Marshall CT Original Marshall CT classification of traumatic brain injury

{
  "categories": [
    {
      "outcome_code": "I",
      "visible_intracranial_pathology": "none",
      "cisterns": "not_abnormal_by_classification",
      "midline_shift_mm": "not_abnormal_by_classification",
      "high_or_mixed_density_lesion": "none"
    },
    {
      "outcome_code": "II",
      "visible_intracranial_pathology": "present",
      "cisterns": "present",
      "midline_shift_mm": "0_to_5_inclusive",
      "high_or_mixed_density_lesion": "none_greater_than_25_cm3"
    },
    {
      "outcome_code": "III",
      "visible_intracranial_pathology": "present",
      "cisterns": "compressed_or_absent",
      "midline_shift_mm": "0_to_5_inclusive",
      "high_or_mixed_density_lesion": "none_greater_than_25_cm3"
    },
    {
      "outcome_code": "IV",
      "visible_intracranial_pathology": "present",
      "cisterns": "any",
      "midline_shift_mm": "greater_than_5",
      "high_or_mixed_density_lesion": "none_greater_than_25_cm3"
    },
    {
      "outcome_code": "V",
      "defining_state": "any_intracranial_lesion_surgically_evacuated",
      "time_dependency": "requires_postoperative_or_treatment_status"
    },
    {
      "outcome_code": "VI",
      "defining_state": "high_or_mixed_density_lesion_greater_than_25_cm3_not_surgically_evacuated",
      "time_dependency": "requires_current_non_evacuated_status"
    }
  ],
  "applicability": {
    "use_for": "Communication and historical risk stratification of acute traumatic brain injury using cranial CT plus known surgical-evacuation status.",
    "classification_unit": "one_patient_at_one_explicit_timepoint",
    "required_inputs": [
      "CT_acquisition_time_from_injury",
      "preoperative_or_postoperative_status",
      "all_visible_traumatic_intracranial_pathology",
      "basal_cistern_status",
      "maximum_midline_shift_in_mm",
      "high_or_mixed_density_lesion_type_and_volume_cm3",
      "whether_any_intracranial_lesion_was_surgically_evacuated"
    ],
    "useful_companion_inputs": [
      "Glasgow_Coma_Scale_and_timing",
      "pupil_response",
      "age",
      "hypotension_and_hypoxia",
      "intracranial_pressure",
      "serial_CT_change",
      "antithrombotic_use",
      "extracranial_injuries",
      "MRI_when_clinically_indicated"
    ],
    "outside_scope": [
      "nontraumatic_hemorrhage",
      "standalone_mild_TBI_exclusion",
      "individual_long_term_prognosis",
      "withdrawal_of_life_support",
      "automatic_surgical_triage",
      "MRI_only_classification_without_corresponding_CT_inputs"
    ]
  },
  "classification_hierarchy": [
    {
      "priority": 1,
      "if": "any_intracranial_traumatic_lesion_has_been_surgically_evacuated_by_the_classification_timepoint",
      "output_code": "V"
    },
    {
      "priority": 2,
      "if": "a_high_or_mixed_density_lesion_greater_than_25_cm3_remains_not_surgically_evacuated",
      "output_code": "VI"
    },
    {
      "priority": 3,
      "if": "midline_shift_is_greater_than_5_mm_and_no_qualifying_greater_than_25_cm3_lesion_exists",
      "output_code": "IV"
    },
    {
      "priority": 4,
      "if": "basal_cisterns_are_compressed_or_absent_shift_is_0_to_5_mm_and_no_qualifying_greater_than_25_cm3_lesion_exists",
      "output_code": "III"
    },
    {
      "priority": 5,
      "if": "traumatic_pathology_is_visible_cisterns_are_present_shift_is_0_to_5_mm_and_no_qualifying_greater_than_25_cm3_lesion_exists",
      "output_code": "II"
    },
    {
      "priority": 6,
      "if": "no_intracranial_traumatic_pathology_is_visible_on_CT",
      "output_code": "I"
    }
  ],
  "threshold_and_state_boundaries": {
    "exactly_5_mm_shift": "Exactly 5 mm remains within the 0-5-mm boundary for class II or III according to cistern status; class IV requires more than 5 mm.",
    "exactly_25_cm3": "Exactly 25 cm3 is not greater than 25 cm3 and therefore is not class VI solely by volume; if not evacuated, classify within II-IV from cisterns and shift while still reporting lesion type and exact volume.",
    "greater_than_25_cm3": "The volume branch applies to a high- or mixed-density intracranial lesion. Record lesion type, site, volume method and whether multiple lesions were combined; do not infer the threshold from vague words such as large.",
    "evacuated_status": "Class V is defined by the fact of surgical evacuation, regardless of the preoperative volume. A planned operation is not an evacuation, and a preoperative CT cannot be retrospectively labeled V unless the classification timepoint is explicitly updated.",
    "category_IV_versus_VI": "If shift is greater than 5 mm but a greater-than-25-cm3 high/mixed lesion remains un-evacuated, the mass-lesion branch VI takes priority; IV is the diffuse-shift category without that qualifying lesion.",
    "category_I_guard": "A normal-appearing CT does not exclude concussion, traumatic axonal or microvascular injury, small lesions below CT resolution or later evolution."
