For healthcare professionals
Marshall CT Classification
The Marshall CT classification sorts the head CT after traumatic brain injury into categories: diffuse injury grades by basal cistern compression and midline shift, and mass lesions by whether a high- or mixed-density lesion is larger than 25 cm3 and whether it was evacuated. This tool is written for clinicians. It is not a self-check and not a diagnosis.
The CT findings
Definitions as printed by Thelin and colleagues (Thelin 2017). Nothing you enter leaves this page.
What you will see
The Marshall CT grade, from I to VI, and the definition it meets.
Educational only, not medical advice. A classification supports clinical judgment; it does not decide treatment on its own.
Source
Marshall LF et al., J Neurosurg 1991 (the original classification, cited only). Definitions as printed in Thelin EP et al., PLoS Med 2017 (CC BY), Table 1 and Methods; types I to VI and in-hospital mortality by type in Rakip U et al., BMC Emerg Med 2026 (CC BY).
Validated in
Patients with traumatic brain injury and a head CT. Thelin 2017 studied 1,115 patients with traumatic brain injury in neuro-intensive care in Stockholm and Helsinki; Rakip 2026 studied 657 adults with traumatic brain injury admitted for more than 24 h at one neurosurgical center in Türkiye (2022 to 2024).
Limits
For clinicians, not a diagnosis. The categories are not an ordinal score, they do not take subarachnoid blood into account and they do not tell epidural from subdural hematoma (Thelin 2017). Grade V depends on whether a mass lesion was evacuated, which makes it hard to compare with scores that use only the pre-operative scan (Thelin 2017). Use it with, not instead of, clinical judgment.
The categories
Marshall and colleagues described the classification in 1991 (Marshall 1991). The definitions below are as printed by Thelin and colleagues, who grouped grade V ("evacuated mass lesion") and grade VI ("non-evacuated mass lesion") together (Thelin 2017).
| Grade | Definition |
|---|---|
| Grade I | No visible intracranial pathology |
| Grade II | Midline shift of 0 to 5 mm, basal cisterns remain visible, no high- or mixed-density lesions > 25 cm3 |
| Grade III (swelling) | Midline shift of 0 to 5 mm, basal cisterns compressed or completely effaced, no high- or mixed-density lesions > 25 cm3 |
| Grade IV (shift) | Midline shift > 5 mm, no high- or mixed-density lesions > 25 cm3 |
| Grade V+VI | High- or mixed-density lesions > 25 cm3 (V: evacuated mass lesion, VI: non-evacuated mass lesion) |
Definitions as printed in Thelin 2017, Table 1; the V and VI labels are from its Methods (Thelin 2017). Thelin notes that the original classification used later CT scans to tell whether a mass lesion had been surgically removed (Thelin 2017). A 2026 study from Türkiye used the same six types, I to VI (Rakip 2026).
For example, a scan with visible injury, no lesion over 25 cm3, a midline shift of 3 mm and compressed basal cisterns is grade III (swelling).
What studies found
In 1,115 patients from Stockholm and Helsinki, the grades were spread as follows: I in 1 (0%), II in 279 (25%), III in 187 (17%), IV in 31 (3%) and V+VI in 617 (55%) (Thelin 2017). For unfavorable outcome (Glasgow Outcome Scale 1 to 3) the area under the curve (AUC) of the Marshall classification in the combined cohort was 0.58 (95% CI 0.55 to 0.61), and for death against survival it was 0.61 (0.57 to 0.65), lower than for the Rotterdam, Stockholm and Helsinki CT scores (Thelin 2017).
