The imaging paradox

Roughly 2.5 million Americans are evaluated for a traumatic brain injury each year, and the overwhelming majority of those injuries are mild. Here is the paradox that defines the field: the injury that brings most patients to the emergency department is the one conventional imaging is least equipped to see.

Concussion is a functional injury, diagnosed clinically rather than radiologically. Decision rules like the Canadian CT Head Rule exist to identify the minority of mild-TBI patients whose scan will change management and spare everyone else the radiation. When a CT is performed, the goal is to rule out what needs a neurosurgeon in the next hour. A normal head CT does not rule out a concussion. It never has.

So where does quantitative CT analysis fit? Not at diagnosis — but in the space after the initial read. The patients who get scanned more than once, who don’t recover on schedule, who carry a history of repeated impacts, or who sit at the moderate-to-severe end of the spectrum where structural change is real and progressive. That space demands precisely what unaided reading does worst: detecting small changes in three-dimensional volume between scans acquired weeks or months apart, on different scanners, at different angles.

Two numbers that matter for monitoring

NeuroQuant CT is FDA-cleared automated post-processing software for non-contrast head CT. It registers the study to an anatomical atlas along the AC-PC plane and returns quantitative measures in minutes. Two of its outputs carry real weight in post-injury monitoring.

Lateral ventricle volume (LVV) is the measure to follow after a significant head injury. Ventricular enlargement following moderate or severe TBI is common, under-recognized, and treatable in a subset of patients — which is exactly why the measurement matters. Enlargement driven by post-traumatic hydrocephalus reflects CSF dysregulation, and a shunt can meaningfully improve that patient’s trajectory. Enlargement from ex vacuo dilatation is the passive consequence of parenchymal volume loss, where a shunt offers nothing. Same finding on the scan, opposite answers on intervention.

The clinical question at each follow-up is the same: is this bigger than the last scan, and by how much? Serial CTs answer it badly by eye. A 10–15% ventricular volume change is close to invisible on side-by-side comparison, particularly when the two scans were acquired at different head angles. Published reader-performance work has measured exactly that task and found meaningful gains when quantification is available: Farid et al. (2017) reported a 19% improvement in radiologist certainty when determining a change in ventricular volume, and Chang et al. (2017) reported a 73% improvement in reading efficiency for the same assessment. Each report quantifies the current study against the immediate prior — and because prior reports remain available for reference, a reader following a patient over months can look back across the full sequence rather than relying on recall.

Ventricle-to-subarachnoid ratio (VSR) addresses what LVV can not: whether the ventricles grew, but why. Comparing ventricular volume against the supratentorial subarachnoid space measures ventricle-to-sulcal concordance — whether the CSF spaces are expanding together or at each other’s expense. In parenchymal atrophy, ventricles and sulci enlarge in parallel, and the ratio stays relatively stable. In hydrocephalus, ventricles expand while the subarachnoid spaces are effaced, and the ratio climbs.

That is the fork in the road in post-traumatic ventriculomegaly. Rising ventricular volume alongside a rising ratio points one direction; rising volume with a stable ratio points the other. Read together, the two measures carry more information than either does alone.

Where subtraction imaging earns its keep

Numbers persuade radiologists; pictures persuade everyone else.

Standardized alignment is the prerequisite. Head CTs are acquired at whatever angle the patient’s head happened to be at, and automatic AC-PC registration means the follow-up is compared to the prior in the same orientation — which is what makes any visual comparison trustworthy at all.

Subtraction maps are the centerpiece, and they map directly onto the concussion-monitoring theme, because the entire clinical difficulty is that the change is too small to see. A subtraction map between a prior and a current head CT renders it visible: ventricular margins expanding outward, a collection evolving or resolving, subtle parenchymal volume loss. What was a subjective “maybe slightly larger” becomes a signal in a specific place — legible not just to the radiologist, but to the neurosurgeon, the intensivist, the rehab physician, and the family.

The image that makes this case best is the simplest one: a prior and current CT pair that look essentially identical, shown beside the subtraction map of what the eye missed.

The honest close

Quantitative CT analysis will not diagnose a concussion, and Concussion Awareness Day is a good moment to say so plainly. What it can do is make the next scan more informative than the last — turning follow-up imaging from a subjective side-by-side into a measured comparison, and making the small changes that precede a bad outcome visible while there is still something to do about them.

For patients whose recovery stalls, that difference is not incremental.