  },
  "imaging_and_measurement_protocol": {
    "timepoint_rule": "Label admission, worst, preoperative, postoperative or follow-up CT. The original framework is often applied to the initial CT, but treatment-defined class V requires subsequent operative information; never mix timepoints silently.",
    "cisterns": "Evaluate the basal cisterns as present, compressed or absent and separate this from ventricular size or focal sulcal effacement.",
    "shift": "Measure maximum midline displacement in millimetres at the relevant axial level and preserve the continuous value before category assignment.",
    "lesion_volume": "Use a documented reproducible method for each high/mixed-density mass lesion and retain the raw dimensions or segmentation method; classification alone does not validate an inferred volume.",
    "serial_change": "Repeat imaging can alter lesion volume, shift, cistern status and operative state. Preserve each time-stamped Marshall output instead of overwriting the admission phenotype."
  },
  "complete_injury_characterization_required": {
    "separately_report": [
      "epidural_hematoma",
      "acute_subdural_hematoma",
      "contusions_and_intraparenchymal_hematoma",
      "traumatic_subarachnoid_hemorrhage",
      "intraventricular_hemorrhage",
      "pneumocephalus",
      "skull_and_skull_base_fractures",
      "herniation_pattern",
      "hydrocephalus",
      "vascular_injury_when_assessed",
      "traumatic_axonal_or_microvascular_injury_when_assessed",
      "posterior_fossa_lesions"
    ],
    "omission_guard": "Marshall does not distinguish epidural from subdural hematoma, does not encode traumatic SAH or IVH directly and is insensitive to many axonal injuries. Never suppress granular findings because a Roman numeral was emitted.",
    "current_NINDS_direction": "Current NINDS work favors a multilevel lesion lexicon recording location, size and number and integrating clinical, blood-biomarker, injury and recovery modifiers; Marshall can be returned as a legacy companion field, not the complete phenotype."
  },
  "nonordinal_and_prognostic_boundary": {
    "nonordinal_rule": "The Roman numerals are category labels, not a monotonic severity ladder. In contemporary validation class IV can have worse outcomes than the treatment-defined V, and V versus VI is confounded by selection for surgery.",
    "modern_validation": "In a 1,115-patient neuro-intensive-care study, Marshall contributed less prognostic information than Stockholm and Helsinki CT scores and had low univariable pseudo-R2 of approximately 0.03-0.05 for long-term functional outcomes.",
    "current_guideline": "The 2024 neuroprognostication guideline recommends that Marshall not be considered a reliable standalone predictor of mortality at 6 months or later and rates evidence for unfavorable functional outcome as low quality.",
    "counseling_guard": "Never turn a category or historical cohort mortality percentage into an individual forecast, ceiling of recovery, treatment limitation or withdrawal-of-support recommendation. Integrate examination trajectory, pupils, age, systemic insults, validated multivariable models and substantial uncertainty."
  },
  "current_management_boundary": {
    "emergency_rule": "Acute TBI with mass effect, neurological deterioration, coma, pupil abnormality, herniation, refractory intracranial hypertension or another critical lesion requires urgent trauma, neurosurgical and neurocritical assessment independent of Marshall category.",
    "classification_threshold_guard": "The greater-than-25-cm3 Marshall boundary is a taxonomy branch, not a universal surgical threshold.",
    "lesion_specific_examples": "Current Brain Trauma Foundation recommendations use lesion- and patient-specific criteria, including EDH greater than 30 cm3, acute SDH thickness greater than 10 mm or shift greater than 5 mm, and parenchymal-lesion combinations involving GCS, deterioration, ICP, location, volume, shift and cisterns. These must not be reverse-engineered from the Marshall numeral.",
    "class_V_guard": "Class V documents that evacuation occurred; it does not prove that surgery was indicated, successful or sufficient and cannot be assigned prospectively as a recommendation.",
    "class_VI_guard": "Class VI documents a qualifying lesion not evacuated at the stated timepoint; it does not mean surgery is contraindicated, futile or permanently withheld.",
    "safety_rule": "Never emit observation, repeat-CT timing, ICP monitoring, craniotomy, decompression, discharge or treatment limitation from Marshall alone."
