Skip to content

Hepatocellular carcinoma — diagnosis and imaging

TL;DR — HCC can often be diagnosed without biopsy in an at-risk liver when multiphase CT or MRI shows the required combination of size and major vascular features (Singal 2023, PMID 37199193). LI-RADS standardizes acquisition, terminology, probability categories, tumor-in-vein assessment, and treatment response, but its categories apply only to defined high-risk populations. CT/MRI LR-5 has high specificity (~92%) but moderate sensitivity (~67%) in meta-analysis, reflecting a deliberate preference to avoid false-positive HCC diagnosis (Lee 2020, PMID 32145134). Indeterminate observations require interval imaging, alternative modality, multidisciplinary review, or biopsy according to their probability and management consequence. Biopsy remains important when imaging is non-diagnostic or the patient lies outside validated non-invasive criteria; contemporary systematic review estimated needle-track seeding at 0.62% across 13,959 biopsies (Nie 2025, PMID 39692333).

Diagnostic context comes first

The same enhancing liver lesion has a different pre-test probability in cirrhosis, chronic HBV without cirrhosis, and a person with no known liver disease. Non-invasive HCC criteria are designed for populations with sufficiently high prior probability; applying them indiscriminately sacrifices positive predictive value (Singal 2023, PMID 37199193).

Before category assignment, establish:

  • whether the patient meets a high-risk definition;
  • whether the study is technically adequate and truly multiphasic;
  • lesion size and whether it is new or growing;
  • prior locoregional or systemic treatment;
  • transplant candidacy and the consequence of a false-positive diagnosis;
  • whether another primary malignancy could metastasize to liver.

From surveillance to diagnosis

Trigger Diagnostic action Reason
New lesion ≥1 cm on surveillance ultrasound Multiphase contrast CT or MRI Characterize vascular features (Singal 2023, PMID 37199193)
Sub-centimetre lesion Repeat ultrasound at short interval CT/MRI performance is limited at very small size
Rising AFP without lesion Repeat/alternative cross-sectional imaging; evaluate non-HCC causes AFP is neither specific nor anatomical
Limited ultrasound visualization Alternative modality, often MRI A technically inadequate test is not a negative test
Incidental lesion in non-at-risk liver General focal-liver-lesion pathway; pathology often required LI-RADS HCC probabilities do not transport automatically

Multiphase technique

HCC diagnosis depends on contrast kinetics, so timing is part of the test. A diagnostic study generally needs late arterial, portal venous, and delayed phases; hepatobiliary contrast MRI adds a later phase that reflects hepatocyte function.

Phase/sequence What it tests HCC relevance
Precontrast Baseline T1 signal, fat, blood products Prevents mistaking intrinsic brightness for enhancement
Late arterial Tumor arterialization Non-rim arterial-phase hyperenhancement is a major feature
Portal venous Relative portal supply Washout may become visible
Delayed Persistent relative hypoenhancement/capsule Supports washout or enhancing capsule
Diffusion-weighted MRI Water mobility/cellularity Ancillary feature; not alone diagnostic
T2-weighted MRI Tissue water and lesion character Ancillary characterization
Hepatobiliary phase Functioning hepatocyte uptake Hypointensity can support suspicion but is not identical to extracellular washout

Motion, mistimed arterial phase, cardiac output, contrast choice, and subtraction technique can change whether a major feature is seen.

Major imaging features

Feature Operational meaning Important mimic or pitfall
Non-rim arterial-phase hyperenhancement Lesion enhances more than liver in arterial phase Perfusion alteration; arterioportal shunt
Non-peripheral washout Lesion becomes relatively hypoenhancing later Background liver enhancement can create pseudo-washout
Enhancing capsule Smooth peripheral enhancing rim Fibrous pseudocapsule; delayed enhancement variation
Threshold growth Size increase meeting defined rate/interval Measurement plane and reader variation
Tumor in vein Enhancing soft tissue within vein Bland thrombus

An individual-patient-data meta-analysis connected CT/MRI and contrast-enhanced-ultrasound LI-RADS major features with histologic HCC, showing that feature performance depends on modality, lesion size, and combinations rather than one universal sign (van der Pol 2022, PMID 34783596).

