How Swelling on a Brain Scan Could Predict Recovery After Road Traffic Accidents

When a motorcyclist or pedestrian is rushed into a trauma center after a road traffic accident, one of the first images clinicians see is a non-contrast computed tomography scan of the head. That single scan carries enormous weight: it reveals fractures, hemorrhages, midline shift, and the state of the fluid-filled spaces at the base of the brain. Yet according to a new study from a trauma center in Chongqing, China, the most widely used CT scoring systems may be missing the very feature that matters most for predicting how these patients will fare months later — cerebral oedema, the swelling of brain tissue that follows mechanical injury.

The research, published in Frontiers in Radiology by a team at Daping Hospital of Army Medical University, set out to answer a deceptively simple question: can the way swelling spreads across the brain’s lobes predict long-term functional outcome better than the established Marshall, Rotterdam, and Helsinki CT scores? The answer, drawn from 227 patients injured in road traffic accidents between January 2019 and December 2023, was a clear yes. A new grading scheme called the A5 (+1) CT classification achieved an area under the curve (AUC) of 0.849 for predicting unfavorable outcomes at three months, outperforming the Marshall classification (0.720), the Rotterdam score (0.745), and the Helsinki score (0.746), with all comparisons reaching statistical significance.

To understand why this matters, it helps to consider what makes road traffic accident injuries different from other causes of traumatic brain injury. A fall produces relatively focused forces, but a collision generates rapid acceleration-deceleration and rotational shear forces that stretch and tear axonal fibers throughout the brain. These biomechanics carry a markedly higher risk of diffuse axonal injury and widespread vasogenic oedema, in which the blood-brain barrier is physically disrupted and fluid leaks into the extracellular space. Conventional CT scores, however, weight their predictions heavily toward hematoma volume and ventricular dimensions, capturing midline shift and basal cistern compression but largely ignoring the spatial distribution of swelling across the hemispheres.

The A5 (+1) classification addresses this blind spot with a straightforward ordinal scale. Grade 0 means no oedema; grade 1 confines swelling to a single cerebral lobe; grade 2 involves multiple lobes within one hemisphere; grade 3 marks bilateral oedema; grade 4 adds effacement of the cortical sulci with visible diffuse oedema extending into the basal cisterns; and grade 5 describes diffuse swelling that obliterates both the sulci and the cisterns entirely. A bonus point — the +1 — is awarded when oedema produces a mass effect that pushes midline structures more than five millimeters off center. In effect, the system maps how far the swelling has traveled and how hard it is squeezing the brain, rather than relying on a single snapshot of hemorrhage geometry.

The study design reflected the biology of traumatic oedema, which typically peaks between two and four days after injury. Every patient underwent head CT at least once within 24 hours of injury and again between days two and four, using either a GE 64-slice or Philips 128-slice scanner with standardized 5-millimeter slice thickness. The researchers recorded the worst oedema observed within 96 hours, quantified hemorrhage and contusion volumes with the Tada formula, and measured perilesional swelling with 3D Slicer software. For patients with contusion-associated oedema, they also computed an oedema index — the ratio of oedema volume to hemorrhage volume — as a quantitative companion to the categorical grading.

The cohort itself was characteristic of road traffic trauma: 73.1 percent male, median age 48, with motorcyclists (41.8 percent) and pedestrians (39.6 percent) dominating the road-user roles. Severity spanned the full spectrum — 36.1 percent had severe injuries with Glasgow Coma Scale scores of 3 to 8, 20.7 percent moderate, and 43.2 percent mild — and 66.1 percent had multiple intracranial lesions. Patients who underwent decompressive craniectomy or hematoma evacuation within four days of injury were excluded, because these operations alter hemodynamics and oedema progression in ways that would confound the imaging analysis. Outcomes were assessed at three months using the Glasgow Outcome Scale, with scores of 4 or 5 counted as favorable and 3 or below as unfavorable.

