About the CT/MRI LI-RADS version 2018
Check tumour in vein first, which gives LR-TIV outright, then LR-M features, which override the table. After that it is a lookup on three things: nonrim arterial phase hyperenhancement, observation size, and how many of the three additional major features are present — nonperipheral washout, enhancing capsule and threshold growth. Without nonrim APHE nothing reaches beyond LR-4 at any size. With it, an observation under 10 mm also caps at LR-4. At 10–19 mm the tie-break matters: one additional feature gives LR-5 if it is washout or threshold growth, but only LR-4 if it is an enhancing capsule alone. At 20 mm or more, nonrim APHE alone is LR-4 and any single additional feature makes it LR-5.
Formula
1. Tumour in vein -> LR-TIV
2. LR-M features -> LR-M
3. Otherwise, look up the table:
NONRIM APHE PRESENT 0 features 1 feature 2 or more
under 10 mm LR-3 LR-4 LR-4
10-19 mm LR-3 LR-5, but LR-4 LR-5
if capsule alone
20 mm or more LR-4 LR-5 LR-5
NO APHE (or rim only) 0 features 1 feature 2 or more
under 20 mm LR-3 LR-3 LR-4
20 mm or more LR-3 LR-4 LR-4
Additional major features = nonperipheral washout,
enhancing capsule,
threshold growth- Nonrim APHE
- The single most important input. Its absence caps the observation at LR-4 whatever the size and however many other features are present.
- The 10 mm floor
- No observation under 10 mm can be LR-5, even with all three additional features. Sub-centimetre lesions are too often benign for that specificity to hold.
- The capsule tie-break at 10–19 mm
- With exactly one additional feature at this size, washout or threshold growth gives LR-5 while an enhancing capsule alone gives LR-4. This is the cell most often got wrong from memory.
- Threshold growth
- 50% or more increase in size in 6 months or less — a single rule in version 2018, where version 2017 had several branches depending on the interval.
- LR-TIV and LR-M are checked before the table and override an otherwise LR-5 appearance.
- LR-1 (definitely benign), LR-2 (probably benign) and LR-NC (not categorisable) are assigned outside the diagnostic table.
- Ancillary features can support upgrading or downgrading in the full LI-RADS framework but do not enter the core table, and cannot be used to reach LR-5.
- LI-RADS applies only to the defined at-risk population; the categories are not validated elsewhere.
Interpreting the result
LR-5 permits a non-invasive diagnosis of hepatocellular carcinoma; refer to the multidisciplinary meeting for staging and treatment allocation, and do not biopsy for confirmation. Before accepting it, confirm that no LR-M or tumour-in-vein feature was overlooked, since both override the category and both change management completely — LR-M usually needs a biopsy, and macrovascular invasion generally precludes transplantation and resection. LR-4 is not sufficient for a non-invasive diagnosis on its own, and management ranges from short-interval reimaging to biopsy to treatment depending on the clinical context and how close the observation sits to LR-5; multidisciplinary discussion is the right route. LR-3 is the category with the widest range of underlying pathology and the least clear path, and the usual course is repeat imaging at roughly three to six months, or an alternative modality or contrast agent, since a proportion of LR-3 observations resolve once better characterised. Where the patient turns out not to be in the at-risk population, do not assign a category at all — report the observation descriptively, because in a non-cirrhotic liver the differential is different and a category implying a high probability of hepatocellular carcinoma would mislead the clinician reading it.
| Score | Band | What it means | Action |
|---|---|---|---|
| LR-TIV | Tumour in vein | Unequivocal enhancing soft tissue within a vein | State whether it is contiguous with a parenchymal mass; macrovascular invasion generally precludes transplant and resection |
| LR-M | Probably or definitely malignant, not HCC specific | Targetoid or other features suggesting cholangiocarcinoma or a combined tumour | Biopsy is usually required; LR-M overrides an otherwise LR-5 appearance |
| LR-5 | Definitely hepatocellular carcinoma | Near-total specificity by design | Non-invasive diagnosis; refer for staging and treatment without biopsy |
| LR-4 | Probably hepatocellular carcinoma | Not sufficient for a non-invasive diagnosis alone | Multidisciplinary discussion; short-interval reimaging, biopsy or treatment depending on context |
| LR-3 | Intermediate probability of malignancy | The widest range of underlying pathology | Repeat imaging at about 3–6 months, or an alternative modality or contrast agent |
| Not applicable | Outside the at-risk population | No cirrhosis, chronic hepatitis B, or current or prior hepatocellular carcinoma | Report descriptively; do not assign a LI-RADS category |
What the LI-RADS v2018 (CT/MRI) needs (8 inputs)
- At-risk population — cirrhosis, chronic hepatitis B, or current/prior HCC
- LI-RADS excludes patients under 18, and those whose cirrhosis is due to congenital hepatic fibrosis or to vascular disorders. Outside the at-risk population the categories carry no validated predictive value and should not be reported.