LI-RADS probability categories

Category Interpretation Typical management logic
LR-1 Definitely benign Routine surveillance
LR-2 Probably benign Routine or individualized follow-up
LR-3 Intermediate probability Repeat/alternative imaging
LR-4 Probably HCC Multidisciplinary decision; biopsy or short-interval imaging
LR-5 Definitely HCC by imaging criteria Treat/stage as HCC in eligible population
LR-M Malignant, not specific for HCC Biopsy often needed because cholangiocarcinoma/combined tumor is possible
LR-TIV Tumor in vein Changes stage and treatment allocation
LR-NC Not categorizable Repeat technically adequate examination

In 14 studies with 2,056 patients, 2,589 observations, and 1,693 HCCs, CT/MRI LR-5 showed pooled per-observation sensitivity 67% (95% CI 62%–72%) and specificity 92% (95% CI 88%–95%) (Lee 2020, PMID 32145134). The asymmetry is intentional: LR-5 is a high-specificity endpoint, not a rule-out test.

CT versus MRI

Dimension Multiphase CT MRI
Acquisition Fast; widely available Longer; motion-sensitive
Radiation Ionizing radiation None
Contrast Iodinated Extracellular or hepatobiliary gadolinium agents
Small-lesion characterization Good Often higher sensitivity with more sequences
Artefacts Arterial timing, beam hardening Motion, susceptibility, arterial timing
Access Generally broader Cost and scanner capacity constraints

Choice should reflect local expertise, prior study quality, kidney function, contrast considerations, body habitus, and whether the alternative modality can resolve the specific uncertainty. Repeating the same technically limited study is rarely informative.

Contrast-enhanced ultrasound

Contrast-enhanced ultrasound evaluates arterial enhancement and washout in real time without CT radiation. CEUS LI-RADS LR-5 had pooled sensitivity 0.71 and specificity 0.88 for HCC across eight studies and 4,215 focal lesions; LR-M had sensitivity 0.85 and specificity 0.86 for other malignancies (Li 2021, PMID 33307595).

Strengths include real-time arterial observation and use when CT/MRI contrast is problematic. Limitations include acoustic-window dependence, inability to stage the entire liver and extrahepatic disease in one examination, operator dependence, and regional differences in availability and guideline integration.

When biopsy matters

Biopsy is considered when:

  • imaging remains indeterminate and the result changes management;
  • the patient is outside a validated high-risk non-invasive pathway;
  • LR-M or another malignancy is suspected;
  • molecular analysis is needed for a trial;
  • atypical progression raises doubt about the original diagnosis.
Biopsy issue Evidence-informed interpretation
Sampling error Small or heterogeneous tumors may yield non-diagnostic tissue
Pathologic continuum High-grade dysplastic nodule versus well-differentiated HCC can be difficult
Bleeding Depends on coagulation, platelets, portal hypertension, approach
Needle-track seeding Systematic review: 0.62% across 13,959 HCC biopsies; reported range 0%–7.77% (Nie 2025, PMID 39692333)
Negative biopsy Does not exclude HCC when pre-test probability and imaging remain high

The decision should compare the consequence of diagnostic uncertainty with procedural risk—not treat biopsy avoidance as an absolute rule.

Pathology and mixed tumors

Histology can establish HCC differentiation and architecture, but small lesions sit on a spectrum from regenerative and dysplastic nodules to early and progressed HCC. Immunohistochemical panels can support diagnosis; no single stain replaces morphology and clinical context.

Combined hepatocellular-cholangiocarcinoma and intrahepatic cholangiocarcinoma matter because an HCC-specific non-invasive label can lead to the wrong transplant or systemic-treatment pathway. LR-M is designed to flag malignancy without claiming HCC specificity (Singal 2023, PMID 37199193).

Macrovascular invasion

Tumor in the portal or hepatic vein is biologically and therapeutically different from bland thrombus. Enhancement within venous soft tissue, continuity with tumor, and vessel expansion support tumor thrombus. Incorrect classification can shift a patient between transplant, locoregional, and systemic pathways.

