How Is a Stroke Diagnosed: From Brain Scans to Blood Tests

Stroke diagnosis unfolds as a rapid, layered process that typically begins with a bedside neurological exam, moves within minutes to a non-contrast CT scan of the brain, and then branches into more advanced imaging and laboratory work depending on what the first scan reveals. No single test confirms a stroke in isolation. Instead, clinicians piece together findings from physical assessment, brain imaging, vascular studies, cardiac evaluation, and blood work to determine whether a stroke has occurred, what type it is, and how best to treat it. The speed of this process matters enormously because treatments for the most common stroke type are time-sensitive, often measured in hours.

Prehospital Screening Before You Reach the Hospital

Diagnosis actually starts before the emergency department. Paramedics and first responders use structured screening tools to decide whether someone’s symptoms look like a stroke and how urgently they need a stroke-capable hospital. The most familiar of these is the FAST test (Face drooping, Arm weakness, Speech difficulty, Time to call emergency services), but there are many variations in clinical use. A newer tool called ACT-FAST has shown generally high accuracy in prehospital settings, with retrospective analyses reporting accuracies above 90% and specificity consistently above 85%, though sensitivity has ranged from 63% to 100% depending on study design.¹

These tools are useful but imperfect, especially for milder strokes where symptoms are subtle. A study comparing several prehospital scales in patients with suspected mild or minor stroke found wide variation in performance. The NIHSS (a more detailed clinical scale used in hospitals) caught 95% of strokes but flagged many non-strokes, while the LAPSS scale caught only 42% but was far more specific at ruling out non-stroke cases.² The tradeoff between catching every possible stroke and avoiding false alarms is a real tension in prehospital care. Most systems err on the side of caution: any sudden-onset neurological problem is treated as a potential stroke until imaging says otherwise.

Non-Contrast CT Scan Is the First Image

Once you arrive at the emergency department with suspected stroke, the first scan ordered is almost always a non-contrast CT (computed tomography) of the head. This scan takes just a few minutes and serves a critical purpose: it identifies or rules out bleeding in the brain. This distinction between a hemorrhagic stroke (caused by a ruptured blood vessel) and an ischemic stroke (caused by a clot blocking blood flow) changes treatment completely. Clot-dissolving drugs that can save brain tissue in an ischemic stroke would be catastrophic if given to someone who is bleeding.

A non-contrast CT is excellent at detecting fresh hemorrhage, which appears as a bright white area on the scan. It is much less sensitive to ischemic strokes, especially in the first few hours. Early ischemic changes on CT can be subtle: a slight loss of the normal contrast between gray and white matter, a faint darkening of tissue, or a barely visible swelling. Research has shown that the visibility of these early signs depends on how much time has passed since symptom onset. Hyperdense artery signs (a bright vessel suggesting a clot) tend to become less common after about six hours, while areas of darkening become more prominent as time goes on. Studies that measured the extent of darkening on CT were able to distinguish patients who were still within treatment windows from those who were beyond them.³ The point is that a “normal-looking” CT in the first hours of an ischemic stroke does not mean the patient is fine. It just means the damage has not yet become visible on this particular type of scan.

CT Angiography and Perfusion Imaging

When the non-contrast CT does not show hemorrhage and ischemic stroke is suspected, most stroke centers immediately add two more CT-based scans: CT angiography (CTA) and CT perfusion (CTP). These are done on the same machine, often during the same session, but they answer different questions.

CTA maps the blood vessels in the neck and brain using an injected contrast dye. Its primary job in acute stroke is to find a large vessel occlusion, meaning a clot blocking one of the brain’s major arteries. Finding that clot matters because patients with large vessel occlusions are candidates for thrombectomy, a procedure where a catheter is threaded into the artery to physically retrieve the clot. CTA is fast and widely available, making it the standard tool for identifying who needs this intervention.

CT perfusion goes a step further. It measures how blood is actually flowing through the brain tissue, creating color-coded maps that distinguish between tissue that is already irreversibly damaged (the ischemic core) and tissue that is starving for blood but still salvageable (the penumbra). This distinction is central to treatment decisions, especially for patients who arrive outside the standard treatment window. Automated CT perfusion software is now widely used and has become a key tool in selecting patients for thrombectomy, particularly in delayed time windows of six to twenty-four hours after symptom onset.⁴ The logic is straightforward: if there is still a large area of salvageable brain tissue relative to the dead core, intervention can still help, even many hours later.

