Return separate P and N assessments, preserve modality/release, and never turn suspicion into pathology, a universal recurrence probability or an autonomous treatment order.
{
"categories": [
{
"outcome_code": "0",
"label": "incomplete",
"defining_state": "new_baseline_without_required_prior_imaging_that_is_known_or_expected_to_become_available"
},
{
"outcome_code": "1",
"label": "no_evidence_of_recurrence",
"defining_state": "expected_posttreatment_findings_without_suspicious_primary_site_or_neck_abnormality"
},
{
"outcome_code": "2",
"label": "low_suspicion",
"primary_site_subcategories": [
"2a_mucosal",
"2b_deep"
],
"neck_subcategory": "N2"
},
{
"outcome_code": "3",
"label": "high_suspicion",
"defining_state": "discrete_or_progressive_abnormality_highly_suspicious_for_viable_tumor"
},
{
"outcome_code": "4",
"label": "known_or_definite_recurrence",
"defining_state": "pathologically_proven_disease_or_definite_radiologic_and_clinical_progression"
}
],
"applicability": {
"intended_use": "Standardized posttreatment imaging surveillance after definitive or curative-intent treatment of head and neck cancer.",
"classification_unit": "one_examination_with_separate_primary_site_P_and_neck_N_assessments",
"required_context": [
"tumor_primary_site_and_histology",
"treatment_type_and_completion_date",
"baseline_and_prior_imaging",
"clinical_examination_or_endoscopy_when_relevant",
"modality_and_protocol",
"separate_primary_site_and_neck_findings"
],
"outside_scope": [
"during_active_treatment_response_scoring",
"untreated_initial_staging",
"pathologic_diagnosis_from_imaging_category",
"single_combined_P_and_N_category",
"automatic_treatment_selection"
]
},
"separate_axes_and_qualifiers": {
"primary_site_axis": "Assign a P category to the treated primary site and use the primary-specific 2a versus 2b branch when category 2 applies.",
"neck_axis": "Assign an N category to treated cervical nodes and the neck independently; the category may legitimately differ from P.",
"unassessable": "Use P-x or N-x when the relevant region cannot be assessed because of technique, artifact, coverage or anatomy. The x qualifier is not category 0 and is not an ordinal risk score.",
"unknown_primary": "Use the declared unknown-primary convention rather than inventing a primary-site category when no treated primary exists.",
"no_collapse_rule": "Never return one overall NI-RADS number by taking the higher or lower of P and N unless an explicitly named local implementation requires a separate summary outside the ACR axes."
},
"category_logic": {
"category_0": {
"gate": "Use only when a new baseline examination lacks a prior study that is known or expected to become available for comparison.",
"action": "Issue an incomplete comparison state and add an addendum after the prior examination becomes available.",
"distinction": "Technical nonassessability is P-x or N-x, not category 0; absence of any prior when none will be available is handled by clinical judgment rather than perpetual category 0."
},
"category_1": {
"primary_and_neck": "Expected posttreatment anatomy, fibrosis, edema or other stable benign change without a suspicious focus.",
"action": "Continue routine surveillance according to the cancer, treatment, timepoint and institutional pathway."
},
"category_2": {
"P_2a_mucosal": "Focal non-mass-like mucosal enhancement or focal reduced diffusion that is low suspicion and accessible to direct clinical inspection.",
"P_2a_action": "Recommend direct visual inspection or endoscopic correlation rather than treating 2a as a biopsy-proven recurrence.",
"P_2b_deep": "Deep ill-defined nonnodular soft tissue or another equivocal deep posttreatment abnormality without a discrete high-suspicion mass.",
"P_2b_action": "Recommend short-interval imaging; MRI is preferred for suspected perineural spread or skull-base evaluation when appropriate.",
"N_2": "Residual heterogeneous enhancement, mild-to-moderate FDG uptake, a new or enlarging node without definitively malignant morphology, or meaningful PET/MRI discordance.",
"N_2_action": "Recommend short-interval imaging and integrate morphology, diffusion, metabolic activity, trajectory and clinical findings."
},
"category_3": {
"any_of": [
"discrete_enhancing_mass_matching_the_tumor_bed",
"intense_focal_FDG_uptake_compatible_with_tumor",
"progressive_or_increasing_perineural_abnormality",
"node_with_highly_suspicious_morphology_or_concordant_progression"
],
"action": "Recommend image-guided or clinical biopsy when clinically indicated and technically feasible; multidisciplinary review can resolve contraindication, inaccessible site or overwhelming concordant evidence."