Imaging should report:

  • vessel and branch involved;
  • extent toward main portal vein or vena cava;
  • enhancement and diffusion features;
  • bland thrombus component;
  • portal-hypertension consequences.

Extrahepatic staging

Once HCC is diagnosed, staging evaluates:

  • number and distribution of intrahepatic lesions;
  • largest viable tumor diameter;
  • macrovascular invasion;
  • lymph nodes;
  • lung, bone, adrenal, and other metastases when clinically indicated;
  • liver reserve and performance status.

Tumor morphology alone does not determine treatment; it enters BCLC and transplant frameworks alongside liver function, performance status, AFP, response to prior therapy, and technical feasibility (Reig 2022, PMID 34801630).

Treatment-response imaging

After embolization or ablation, size alone can remain stable despite necrosis. Viable tumor is inferred from residual or recurrent mass-like enhancement. The LI-RADS treatment-response algorithm standardizes viable, equivocal, and nonviable categories.

In a transplant-explant validation after bland embolization, the LI-RADS treatment-response algorithm was compared with histopathologic viability; accuracy and inter-reader agreement supported standardization while revealing an equivocal category that cannot be treated as pathologic certainty (Shropshire 2019, PMID 31038409).

Response assessment depends on therapy:

Therapy Key imaging question
Thermal ablation Is there nodular enhancement at or around the margin?
TACE Is viable enhancing tissue present despite retained embolic material?
TARE Is delayed radiation effect being mistaken for progression?
External-beam radiation Is persistent enhancement evolving as expected after radiation?
Systemic immunotherapy Is new/enlarged disease true progression, atypical response, or mixed response?

Reporting minimum dataset

  • High-risk context and modality/contrast.
  • Technical adequacy and arterial timing.
  • Observation number, segment, size, and comparison date.
  • Major and ancillary features.
  • LI-RADS category.
  • Tumor-in-vein status.
  • Extrahepatic findings relevant to stage.
  • Treatment-response category when applicable.
  • Explicit recommendation for indeterminate findings.

Diagnostic-performance ledger

Question Best pooled estimate Consequence
LI-RADS ≥3 on CT/MRI Sensitivity 0.86 (95% CI 0.78–0.91), specificity 0.85 (0.78–0.90), AUC 0.92 Useful for risk stratification, but too nonspecific to equate every LR-3/4 lesion with HCC (Liang 2021, PMID 33276248)
LR-5: MRI versus CT Sensitivity 61% (95% CI 43–76) versus 48% (31–65); specificity 93% versus 96% MRI gains sensitivity while both retain high specificity; heterogeneity was high (Kim 2022, PMID 35849177)
All-size HCC: MRI versus CT Sensitivity 0.82 versus 0.66 in comprehensive comparative studies Modality choice matters most for small lesions; evidence quality and contrast-agent differences limit a single universal hierarchy (Roberts 2018, PMID 28859233)
CEUS LR-5 Sensitivity 69% (95% CI 64–73), specificity 92% (83–96) A positive category is persuasive, but CEUS does not survey the entire liver equivalently to CT/MRI (Shin 2020, PMID 32722894)
HCC proportion in LR-5 96% for CEUS and 95% for CT/MRI Similar positive predictive enrichment can coexist with different lesion sampling and washout definitions (Zhou 2022, PMID 35425706)
CEUS inter-reader agreement Pooled κ 0.73 for non-rim APHE; 0.69 for mild washout; 0.54 for late washout Feature interpretation is reproducible but not deterministic (Kang 2021, PMID 34156509)
Historical biopsy seeding 2.7% (95% CI 1.8–4.0) Older techniques likely overestimate contemporary risk, but biopsy still requires a management-changing indication (Silva 2008, PMID 18669577)

Where imaging categories break down

Combined hepatocellular–cholangiocarcinoma can show HCC-like arterialization in one region and biliary/stromal features elsewhere. Its diagnosis remains morphologic, with immunohistochemistry refining rather than replacing routine histology; small biopsies can sample only one lineage (Beaufrère 2021, PMID 33545267; Roßner 2023, PMID 36672443). A deep-learning analysis of 405 combined tumors could reclassify cases toward HCC or intrahepatic cholangiocarcinoma in ways associated with outcomes and molecular features, but this is a research phenotype rather than a validated clinical diagnostic standard (Calderaro 2023, PMID 38092727). The absence of an established systemic-treatment standard for this rare entity makes correct classification clinically material (Ye 2024, PMID 38344449).

