Dementing Disorders
Amyloid PET uses a tracer that binds to amyloid plaques, one of the hallmark proteins of Alzheimer’s disease. It can confirm or rule out amyloid pathology with high sensitivity. In an autopsy-confirmed study, amyloid PET correctly identified Alzheimer’s neuropathology with about 96% sensitivity, significantly better than FDG-PET’s 80% for the same purpose.11PubMed Central. Diagnostic Accuracy of Amyloid versus FDG PET in Autopsy-Confirmed Dementia A negative amyloid PET scan makes Alzheimer’s disease very unlikely, which can be extremely helpful in redirecting the diagnostic workup.
Tau PET is newer and targets the other major Alzheimer’s protein: tangles of tau. Unlike amyloid, which can accumulate for years without symptoms, tau pathology correlates more closely with where and how severe the cognitive decline is. Both amyloid and tau PET have formal appropriate-use criteria developed by expert workgroups. For amyloid PET, seven clinical scenarios are rated as appropriate, while for tau PET, five scenarios are considered appropriate.12PubMed Central. Updated appropriate use criteria for amyloid and tau PET: A report from the Alzheimer’s Association and Society for Nuclear Medicine and Molecular Imaging Workgroup These scans are not ordered routinely. They tend to be reserved for diagnostically uncertain cases or to guide eligibility for certain treatments.
Cerebrospinal Fluid Biomarkers
Before blood tests emerged, measuring Alzheimer’s proteins in cerebrospinal fluid (CSF) collected by lumbar puncture was the main way to check for biological evidence of disease. The core CSF biomarkers are amyloid-beta 42 (which drops when amyloid is accumulating in the brain), phosphorylated tau (which rises with tangle formation), and total tau (which reflects general nerve cell damage). Using ratios of these markers rather than any single one improves accuracy. The ratio of amyloid-beta 42 to phosphorylated tau is better than amyloid-beta 42 alone, and it agrees more closely with amyloid PET results.13PubMed Central. Clinical application of CSF biomarkers for Alzheimer’s disease: From rationale to ratios In one clinical cohort, combining low amyloid-beta 42 with a low amyloid/phospho-tau ratio yielded about 78% sensitivity and 83% specificity for Alzheimer’s disease versus healthy controls.14PubMed Central. Cerebrospinal fluid biomarkers in Alzheimer’s disease: Diagnostic accuracy and prediction of dementia
CSF testing is also expanding beyond Alzheimer’s. A technique called alpha-synuclein seed amplification assay (SAA) detects the misfolded protein that characterizes Lewy body diseases. In one study of people clinically diagnosed with dementia with Lewy bodies, about 72% tested positive on the alpha-synuclein SAA, compared with only 4% of controls.15PubMed Central. Association of CSF α-Synuclein Seeding Amplification Assay Results With Clinical Features of Possible and Probable Dementia With Lewy Bodies People who tested positive also showed worse motor symptoms and more pronounced smell loss, two hallmarks of Lewy body pathology, and experienced more progressive motor decline over time.16PubMed Central. CSF α-Synuclein Seed Amplification Assays and Alzheimer Disease Biomarkers in Dementia With Lewy Bodies: Presentation and Progression The test is not yet routine in all clinics, but it represents a genuine step toward molecular-level identification of Lewy body disease during life rather than after death.
Blood Biomarkers
The most significant recent shift in dementia diagnosis is the move toward blood-based biomarkers. A lumbar puncture is uncomfortable, and PET scans are expensive and not widely available. A blood draw, on the other hand, is something any clinic can do. The leading candidate is plasma phosphorylated tau 217 (p-tau217). Across multiple independent cohorts, p-tau217 predicted abnormal amyloid buildup with an accuracy above 0.92 on a 0-to-1 scale, and predicted abnormal tau pathology with similar or even higher accuracy.17JAMA Neurology. Diagnostic Accuracy of a Plasma Phosphorylated Tau 217 Immunoassay for Alzheimer Disease Pathology These numbers approach what CSF markers and PET scans achieve, which is remarkable for a simple blood test.