Automated CTP software has improved the speed and consistency of these measurements, identifying blood-flow abnormalities that enhance diagnosis of large and medium vessel occlusions when compared with CTA alone.⁵ However, the technical details of how the scan is acquired can influence the results. Research has shown that the standard scan duration of about 48 seconds can sometimes cut off the data before blood flow through the tissue is fully captured, leading to overestimation of the dead core and underestimation of the salvageable penumbra. Extending the scan time to roughly 210 seconds eliminated this problem, resulting in smaller estimated core volumes and larger penumbra volumes.⁶ This technical nuance can change whether a patient is deemed eligible for clot retrieval.

MRI and Diffusion-Weighted Imaging

MRI is the most sensitive and specific imaging tool for diagnosing ischemic stroke, but it takes longer, is not available everywhere around the clock, and requires the patient to lie still in a scanner for an extended period. That combination means MRI is often used as a second-line study when the initial CT-based workup is ambiguous or when more precision is needed.

The workhorse MRI sequence in stroke is diffusion-weighted imaging (DWI). It detects changes in how water molecules move through brain tissue. When cells are starved of oxygen and begin to swell, water movement becomes restricted in a characteristic way that DWI picks up within minutes of stroke onset.⁷ Studies report DWI sensitivity ranging from 88% to 100% with specificity of 95% to 100% for detecting ischemic lesions.⁸ Early studies comparing DWI to older MRI sequences found that DWI detected 98% of ischemic lesions, compared with 71% for standard T2-weighted scans.⁹

DWI is particularly valuable in ambiguous cases: small strokes in the brainstem or posterior brain that CT often misses, strokes that look like other conditions, or situations where the clinical picture is confusing. It is also the tool that defines the modern understanding of transient ischemic attacks, as we will see shortly.

Detecting Hemorrhage and Microbleeds With MRI

While CT is the go-to for detecting acute bleeding, MRI has its own hemorrhage-sensitive sequences. Susceptibility-weighted imaging (SWI) is a specialized MRI technique that is exquisitely sensitive to blood products in the brain. It can detect acute hemorrhage within six hours of symptom onset and is also capable of revealing old, tiny bleeds called cerebral microbleeds that may not show on any other scan.¹⁰

Microbleeds matter for stroke diagnosis and management for several reasons. Their presence can signal underlying small-vessel disease, and in patients being considered for clot-dissolving drugs, a heavy microbleed burden raises concern about bleeding risk. SWI can also distinguish between tiny bleeds and calcium deposits in the brain, which can look similar on other types of scans.¹¹ Longitudinal studies using SWI have found that individual microbleeds can grow over time, which has implications for conditions beyond stroke, including certain forms of dementia.¹²

Telling Stroke Apart From Its Mimics

Somewhere between a quarter and a third of patients who arrive at the emergency department with suspected stroke turn out to have something else. Stroke mimics are conditions that produce sudden neurological symptoms resembling a stroke, and sorting them out is one of the harder parts of stroke diagnosis. The list of mimics is long and includes seizures, severe migraines, low blood sugar, infections, brain tumors, and functional neurological disorders (where the nervous system produces real symptoms without structural damage).¹³

The diagnostic principle is conservative: any sudden-onset, objective, focal neurological deficit should be assumed to be a stroke until proven otherwise. Non-contrast CT is the first imaging step, and MRI is the most sensitive and specific tool for making the distinction. Some mimics leave their own imaging fingerprints. Seizures, for example, can cause distinctive perfusion abnormalities on CT perfusion scans. A study of seizure-related stroke mimics found that about 60% showed cortical hyperperfusion (too much blood flow), while about 40% showed hypoperfusion (too little), but in patterns that crossed normal vascular territory boundaries, unlike typical stroke patterns.¹⁴ Recognizing these atypical patterns prevents unnecessary and potentially harmful treatment.

Transient Ischemic Attacks and the Tissue-Based Definition

A transient ischemic attack, or TIA, has traditionally been defined by symptoms that resolve completely within 24 hours. The modern understanding has shifted. The American Heart Association and American Stroke Association endorsed a tissue-based definition: a TIA is a transient episode of neurological dysfunction caused by focal ischemia, without evidence of acute infarction on imaging.¹⁵ In practice, this means that if someone’s symptoms clear up completely but DWI on MRI shows a fresh area of restricted diffusion (evidence of tissue injury), that event is classified as a stroke rather than a TIA, regardless of how brief the symptoms were.