In 657 adults at one center in Türkiye, in-hospital mortality by Marshall type was as follows (Rakip 2026):
| Type | Total (n) | Deaths (n) | Mortality (%) |
|---|---|---|---|
| I | 135 | 3 | 2.2 |
| II | 10 | 1 | 10.0 |
| III | 61 | 6 | 9.8 |
| IV | 212 | 12 | 5.7 |
| V | 68 | 20 | 29.4 |
| VI | 171 | 30 | 17.5 |
As printed in Rakip 2026, Table 3. In that study the Marshall AUC for in-hospital mortality was 0.682 (95% CI 0.625 to 0.739), lower than the Rotterdam (0.766) and Helsinki (0.793) CT scores, and none of the CT scores added meaningfully to a model of age and admission GCS (Rakip 2026). These are figures for groups in one hospital, not a prediction for one patient.
Limits
Thelin and colleagues point out that the Marshall classification was never meant for outcome prediction, as it is not an ordinal score (the original authors acknowledge that grade IV is worse than grades V and VI), that it does not take subarachnoid hemorrhage into account or tell epidural from subdural hematoma, and that the 25 cm3 cutoff puts almost all extra-parenchymal bleeds into grade VI (Thelin 2017). A 2025 NINDS working group notes that the Marshall and Rotterdam systems have limits in milder injuries, for longer-term and broader outcomes, and for the wider range of lesions seen after brain injury (Mac Donald 2025).
Related tools
For points scores from the same admission CT, see the Rotterdam CT score and the Helsinki CT score. For the bedside level of consciousness, use the Glasgow Coma Scale calculator; to add the pupil response, see the GCS-Pupils score. To decide which adults with a minor head injury need a CT, see the Canadian CT Head Rule. To grade outcome after a brain injury, see the Glasgow Outcome Scale. Other bedside scores are listed under For healthcare professionals.
Sources
- Marshall LF, Marshall SB, Klauber MR, et al. A new classification of head injury based on computerized tomography. J Neurosurg. 1991;75(Suppl):S14-S20. doi:10.3171/sup.1991.75.1s.0s14. Journal and pages confirmed via CrossRef. Original classification (cited only). doi.org/10.3171/sup.1991.75.1s.0s14
- Thelin EP, Nelson DW, Vehvilainen J, et al. Evaluation of novel computerized tomography scoring systems in human traumatic brain injury: an observational, multicenter study. PLoS Med. 2017;14(8):e1002368. doi:10.1371/journal.pmed.1002368. PMID 28771476. Open access (CC BY 4.0). Table 1 (Rotterdam CT score items, points, sum + 1, range 1 to 6; Helsinki CT score items, points, sum, range −3 to 14); Table 2 (1,115 patients, score distribution); Table 3 (AUCs). The 2023 republication notice (PLoS Med 20(11):e1004320) removed a supporting file only. pmc.ncbi.nlm.nih.gov/articles/PMC5542385/
- Rakip U, Yıldızhan S, Canbek İ, et al. External validation and comparative performance of Marshall, Rotterdam, and Helsinki computed tomography scores in adult traumatic brain injury. BMC Emerg Med. 2026;26:242. doi:10.1186/s12873-026-01655-z. PMID 42332581. Open access (CC BY 4.0). Journal confirmed via CrossRef. Table 2 (types I to VI), Table 3 (in-hospital mortality by type), Table 4 (AUCs). pmc.ncbi.nlm.nih.gov/articles/PMC13548489/
- Mac Donald CL, Yuh EL, Vande Vyvere T, et al. Neuroimaging characterization of acute traumatic brain injury with focus on frontline clinicians: recommendations from the 2024 National Institute of Neurological Disorders and Stroke Traumatic Brain Injury Classification and Nomenclature Initiative Imaging Working Group. J Neurotrauma. 2025;42(13-14):1056-1064. doi:10.1089/neu.2025.0079. PMID 40393517. Open access (CC BY 4.0). Introduction: Marshall and Rotterdam CT scores and their limits. pmc.ncbi.nlm.nih.gov/articles/PMC12409119/
Written and maintained by the uncovermyrisk.com editorial team. We do not have a medical reviewer; every claim links to its primary source above. Last reviewed: 2026-11-04. Spotted an error? See About.