Another blood marker, neurofilament light chain (NfL), works differently. It does not point to Alzheimer’s specifically. Instead, it signals that nerve cells are being damaged, regardless of the cause. That makes it useful as a general indicator of neurodegeneration. Higher baseline NfL levels in blood are associated with worse cognitive performance across domains including memory, language, and executive function, and they predict faster cognitive decline over time in people with neurodegenerative conditions.18PubMed Central. Plasma neurofilament light chain as prognostic marker of cognitive decline in neurodegenerative diseases, a clinical setting study Because NfL is elevated in many conditions, from Alzheimer’s to frontotemporal dementia to multiple sclerosis, it cannot tell you which disease is present. But a normal NfL level in someone with cognitive complaints offers some reassurance that rapid neurodegeneration is not underway.19PubMed Central. Plasma Neurofilament Light Chain in Patients Affected by Alzheimer’s Disease with Different Rate of Progression: A Retrospective Study on an ADNI Cohort
In practice, p-tau217 and NfL answer complementary questions. The first asks, “Is Alzheimer’s pathology present?” The second asks, “How much brain damage is happening right now?” Used together with clinical information, they can sketch a surprisingly detailed biological picture from a few tubes of blood.
The ATN Framework
All of these biomarkers need a way to be organized and interpreted. The field has adopted a system called A/T/N, where “A” stands for amyloid pathology, “T” stands for tau pathology, and “N” stands for neurodegeneration or neuronal injury. Each category is rated as positive or negative based on whatever biomarker is available, whether that is a PET scan, CSF test, blood test, or MRI measurement. A person’s biological profile might be written as A+/T+/N+, meaning amyloid plaques are present, tau tangles are present, and nerve cell damage is detectable.20PubMed Central. A/T/N: An unbiased descriptive classification scheme for Alzheimer disease biomarkers The NIA-AA research framework uses this system to define Alzheimer’s disease biologically rather than purely by symptoms, grouping imaging and fluid biomarkers together by the pathological process each one measures.21PubMed Central. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease
This shift matters because someone can have Alzheimer’s biology in their brain (A+/T+) without yet showing obvious symptoms. Conversely, someone with clear cognitive decline might be A-negative, meaning their symptoms come from something other than Alzheimer’s. The framework forces clinicians to separate the question “What is wrong with this person’s brain?” from the question “What symptoms are they experiencing?” Answering both gives a much sharper diagnosis.
Telling Dementia Types Apart
Dementia is not one disease. Alzheimer’s disease is the most common cause, but dementia with Lewy bodies, vascular dementia, frontotemporal dementia, and several rarer conditions each require different management and carry different prognoses. Getting the type right matters, not just for treatment planning but because some medications appropriate for one type can worsen another.
Vascular dementia often results from accumulated damage caused by strokes or chronic small vessel disease in the brain. Even a relatively small stroke in a critical area, such as a speech-related region or specific white matter pathways, can cause disproportionate cognitive effects.22PubMed. Emerging Concepts in Vascular Dementia: A Review MRI is especially helpful here, because it can show infarcts, white matter lesions, and other vascular damage that point to a vascular contribution. In many older adults, Alzheimer’s pathology and vascular damage coexist, making the picture more complicated than a single label suggests.
The accuracy of clinical diagnosis alone, without biomarker confirmation, is less impressive than many people assume. One autopsy study found that the clinical diagnosis of Alzheimer’s was correct only about 78% of the time overall, and in cases with a short disease duration, clinical accuracy dropped to roughly 52% while a biomarker-based test achieved 100% accuracy against autopsy findings.23PubMed. Early diagnostic accuracy and pathophysiologic relevance of an autopsy-confirmed Alzheimer’s disease peripheral biomarker For dementia with Lewy bodies, clinical criteria for “probable DLB” yielded about 73% sensitivity and 93% specificity when checked against neuropathology, meaning a confident clinical diagnosis was usually correct but a substantial number of true cases were missed.24PubMed. Accuracy of Clinical Diagnosis of Dementia with Lewy Bodies versus Neuropathology These numbers are the core argument for adding biomarkers to the diagnostic process whenever possible.