  },
  "comparison_with_Rotterdam": {
    "distinction": "Rotterdam separates cistern status, shift, epidural mass lesion and traumatic SAH or IVH into an additive prognostic score; it is not a numeric conversion of Marshall.",
    "interoperability_guard": "Return the named system and its raw components. Do not map Marshall I-VI to Rotterdam 1-6 by matching numbers or assumed severity."
  },
  "agent_output_contract": [
    "Marshall_1991_system_identity",
    "CT_time_from_injury_and_named_timepoint",
    "preoperative_or_postoperative_state",
    "visible_pathology_and_complete_granular_lesion_list",
    "basal_cistern_status",
    "maximum_midline_shift_mm_with_exact_boundary_handling",
    "each_high_or_mixed_density_lesion_type_volume_cm3_and_measurement_method",
    "surgical_evacuation_status_and_time",
    "single_supported_I_to_VI_category_or_explicit_unclassifiable_state",
    "nonordinal_and_nonprognostic_warning",
    "urgent_clinical_flags_without_grade_only_management",
    "missing_inputs_and_uncertainty"
  ],
  "missing_input_behavior": [
    "If operative status or the classification timepoint is unknown, do not choose V or VI when a mass lesion could cross those branches; request the missing status.",
    "If shift is described but not measured near 5 mm, return II/III versus IV unresolved and request the continuous measurement.",
    "If lesion volume is unknown or approximately 25 cm3, do not infer greater than 25 and do not choose VI solely from words such as large or mass effect.",
    "If cisterns are not evaluable with shift 0-5 mm, return II versus III unresolved rather than guessing.",
    "If CT is negative but symptoms or examination support TBI, return class I only as no visible CT pathology and preserve the clinical diagnosis and MRI limitations separately.",
    "If multiple CTs are provided, classify each named timepoint or require the caller to select admission, worst, preoperative or follow-up; never blend their findings."
  ],
  "supporting_sources": [
    {
      "role": "original_classification",
      "citation": "Marshall et al. J Neurosurg. 1991;75 Suppl:S14-S20",
      "doi": "10.3171/sup.1991.75.1s.0s14"
    },
    {
      "role": "modern_CT_score_validation_and_nonordinal_limit",
      "citation": "Thelin et al. PLOS Medicine. 2017;14:e1002368",
      "doi": "10.1371/journal.pmed.1002368",
      "pmcid": "PMC5542385"
    },
    {
      "role": "current_neuroprognostication_boundary",
      "citation": "Guidelines for Neuroprognostication in Critically Ill Adults with Moderate-Severe TBI. 2024",
      "doi": "10.1007/s12028-023-01902-2",
      "pmcid": "PMC10959796"
    },
    {
      "role": "current_granular_imaging_characterization",
      "citation": "Mac Donald et al. J Neurotrauma. 2025;42:1056-1064",
      "doi": "10.1089/neu.2025.0079",
      "pmcid": "PMC12409119"
    },
    {
      "role": "lesion_specific_surgical_boundary",
      "citation": "Brain Trauma Foundation Guidelines for the Surgical Management of TBI",
      "url": "https://braintrauma.org/coma/guidelines/surgical"
    }
  ],
  "source_locator": "Marshall et al. J Neurosurg 1991;75 Suppl:S14-S20, DOI 10.3171/sup.1991.75.1s.0s14, original category table; Thelin et al. 2017, DOI 10.1371/journal.pmed.1002368, PMC5542385, category construction, timepoint and comparative prognostic limits; 2024 neuroprognostication guideline, DOI 10.1007/s12028-023-01902-2, PMC10959796, Marshall recommendations; Mac Donald et al. 2025, DOI 10.1089/neu.2025.0079, PMC12409119, granular NINDS imaging model; current Brain Trauma Foundation surgical guideline, lesion-specific thresholds."
}