- Tumour in vein
- Unequivocal enhancing soft tissue within a vein. Assigned LR-TIV and takes precedence over the diagnostic table regardless of every other feature.
- LR-M features
- Rim arterial phase hyperenhancement, peripheral washout, delayed central enhancement, targetoid appearance or restriction, infiltrative appearance, marked diffusion restriction or necrosis. LR-M means probably or definitely malignant but not specific for hepatocellular carcinoma, and it is checked before the table.
- Nonrim arterial phase hyperenhancement
- Nonrim enhancement in the arterial phase unequivocally greater than background liver. This is the entry point to the higher categories — without it, no observation of any size can exceed LR-4 on this table.
- Observation size
- Measured on the sequence where the margins are sharpest. The bands are under 10 mm, 10–19 mm and 20 mm or more when nonrim APHE is present, and under 20 mm and 20 mm or more when it is not.
- Nonperipheral washout
- A temporal reduction in enhancement relative to liver from an earlier to a later phase, and not at the periphery. Peripheral washout is an LR-M feature instead, so the distinction changes the category rather than refining it.
- Enhancing capsule
- A uniform, sharp, smooth rim of hyperenhancement around most or all of the observation, unequivocally thicker or more conspicuous than surrounding fibrotic tissue. The one major feature that cannot on its own lift a 10–19 mm observation to LR-5.
- Threshold growth
- An increase in size of 50% or more within 6 months or less. Version 2018 simplified this from the multi-branch version 2017 definition to this single rule.
What it returns
- LI-RADS category
- LR-3, LR-4, LR-5, LR-M or LR-TIV. LR-1, LR-2 and LR-NC are assigned outside this table and are not computed here.
- Size band and feature count
- Reported explicitly, since these two determine the cell and are where transcription errors occur.
- What the category means for management
- LR-5 permits a non-invasive diagnosis; LR-4 generally warrants multidisciplinary discussion; LR-3 usually means short-interval reimaging.
How it is calculated
Hepatocellular carcinoma is one of the few solid cancers routinely diagnosed and treated without histology, and that is possible because the imaging appearance in a cirrhotic liver is distinctive. As a regenerative nodule progresses to dysplasia and then to carcinoma, its blood supply shifts: portal venous inflow falls away and unpaired arterial supply increases, which produces arterial phase hyperenhancement followed by relative washout as the surrounding liver enhances later. A fibrous pseudocapsule develops around the expanding lesion, and the lesion grows. Those four observations — arterial hyperenhancement, washout, capsule and growth — are exactly the major features in the table, and the categories are an ordered statement of how confidently that sequence has been observed. The size bands exist because the same features mean different things at different scales: a 6 mm arterially enhancing focus in a cirrhotic liver is far more often a perfusion anomaly or a regenerative nodule than a cancer, whereas the same features at 3 cm rarely are. LR-M sits outside this logic because cholangiocarcinoma and combined tumours enhance in a different pattern — rim rather than nonrim, peripheral rather than nonperipheral washout — and calling those hepatocellular carcinoma would be a serious error, which is why they are checked first.
Facts & figures
| Arterial phase | Size | 0 additional features | 1 additional feature | 2 or more |
|---|---|---|---|---|
| Nonrim APHE | Under 10 mm | LR-3 | LR-4 | LR-4 |
| Nonrim APHE | 10–19 mm | LR-3 | LR-5 — but LR-4 if the single feature is an enhancing capsule | LR-5 |
| Nonrim APHE | 20 mm or more | LR-4 | LR-5 | LR-5 |
| No APHE, or rim only | Under 20 mm | LR-3 | LR-3 | LR-4 |
| No APHE, or rim only | 20 mm or more | LR-3 | LR-4 | LR-4 |
Additional major features are nonperipheral washout, enhancing capsule and threshold growth. LR-TIV and LR-M are assigned before this table and override it.
| Feature | Definition | What it reflects |
|---|---|---|
| Nonrim APHE | Nonrim arterial enhancement unequivocally greater than background liver | Unpaired arterial supply replacing portal inflow during carcinogenesis |
| Nonperipheral washout | Temporal reduction in enhancement relative to liver from an earlier to a later phase | Loss of portal venous supply; peripheral washout instead suggests a non-HCC malignancy |
| Enhancing capsule | Uniform, sharp, smooth rim thicker or more conspicuous than surrounding fibrosis | A fibrous pseudocapsule around an expansile lesion |
| Threshold growth | 50% or greater increase in size in 6 months or less | Behaviour over time, added where a single study is equivocal |
The definition of threshold growth was simplified in version 2018; version 2017 used several branches depending on the imaging interval.