},
"category_4": {
"any_of": [
"pathologically_proven_residual_or_recurrent_tumor",
"definite_radiologic_progression_with_concordant_clinical_evidence_when_tissue_is_unnecessary_or_not_feasible"
],
"action": "Route to multidisciplinary clinical management while preserving site, extent and proof basis; category 4 is not a treatment regimen."
}
},
"modality_and_version_boundary": {
"CT_and_PET_CT": "Use the 2018 ACR white-paper framework and its modality-specific descriptors.",
"MRI": "Use the current MRI v2025 release and its assessment-category table; do not assume every CT/PET descriptor has identical MRI wording.",
"longitudinal_consistency": "State the release and modality used whenever comparison or automation could mix implementations.",
"PET_limitation": "Inflammation, physiologic uptake and treatment timing can mimic disease; metabolic intensity must be integrated with anatomy and trajectory."
},
"decisive_boundaries": {
"zero_versus_x": "Category 0 means comparison is incomplete because a required prior is expected; x means the anatomy is unassessable on the current examination.",
"one_versus_two": "Expected stable posttreatment change without a focal concern is 1; a focal low-suspicion mucosal, deep or nodal abnormality is 2.",
"P_2a_versus_P_2b": "A mucosal focus suitable for direct inspection is 2a; a deep ill-defined nonnodular abnormality requiring imaging surveillance is 2b.",
"two_versus_three": "Equivocal low-suspicion change without a discrete convincing tumor pattern is 2; a discrete mass, intense concordant uptake or convincing progression is 3.",
"three_versus_four": "High imaging suspicion awaiting confirmation is 3; pathology or definite radiologic-plus-clinical progression is 4."
},
"risk_interpretation": {
"qualitative_only": "The ordinal category communicates suspicion and linked workup, not a universal recurrence percentage.",
"why_probabilities_vary": [
"primary_versus_neck_axis",
"CT_PET_CT_or_MRI_release",
"tumor_site_and_histology",
"time_since_treatment",
"prevalence_and_referral_setting",
"reader_and_followup_reference_standard"
],
"prohibited_inferences": [
"category_1_equals_zero_recurrence_risk",
"category_2_has_one_portable_percentage",
"category_3_is_pathology",
"category_4_selects_a_regimen"
]
},
"agent_output_contract": [
"Return treatment type, completion date, modality, protocol, current release and comparison adequacy.",
"Return P and N separately and use 2a or 2b only on the primary-site axis.",
"Describe the exact mucosal, deep, nodal, diffusion, enhancement, FDG and trajectory evidence before the category.",
"Return category-linked next step as a recommendation with clinical and technical qualifiers.",
"Distinguish 0, P-x or N-x, unknown primary, suspicion, proof and treatment selection."
],
"missing_input_behavior": [
"If a known prior will become available, return category 0 and an addendum pathway rather than a false definitive category.",
"If the primary site or neck is technically unassessable, return P-x or N-x for that axis and still classify the other axis when possible.",
"If treatment status is not post-definitive surveillance, withhold NI-RADS and route to the appropriate staging or response framework.",
"If P-2a versus P-2b cannot be resolved, expose the ambiguity and the finding location rather than collapsing it to a generic 2 without explanation."
],
"supporting_sources": [
{
"role": "CT_PET_framework",
"citation": "Aiken et al. J Am Coll Radiol. 2018;15:1097-1108",
"doi": "10.1016/j.jacr.2018.05.006"
},
{
"role": "current_MRI_release",
"citation": "Aiken et al. J Am Coll Radiol. 2025",
"doi": "10.1016/j.jacr.2025.07.023"
},
{
"role": "official_release_index",
"citation": "ACR NI-RADS current-release page",
"url": "https://www.acr.org/Clinical-Resources/Clinical-Tools-and-Reference/Reporting-and-Data-Systems/NI-RADS"
}
],
"source_locator": "ACR NI-RADS MRI v2025 Assessment Categories table for 0, 1, primary 2a/2b, neck 2, 3, 4 and x qualifiers; Aiken et al. 2018, DOI 10.1016/j.jacr.2018.05.006, CT/PET white paper; ACR current-release index."
}