Radiomics is similarly promising but immature. Eleven grade-prediction studies included 2,245 patients, none prospectively designed and only two with external test cohorts (Wang 2023, PMID 37541183). A later meta-analysis reported encouraging discrimination of intrahepatic cholangiocarcinoma from HCC, yet retrospective design, segmentation variation, scanner dependence, and limited external validation prevent routine substitution for expert imaging or tissue (Wang 2025, PMID 40571544).

Response imaging is treatment-specific

The 2024 LI-RADS treatment-response revision separates non-radiation from radiation pathways because persistent enhancement after radiation may reflect delayed necrosis rather than viable tumor (Aslam 2024, PMID 39530896). In explant-correlated SBRT data, sensitivity for incomplete necrosis ranged 71%–86% across readers, but specificity was less stable, underscoring that an early enhancing focus after radiation cannot be interpreted exactly like enhancement after thermal ablation (Mendiratta-Lala 2022, PMID 34644607). After ablation, a 119-patient explant series found 32% of treated tumors still viable pathologically, providing a stringent test of imaging rather than clinical follow-up alone (Shenoy-Bhangle 2025, PMID 40413459). CEUS now has a parallel non-radiation response lexicon, but evidence remains less extensive than for CT/MRI (Lyshchik 2024, PMID 38805727).

Controversies

  • Highest sensitivity versus highest specificity. MRI detects more HCC than CT in pooled comparisons, while CT LR-5 may be marginally more specific; availability, motion, renal function, contrast agent, and transplant policy can decide which error matters more (Kim 2022, PMID 35849177; Roberts 2018, PMID 28859233).
  • Biopsy versus serial imaging. Tissue resolves mixed histology and enables molecular work, but sampling error and tract seeding remain; serial imaging avoids an invasive procedure but can delay a non-HCC diagnosis (Silva 2008, PMID 18669577; Beaufrère 2021, PMID 33545267).
  • CEUS as problem-solver versus primary modality. CEUS offers real-time arterial phase and no nephrotoxic iodinated contrast, but field-of-view, operator dependence, and differing washout behavior constrain interchangeability (Qin 2022, PMID 35164930; Zhou 2022, PMID 35425706).
  • AI promise versus readiness. Retrospective radiomics performance is not evidence of prospective clinical utility; locked models, external calibration, acquisition harmonization, and decision-impact trials remain prerequisites (Wang 2023, PMID 37541183; Wang 2025, PMID 40571544).

Open questions

  • Can quantitative enhancement or radiomics improve sensitivity without eroding LR-5 specificity (Lee 2020, PMID 32145134)?
  • Which LR-3/LR-4 lesions should undergo biopsy rather than serial imaging?
  • How should hepatobiliary-phase features be harmonized across contrast agents and guidelines?
  • Can liquid biopsy resolve LR-M or indeterminate nodules without tissue?
  • How should treatment-response criteria differ among TACE, TARE, radiation, ablation, and immunotherapy (Shropshire 2019, PMID 31038409)?

