Blood leaves a permanent mark. When a small vessel inside the brain tears, the escaped red cells break down over weeks into hemosiderin, an iron-storage compound, and the iron stays where it settled. Years later it is still in the tissue.
Susceptibility-weighted imaging exists to find that iron. It is the third modality in our series on the eight imaging types used to evaluate [mild traumatic brain injury](condition-concussion-and-mtbi.html), and it asks one narrow question: did small vessels bleed, and where.
How Does Susceptibility-Weighted Imaging Work?
Most MRI sequences image water, the hydrogen in soft tissue. SWI images something different: how strongly a substance distorts the magnetic field it sits in. That property is magnetic susceptibility, and iron distorts a field far more than water does. SWI is a gradient-echo sequence combining magnitude and phase information to amplify those differences rather than average them away.
The consequence is useful. A high-susceptibility deposit does not merely appear dark; it **blooms**, perturbing the field around itself so it renders larger than it is. A deposit too small to resolve becomes visible through the shadow it casts.
That brings a short list into view: blood breakdown products, small veins, and iron deposits. Standard T1 and T2 sequences do not reliably show any of them, which is part of why [a normal-looking scan does not settle the question](post-what-imaging-shows-after-a-head-injury-and-what-it-misses.html).
What Is SWI Looking For After a Head Injury?
Microhemorrhages — the small-vessel bleeding that marks traumatic vascular injury. On the image they are punctate: dots, sometimes short streaks along a vessel. Old or new, they persist, which is why an SWI performed months later still registers what happened.
Location carries most of the interpretive weight. Three regions matter.
**The gray-white junction.** Gray and white matter differ in density, so in a rotational acceleration injury they decelerate at different rates and the boundary between them becomes a shear plane — the field's working account of traumatic axonal injury (Bruggeman et al., 2021). Small vessels crossing that boundary take the same shear as the axons.
**The deep nuclei.** The basal ganglia and thalamus sit at interfaces of the same kind, deeper in.
**The brainstem.** The most rotationally stressed structure in the head, where a small lesion buys the largest symptom burden.
Bruggeman and colleagues wrote a narrative review, not an accuracy study, so read that mechanism as a model rather than a measured probability. It still matters: it explains why microbleeds in those three places mean something different from microbleeds anywhere else.
Can SWI Find Anything When Everything Else Is Normal?
That question decides whether the sequence is worth ordering, and one study answers it. Huang and colleagues imaged 111 patients with mTBI **without parenchymal hemorrhage on CT and conventional MRI** (Huang et al., 2015). Everyone in the cohort had already been scanned and told there was nothing there.
SWI found microbleeds in 26 of them. It also found microbleeds in 12 control subjects (p = 0.0197) — the most important number in the paper, and one I will come back to.
Where the groups separated sharply was distribution. In cortical and subcortical regions, the mTBI group showed 52 microbleeds, 86.7%, against 3 microbleeds, 20%, in controls (p < 0.0001). The trauma pattern was not simply *more* — it was *elsewhere*, where the shear model predicts.
It also tracked with function. Patients with mTBI who had detected microbleeds had lower digit span scores than patients with negative SWMRI findings (p = 0.017). Digit span measures working memory and attention — one narrow test, not a battery, but it moved as the imaging predicted.
Does a Microbleed Prove a Head Injury Caused It?
No, and here I have to amend something I have said from a podium. In my lectures I frame a microbleed as objective evidence that real force reached the tissue. I still hold that, with the correction Huang's controls require: twelve people with no traumatic injury also had microbleeds. A positive SWI is evidence. It is not proof.
A microbleed records that a vessel bled, not why. What carried statistical weight in Huang's data was location, not presence — the cortical and subcortical concentration separating the groups at p < 0.0001. A single dot in an unremarkable spot says nothing about cause.
Read in context, though — a known mechanism, symptoms that began at that moment, microbleeds at the gray-white junction or in the brainstem — SWI moves the case from "we cannot find anything" to "here is a physical finding consistent with what you described."
Do Microbleeds Predict How Someone Will Do?