This distinction is more than academic. Research has shown that patients whose clinical TIA symptoms resolve but whose DWI is positive carry a higher long-term risk of recurrent stroke compared with those who are DWI-negative.¹⁶ Getting an MRI after a TIA, then, is not about being thorough for its own sake. It directly affects how aggressively clinicians pursue prevention strategies.

Looking for the Source of the Clot

Confirming that a stroke has happened is only part of the job. The other half is figuring out why it happened, because the cause determines how to prevent the next one. A substantial portion of ischemic strokes originate from the heart, where abnormal rhythms or structural problems allow clots to form and then travel to the brain.

The standard cardiac workup after an ischemic stroke includes an electrocardiogram (ECG), echocardiography, and cardiac monitoring.¹⁷ Echocardiography comes in two forms. Transthoracic echocardiography (TTE) is the standard initial ultrasound of the heart, useful for assessing heart function in the acute phase. Transesophageal echocardiography (TEE), where the probe goes down the throat to get a closer view, is considered the gold standard for detecting clots in the left atrial appendage, a common hiding spot for emboli in patients with atrial fibrillation.¹⁸

Cardiac monitoring may be brief (a 24-hour Holter monitor) or extended (weeks of monitoring with a wearable or implantable device). The goal is to catch intermittent atrial fibrillation that may not show up on a standard ECG taken at one moment in time. Discovering atrial fibrillation changes management dramatically, typically adding anticoagulation therapy to the prevention plan.

Evaluating the Carotid Arteries

The carotid arteries in the neck are another major source of strokes. Plaque buildup can narrow these arteries severely, and pieces of plaque or clot can break off and lodge in the brain. Carotid evaluation is a routine part of stroke workup, especially for strokes affecting the front of the brain.

Duplex ultrasound (color Doppler sonography) is usually the first test because it is non-invasive, inexpensive, and available at the bedside. It measures blood flow velocity through the carotid arteries; higher velocities mean greater narrowing. MR angiography provides another non-invasive option. A large meta-analysis found that contrast-enhanced MR angiography detected significant carotid narrowing (70% to 99% stenosis) with about 95% sensitivity and 92% specificity, while the older time-of-flight technique achieved roughly 91% sensitivity and 88% specificity.¹⁹ Earlier comparative work found that both MR angiography and color Doppler ultrasound had similarly high sensitivity for detecting 50% or greater narrowing, though ultrasound showed a stronger correlation with the actual degree of narrowing as measured by catheter angiography.²⁰ In practice, clinicians often use two non-invasive tests to confirm significant stenosis before recommending surgery or stenting.

Blood Tests and Emerging Biomarkers

Blood work in acute stroke serves a different purpose than imaging. Standard labs drawn on arrival, including blood glucose, complete blood count, and coagulation tests, are not there to diagnose the stroke itself. They are there to rule out metabolic conditions that can mimic stroke (low blood sugar is the classic example), identify conditions that might make treatment dangerous (such as abnormal clotting), and establish a baseline for the patient’s overall health.

What the field has long wanted, but not yet achieved, is a blood test that could diagnose stroke the way troponin diagnoses a heart attack. Two proteins have emerged as the most promising candidates. S100B and GFAP (glial fibrillary acidic protein) are both released when brain cells are damaged. Reviewed studies highlight their potential as biomarkers with high sensitivity and specificity in differentiating between ischemic and hemorrhagic stroke, and they correlate with stroke severity and long-term outcomes.²¹ GFAP in particular has attracted attention because its levels rise much faster in hemorrhagic stroke than in ischemic stroke, offering a potential way to distinguish the two types before imaging is available. Neither biomarker has yet reached the point of routine clinical use for acute diagnosis, but they are actively being studied in ambulance-based and emergency department settings.