Evidence
Derivation — ACR LI-RADS Steering Committee, version 2018
2018Expert consensus of the American College of Radiology LI-RADS Steering Committee, published in Radiology in 2018 and harmonised with the AASLD hepatocellular carcinoma guidance of the same year.
LR-5 is constructed for very high specificity rather than balanced accuracy, since the category is intended to license treatment without histology. Reported specificity for hepatocellular carcinoma is consistently in the mid-90s or higher, with correspondingly moderate sensitivity.
Harmonisation with AASLD 2018
2018AASLD guidance on the diagnosis, staging and management of hepatocellular carcinoma.
Version 2018 was explicitly aligned with AASLD, so an LR-5 observation satisfies the guideline's imaging criteria for a non-invasive diagnosis of hepatocellular carcinoma in an at-risk patient.
Reported diagnostic performance across modalities
2018Series applying the major features at CT and at MRI in at-risk populations.
Nonperipheral washout is detected less often at CT than at MRI, which is one reason the same observation can categorise differently between modalities and why a change of modality is a reasonable next step for an LR-3.
How it compares
LI-RADS v2018 (CT/MRI) vs Milan criteria
LI-RADS decides whether a lesion is hepatocellular carcinoma; Milan decides whether the patient can be transplanted for it.
The two sit in sequence rather than in competition. LI-RADS operates per observation and answers a diagnostic question, using enhancement pattern and size to reach a category. Milan operates per patient and answers an allocation question, using the number and size of confirmed tumours plus the absence of vascular invasion and extrahepatic spread. An LR-5 observation is what allows a lesion to be counted for Milan without biopsy — and an LR-TIV finding is what usually takes the patient outside Milan altogether, since macrovascular invasion is an absolute exclusion.
LI-RADS v2018 (CT/MRI) vs ALBI grade
Complementary halves of the same decision — what the tumour is, and what the liver can tolerate.
LI-RADS characterises the observation and says nothing about hepatic reserve. ALBI grades hepatic function from bilirubin and albumin alone, without the subjective ascites and encephalopathy components of Child-Pugh, and is what determines whether resection, ablation or transarterial therapy is survivable once an LR-5 lesion has been identified. In practice a multidisciplinary meeting needs both: a category from the imaging and a functional grade from the bloods.
LI-RADS v2018 (CT/MRI) vs Child-Pugh score
Defines much of the at-risk population that LI-RADS is validated in, and constrains what can be done about an LR-5 finding.
Cirrhosis is the commonest route into the LI-RADS at-risk population, and Child-Pugh class is the traditional measure of how decompensated that cirrhosis is. The relationship runs both ways: the presence of cirrhosis is what makes LI-RADS applicable at all, and the severity of it is what determines whether an identified hepatocellular carcinoma can be resected, ablated, embolised or only managed supportively. An LR-5 lesion in a Child-Pugh C liver is a different clinical problem from the same lesion in a Child-Pugh A liver.
Pearls & pitfalls
- Check the at-risk population first. Outside it, LI-RADS categories are not validated and should not be reported.
- At 10–19 mm with one additional feature, an enhancing capsule alone gives LR-4, while washout or threshold growth gives LR-5.
- Nothing under 10 mm can be LR-5, even with all three additional features present.
- Without nonrim APHE, no observation reaches beyond LR-4 at any size.
- Rim APHE is not nonrim APHE — it is an LR-M feature and points away from hepatocellular carcinoma.
- Peripheral washout is likewise an LR-M feature; only nonperipheral washout counts as a major feature.
- LR-M and LR-TIV are checked before the table and override an otherwise LR-5 appearance.
- Threshold growth in version 2018 is a single rule — 50% or more in 6 months or less.
- Ancillary features can support upgrading within the wider framework but never to LR-5, and they do not enter the core table.
- Washout is detected less often at CT than at MRI, so the same observation can categorise differently between modalities.
- LR-5 licenses treatment without biopsy, which makes an error in either direction consequential.
Critical actions
- Confirm the patient meets the LI-RADS at-risk definition before assigning any category.
- Look for tumour in vein and LR-M features before applying the diagnostic table.
- Distinguish rim from nonrim arterial phase hyperenhancement, and peripheral from nonperipheral washout.
- Measure the observation on the sequence with the sharpest margins and record which band it falls in.
- When exactly one additional feature is present at 10–19 mm, record which feature it is.
- State the category explicitly in the report, alongside the features that produced it.
- Refer LR-5 observations for multidisciplinary staging without requesting a confirmatory biopsy.