Griffin and colleagues studied outcome in 439 patients with TBI (Griffin et al., 2019). Of those, 31% (134/439) had evidence of punctate and/or linear traumatic microbleeds on MRI, and the presence of traumatic microbleeds was an independent predictor of disability (P < 0.05; odds ratio = 2.5). Independent means the association survived adjustment — not merely a proxy for other things predicting a bad outcome. Prevalence rose with severity: 27% of mild patients, 47% of moderate patients, and 58% of severe TBI patients had TMBs.
Two limits belong with that. This was a mixed-severity cohort, not an mTBI cohort, so it does not transfer cleanly to someone concussed who walked out of the emergency department. And the authors state the interpretive problem themselves: it is "difficult to determine whether TMBs are simply a signature of more severe injury or whether they could be causally associated with worse outcome."
That distinction is not academic. If microbleeds mark how hard the brain was hit, they are prognostic; if they drive dysfunction, they are a treatment target. The data cannot yet separate those, and I will not pretend otherwise to patients living with [persistent symptoms](condition-post-concussion-syndrome.html).
What Are the Limits of This Sequence?
A negative SWI rules out very little. Most of Huang's 111 mTBI patients had no detectable microbleeds, and every one still had a mild traumatic brain injury. The diagnosis rests on characteristic symptoms and neurological effects. Imaging can support it; imaging cannot revoke it.
SWI also images bleeding, not axons. It catches the vascular consequence of shear, so a normal SWI alongside abnormal white-matter findings on another sequence is coherent, not contradictory.
And Wortzel's review concludes that the evidence remains insufficient to support routine use of advanced neuroimaging for clinical decisions about individual patients (Wortzel, 2022). That includes SWI. My reading is not that the sequence should go unordered — it is that no single scan should carry a diagnosis, a prognosis, or a legal argument alone. Which is this series' argument: eight modalities, eight questions.
Does This Help Separate a Brain Problem From a Neck Problem?
It does, and that is a large part of why it is in the protocol.
Most of my work concerns the craniocervical junction, so it is reasonable to ask whether I simply find the neck wherever I look. I would rather not, and a protocol that could only produce one answer would not be worth the radiologist's time.
Susceptibility-weighted imaging is a brain sequence. A microbleed at the gray-white junction or in the deep nuclei is located in brain tissue, and no amount of upper cervical correction will change it. When that finding is present, it tells me something happened inside the skull that my hands do not reach, and it tells me the person in front of me needs co-management rather than a single approach.
The reverse case is just as informative. A clean susceptibility study in someone with disabling symptoms and a measurably displaced junction shifts the weight of the explanation toward the structure I can actually address.
Sometimes it is the brain. Sometimes it is the neck. Sometimes both, which is the most common answer in the population I see. Running the brain sequences alongside imaging of the junction is what allows that to be a finding rather than an assumption.
Have You Been Told There Was Nothing There?
If your CT and standard MRI were read as normal after a head injury, that is the population Huang studied — and roughly a quarter had microbleeds their earlier imaging missed.
Ordering the sequence that answers your particular question, and reading the result honestly against what it can and cannot establish, is the work we do at Cerebral. If you'd like a real evaluation, we're here.
References
- Huang YL, Kuo YS, Tseng YC, et al. Susceptibility-weighted MRI in mild traumatic brain injury. *Neurology*. 2015;84(6):580–585. https://pubmed.ncbi.nlm.nih.gov/25576634/
- Griffin AD, Turtzo LC, Parikh GY, et al. Traumatic microbleeds suggest vascular injury and predict disability in traumatic brain injury. *Brain*. 2019;142(11):3550–3564. https://pubmed.ncbi.nlm.nih.gov/31608359/
- Bruggeman GF, Haitsma IK, Dirven CMF, et al. Traumatic axonal injury (TAI): definitions, pathophysiology and imaging — a narrative review. *Acta Neurochirurgica*. 2021;163(1):31–44. https://pubmed.ncbi.nlm.nih.gov/33006648/
- Wortzel HS. Advanced neuroimaging and mild traumatic brain injury litigation, revisited. *Journal of the American Academy of Psychiatry and the Law*. 2022;50(3):336–341. https://jaapl.org/content/50/3/336