Thrombophilia Testing in Younger Stroke Patients

When a stroke happens in someone under about 50 with no obvious risk factors, the diagnostic workup often expands to include thrombophilia testing. These are blood tests looking for inherited or acquired conditions that make blood clot too easily. The rationale makes intuitive sense, but the evidence is nuanced. A topical review found that inherited thrombophilias have an uncertain relationship with ischemic stroke, while antiphospholipid syndrome, an acquired immune-mediated condition, appears to be a stronger risk factor, especially in young patients.²²

A recent study of young adults with cryptogenic ischemic stroke (meaning no identified cause) found that over a third had some thrombophilia deviation on admission blood tests, although high-risk results remained rare. The authors suggested screening may be worth considering in young patients who also have a history of venous blood clots, physical inactivity, and low hemoglobin or HDL cholesterol.²³ Broadly, thrombophilia testing remains a targeted tool rather than a routine part of every stroke workup.

Artificial Intelligence in Stroke Imaging

AI tools are increasingly embedded in the stroke diagnostic workflow, and in some hospitals they are already running automatically whenever a stroke scan comes through. These systems can detect hemorrhage on plain CT, identify large vessel occlusions on CTA, grade the extent of early ischemic changes, and estimate core and penumbra volumes on perfusion studies.²⁴

A multicenter study evaluating one such AI tool reported about 95% accuracy for detecting hemorrhage on non-contrast CT and about 86% accuracy for identifying large vessel occlusions on CTA.²⁵ These tools are not replacing radiologists. Their main value is speed: they can flag a likely large vessel occlusion and push an alert to the stroke team’s phones within minutes of the scan completing, often before a radiologist has formally reviewed it. In a condition where every minute of delay translates to lost brain tissue, shaving even ten or fifteen minutes off the notification time matters.

Portable MRI and Bedside Diagnostics

One of the biggest bottlenecks in stroke diagnosis is getting the patient to the scanner. Standard MRI machines are expensive, immobile, and often booked. A newer generation of portable, ultra-low-field MRI devices is being tested for use at the bedside or even in ambulances. These machines use much weaker magnets and produce lower-resolution images, but they can go to the patient instead of the other way around.

A pilot study of portable ultra-low-field MRI in acute stroke found that it detected ischemic lesions in eight of twelve patients who had confirmed infarcts on standard high-field MRI. The four missed infarcts were all smaller than 6 mm. In every case studied, a treatment decision made by a blinded clinical team based on the portable MRI alone matched the actual clinical decisions that had been made using conventional imaging.²⁶ This is early-stage work with a small sample, but it points toward a future where brain imaging could begin before the patient reaches the hospital, particularly in rural areas where transport times are long.

Who Gets Left Behind by the Diagnostic System

The sophistication of modern stroke diagnostics, from automated perfusion maps to AI-powered alerts, means very little if you cannot reach a facility that has them. Roughly 20% of the adult population in the contiguous United States, about 49 million people, lives more than a 60-minute drive from a hospital with advanced stroke care capabilities. These areas are overwhelmingly rural. The populations living in those underserved areas already carry higher rates of the very conditions that cause strokes: higher blood pressure, more diabetes, more smoking, and more obesity. The median stroke prevalence in those remote areas is 3.5%, compared with 2.8% in areas with timely access to stroke centers.²⁷

Even within hospitals, access to the full diagnostic toolkit is uneven. A population-based study found that younger patients were substantially more likely to receive advanced neuroimaging compared with older patients, and patients from lower socioeconomic backgrounds were less likely to receive advanced imaging compared with those from higher socioeconomic groups.²⁸ This means that the patients who might benefit most from thorough imaging, older adults and those with more comorbidities, are sometimes the least likely to get it. Telemedicine, mobile stroke units, and portable imaging technologies are all attempts to close this gap, but none has eliminated it yet.

Stroke Diagnosis During Pregnancy

Pregnancy and the postpartum period carry an increased risk of stroke, and diagnosing one during this time involves unique challenges. The symptoms of stroke can overlap with complications like eclampsia and posterior reversible encephalopathy syndrome. Imaging decisions are complicated by the understandable desire to limit radiation exposure, though non-contrast CT is generally considered safe in pregnancy when the clinical situation demands it, and MRI, which uses no radiation, is an attractive alternative when available. The physiological changes of pregnancy, including changes in blood clotting, blood volume, and vessel wall structure, also mean the list of possible stroke causes is different than in the general population, requiring clinicians to think about conditions like cerebral venous thrombosis and vasculopathies specific to pregnancy.²⁹ The diagnostic principles are the same as for anyone else, but the urgency of confirming the diagnosis is compounded by the need to protect two patients at once.