Genetic Testing in Dementia Workups
Genetics occasionally plays a role in diagnosing dementia, though its place differs from the biomarkers discussed above. The APOE gene, specifically the ε4 variant, is the best-known genetic risk factor for late-onset Alzheimer’s. Carrying one copy roughly triples the risk, and carrying two copies raises it further. But APOE status alone is not diagnostic. Many ε4 carriers never develop Alzheimer’s, and many Alzheimer’s patients do not carry the variant. In memory clinic patients, both APOE genotype and a broader polygenic risk score (a composite measure of many small genetic contributions) were independently associated with Alzheimer’s disease.25PubMed Central. Genetic testing of common and rare variants in dementia patients from a memory clinic Genetic testing can help estimate risk, identify rare hereditary forms of dementia, or clarify ambiguous cases, but it is not a standalone diagnostic tool for typical late-onset disease.
Voice Analysis and Digital Tools
One of the more unexpected frontiers in dementia detection involves listening to how people talk. Cognitive impairment changes speech in measurable ways: word-finding becomes harder, sentences get simpler, the overall coherence of what someone says tends to drop. Automated speech analysis tools can pick up on these changes, sometimes before standard screening tests flag a problem. One study found significant differences in word-finding difficulty, information content, global coherence, and syntactic complexity between cognitively healthy people and those with mild cognitive impairment.26PubMed Central. Using Digital Speech Assessments to Detect Early Signs of Cognitive Impairment
More recent work has pushed this further using artificial intelligence. In a study analyzing voice recordings from a story-recall task, a machine learning model that combined acoustic and linguistic features detected cognitive impairment with very high accuracy, and a deep learning model incorporating a large language model performed even better on a held-out test set.27PubMed Central. Voiceprints of cognitive impairment: analyzing digital voice for early detection of Alzheimer’s and related dementias These tools are still in the research phase, but the appeal is obvious: they could enable screening through a phone call or tablet app, without requiring a clinic visit at all. For populations with limited access to specialists, that kind of scalability could be transformative.
Disparities in Who Gets Diagnosed
The advances in biomarker testing and imaging carry a risk: they could widen existing gaps in who actually receives an accurate, timely dementia diagnosis. In many healthcare settings, access to PET scans, lumbar punctures, or even specialized memory clinics is uneven. Rural communities, racial and ethnic minorities, and people with lower socioeconomic status are already less likely to receive a dementia diagnosis, and the arrival of expensive new technologies threatens to deepen that divide.28Alzheimer’s & Dementia: Behavior & Socioeconomics of Aging. Mitigating disparities in diagnosis of dementia Blood-based biomarkers like p-tau217 have the potential to narrow the gap, precisely because a blood draw is cheap and universally available. But the assays themselves are still being rolled out, and insurance coverage remains inconsistent.
The Ethics of Early Biomarker Results
As biomarkers improve, they are increasingly able to detect Alzheimer’s pathology in people who have no symptoms yet. This raises a difficult question: should you tell someone they have amyloid plaques building up in their brain years before they might develop dementia? Evidence from disclosure studies suggests that the psychological risk of sharing this information is low, and that many people find it beneficial for planning and for exercising personal autonomy over future decisions.29PubMed Central. Disclosure of preclinical Alzheimer’s disease biomarker results in research and clinical settings: Why, how, and what we still need to know However, that evidence comes mostly from well-educated, high-socioeconomic-status, and predominantly White study samples, so how broadly it applies is uncertain.
The situation is even more complex for genetic forms of dementia like familial frontotemporal dementia, where a positive biomarker result might predict onset with uncomfortable certainty. Disclosure involves balancing the relief of resolved uncertainty and the chance to plan against the psychological burden of knowing what is coming. Autonomy, actionability, and social impact all factor into these conversations.30PubMed Central. Ethics of disclosure of onset-predictive biomarker test results for genetic frontotemporal dementia in the research context As disease-modifying treatments for Alzheimer’s have begun to reach the market, the argument for early detection has strengthened, because knowing early could mean intervening early. But the conversation between a clinician and a patient about what a preclinical biomarker result means, and what it does not mean, remains one of the more delicate in medicine.