The statistical picture was striking. The IMPACT model — a clinical prognostic tool built from age, motor score, and pupillary light reflex — achieved an AUC of 0.838 on its own, already beating every conventional CT score. But when the A5 (+1) classification was added to IMPACT, the AUC climbed to 0.914, a gain of 0.076 that was highly significant (p = 0.001). By contrast, adding the Rotterdam score improved the AUC by only 0.038 (p = 0.028), while the Marshall classification and Helsinki score produced no significant improvement at all. Net reclassification analysis reinforced the point: incorporating the A5 (+1) grade raised the NRI by 86.8 percent, and decision curve analysis showed the combined model delivered high clinical value across a meaningful range of risk thresholds.

Among the 47.6 percent of patients who actually showed cerebral oedema on their scans, the differences were even more vivid. Those with unfavorable outcomes had a median oedema volume of 5.463 milliliters versus 1.343 milliliters in those who recovered well, and higher oedema index values (2.864 versus 2.287). The A5 (+1) grades told the same story from the other direction: 71.2 percent of patients with poor outcomes sat in grades 3 to 5, while 77.6 percent of those with good outcomes clustered in grades 1 and 2. In this oedema subgroup, however, the picture grew more nuanced — the IMPACT model alone reached an AUC of 0.798, and while adding the oedema index lifted the combined AUC by 0.05 (p = 0.035), adding the A5 (+1) grade did not produce a statistically significant further gain, possibly because dehydrating agents given during acute care increase tissue density on CT and can mask radiographic signs of swelling, and because the oedema subgroup was relatively small.

The authors place their findings in a broader pathophysiological context. Traumatic oedema differs fundamentally from the swelling that follows ischemic stroke: trauma mechanically ruptures the blood-brain barrier and triggers inflammatory cascades that drive vasogenic oedema, which can persist for up to seven days, whereas early stroke oedema is primarily cytotoxic, arising from energy failure, and follows a more predictable sequence confined to the vascular territory of the blocked artery. Traumatic oedema is instead diffuse and heterogeneous, involving both the injured region and remote brain areas — a pattern that a lobe-by-lobe grading system is inherently positioned to capture. Autopsy studies cited in the paper support the biological relevance of CT-based grading, showing that brain water content in patients with Marshall grades I and II approached normal levels, while patients with grade III or higher showed markedly elevated tissue densities.

The study’s limitations are acknowledged candidly: it was single-center, retrospective, and modest in size; it lacked external validation; and factors such as hypoxia, cerebrovascular autoregulation, blood glucose, and hemoglobin could not be incorporated. Interventions like intracranial pressure monitoring and external ventricular drainage were permitted and may have influenced imaging findings. Still, the core message is compelling and practical. For the leading cause of traumatic brain injury worldwide — and the dominant one in low- and middle-income countries — a simple, repeatable reading of how far swelling has spread across the brain’s lobes, layered onto an established clinical model, appears to sharpen prognosis considerably. The team advocates multicenter validation to confirm generalizability, but the implication for emergency radiology is already clear: the pattern of oedema on a routine head CT is not just a descriptive finding, it is a prognostic signal that conventional scores have been leaving on the table.

Subject of Research: Prognostic value of cerebral oedema grading on head CT for traumatic brain injury following road traffic accidents

Article Title: Prognostic value of oedema findings on head CT in traumatic brain injury following road traffic accidents

Article References: Chen, D., Ma, W., Zhan, L., Zhao, H., & Xiong, K. (2026). Prognostic value of oedema findings on head CT in traumatic brain injury following road traffic accidents. Frontiers in Radiology, 6, Article 1941750. https://doi.org/10.3389/fradi.2026.1941750

Image Credits: AI Generated

DOI: 10.3389/fradi.2026.1941750

Keywords: traumatic brain injury, cerebral oedema, head CT, road traffic accidents, A5 (+1) CT classification, IMPACT model, Marshall CT classification, Rotterdam CT score, Helsinki CT score, prognosis, emergency radiology, Glasgow Outcome Scale