References

  1. Singal AG, et al. AASLD Practice Guidance on prevention, diagnosis, and treatment of hepatocellular carcinoma. Hepatology. 2023;78:1922-1965. PMID 37199193
  2. Lee S, et al. Diagnostic performance of CT/MRI Liver Imaging Reporting and Data System v2017 for hepatocellular carcinoma: a systematic review and meta-analysis. Liver Int. 2020;40:1488-1497. PMID 32145134
  3. van der Pol CB, et al. CT/MRI and CEUS LI-RADS major features association with hepatocellular carcinoma: individual patient data meta-analysis. Radiology. 2022;302:326-335. PMID 34783596
  4. Li J, et al. Diagnostic accuracy of contrast-enhanced ultrasound LI-RADS for differentiating HCC and other hepatic malignancies in high-risk patients: a meta-analysis. Ultraschall Med. 2021;42:187-193. PMID 33307595
  5. Nie C, et al. Tumor seeding with needle biopsy of hepatocellular carcinoma: a systematic review. Am J Gastroenterol. 2025;120:1144-1151. PMID 39692333
  6. Shropshire EL, et al. LI-RADS Treatment Response Algorithm: performance and diagnostic accuracy. Radiology. 2019;292:226-234. PMID 31038409
  7. Reig M, et al. BCLC strategy for prognosis prediction and treatment recommendation: the 2022 update. J Hepatol. 2022;76:681-693. PMID 34801630
  8. Liang Y, et al. Diagnostic performance of LI-RADS for MRI and CT detection of HCC: a systematic review and diagnostic meta-analysis. Eur J Radiol. 2021;134:109404. PMID 33276248
  9. Kim YY, et al. Diagnostic performance of CT versus MRI LI-RADS category 5 for hepatocellular carcinoma: a systematic review and meta-analysis. Eur Radiol. 2022;32:6723-6729. PMID 35849177
  10. Roberts LR, et al. Imaging for the diagnosis of hepatocellular carcinoma: a systematic review and meta-analysis. Hepatology. 2018;67:401-421. PMID 28859233
  11. Qin Z, et al. Risk stratification for HCC of CEUS LI-RADS and diagnostic performance of LR-5 and LR-M: a systematic review and meta-analysis. Clin Radiol. 2022;77:e280-e286. PMID 35164930
  12. Kang JH, et al. Inter-reader reliability of contrast-enhanced ultrasound LI-RADS: a meta-analysis. Abdom Radiol (NY). 2021;46:4671-4681. PMID 34156509
  13. Silva MA, et al. Needle-track seeding following biopsy of liver lesions in the diagnosis of hepatocellular cancer: a systematic review and meta-analysis. Gut. 2008;57:1592-1596. PMID 18669577
  14. Calderaro J, et al. Deep learning-based phenotyping reclassifies combined hepatocellular-cholangiocarcinoma. Nat Commun. 2023;14:8290. PMID 38092727
  15. Beaufrère A, et al. Combined hepatocellular-cholangiocarcinoma: an update. J Hepatol. 2021;74:1212-1224. PMID 33545267
  16. Aslam A, et al. CT/MRI LI-RADS 2024 update: treatment response assessment. Radiology. 2024;313:e232408. PMID 39530896
  17. Mendiratta-Lala M, et al. LI-RADS treatment response algorithm with radiologic-pathologic explant correlation after SBRT. Int J Radiat Oncol Biol Phys. 2022;112:704-714. PMID 34644607
  18. Shenoy-Bhangle AS, et al. Accuracy of LI-RADS treatment response after percutaneous ablation with explant correlation. Cancer Imaging. 2025;25:65. PMID 40413459
  19. Wang Q, et al. Radiomics models for preoperative prediction of histopathological grade of HCC: a systematic review. Eur J Radiol. 2023;166:111015. PMID 37541183
  20. Zhou Y, et al. Risk stratification and distribution of HCC in CEUS and CT/MRI LI-RADS: a meta-analysis. Front Oncol. 2022;12:873913. PMID 35425706
  21. Shin J, et al. CEUS LI-RADS for diagnosing hepatocellular carcinoma: a meta-analysis. Liver Int. 2020;40:2345-2352. PMID 32722894
  22. Wang D, et al. Diagnostic performance of radiomics for differentiating intrahepatic cholangiocarcinoma from HCC: a meta-analysis. Acad Radiol. 2025;32:6555-6569. PMID 40571544
  23. Ye L, et al. Combined hepatocellular-cholangiocarcinoma: biology, diagnosis, and management. Liver Cancer. 2024;13:6-28. PMID 38344449
  24. Roßner F, et al. Pathology of combined hepatocellular carcinoma-cholangiocarcinoma: an update. Cancers. 2023;15. PMID 36672443
  25. Lyshchik A, et al. LI-RADS contrast-enhanced US nonradiation treatment response assessment version 2024. Radiology. 2024;311:e232369. PMID 38805727