- Arrange short-interval reimaging or an alternative modality for LR-3, and multidisciplinary discussion for LR-4.
- For a non-at-risk patient, report descriptively rather than assigning a category.
Why this score exists
The capsule tie-break at 10–19 mm is the most revealing thing in the table, because it is the one place where the count of features is not enough and the identity of the feature matters. Everywhere else, LI-RADS treats washout, capsule and threshold growth as interchangeable units to be added up. At 10–19 mm with exactly one of them, it stops doing that: washout or threshold growth reaches LR-5, and an enhancing capsule alone does not. The reason is that an enhancing capsule is the least specific of the three at that size — fibrotic pseudocapsules form around benign nodules in a cirrhotic liver too — and LR-5 is the category that permits treating a patient for cancer without ever obtaining tissue. A committee willing to complicate an otherwise clean lookup table for one cell is a committee that understands what its own category is being used for. It is also, predictably, the cell that gets recalled wrongly most often, which is exactly why the fixtures behind this calculator pin every cell of the table rather than a representative sample.
About the creator
First author, CT/MRI LI-RADS version 2018
Led the version 2018 publication and its harmonisation with AASLD guidance.
Co-author; abdominal imaging and LI-RADS Steering Committee
Contributed to the major and ancillary feature definitions.
LI-RADS founding contributor; liver imaging standardisation
Central to the development of LI-RADS across its successive versions.
Limitations
- Expert consensus rather than a fitted model, so the category boundaries are agreed rather than derived from outcome data.
- Several major features are subjective — whether enhancement is unequivocally greater than liver, and whether a capsule is unequivocally more conspicuous than surrounding fibrosis — and interobserver agreement is imperfect.
- Washout is detected less often at CT than at MRI, so the same observation can categorise differently by modality.
- The core table excludes ancillary features, which many radiologists use in practice and which can shift a category outside the table.
- LR-3 is heterogeneous and the framework gives limited guidance on what to do with it.
- Categories are not validated outside the at-risk population and can mislead if applied there.
- Size measurement is operator-dependent and matters most exactly where the bands change, at 10 mm and 20 mm.
- LR-5 is tuned for specificity, so its sensitivity is moderate and a proportion of hepatocellular carcinomas are categorised lower.
If you are the patient
LI-RADS is a scoring system radiologists use when they look at a liver scan in someone at higher risk of liver cancer — usually people with cirrhosis, long-standing hepatitis B, or a liver cancer in the past. It gives each spot found on the scan a category, and the category says how likely that spot is to be a liver cancer. The main categories run from LR-3, which means uncertain and usually leads to another scan in a few months, through LR-4, which means probably cancer and needs a team discussion, to LR-5, which means the appearance is characteristic enough that it can be treated as liver cancer without a biopsy. There are two extra categories: LR-M, meaning it looks like a cancer but not the usual liver cancer type, which normally does need a biopsy; and LR-TIV, meaning tumour has grown into a vein. The categories come from a few specific things on the scan — whether the spot lights up when contrast dye is given, how big it is, whether the dye fades from it faster than from the surrounding liver, whether there is a rim around it, and whether it has grown by half or more in six months. The reason LR-5 matters so much is that it means treatment can start without a biopsy, which spares an invasive test. It is also why radiologists are careful about the boundaries, and why the same spot can be scored differently on a CT and an MRI scan — MRI shows some of these features more clearly. If you have a report with one of these categories on it, it is worth asking your team what the plan is for that category specifically, since each one leads somewhere different.
Frequently asked questions
What is LI-RADS v2018?#
The American College of Radiology's system for categorising liver observations on multiphase CT or MRI in patients at risk of hepatocellular carcinoma. The version 2018 CT/MRI algorithm assigns LR-1 to LR-5 plus LR-M and LR-TIV, with the core diagnostic table driven by nonrim arterial phase hyperenhancement, observation size, and the count of additional major features.
What are the criteria for LR-5?#
Nonrim arterial phase hyperenhancement, size of at least 10 mm, and at least one additional major feature — nonperipheral washout, enhancing capsule or threshold growth. The exception is at 10–19 mm, where an enhancing capsule alone is not sufficient and gives LR-4; washout or threshold growth alone does reach LR-5.
Why can't a lesion under 10 mm be LR-5?#
Because at that size the major features are not specific enough. Sub-centimetre arterially enhancing foci in a cirrhotic liver are frequently perfusion anomalies or regenerative nodules, and LR-5 is designed for near-total specificity since it permits treatment without a biopsy. A sub-centimetre observation caps at LR-4 however many additional features it has.
What is the difference between an enhancing capsule and the other major features at 10–19 mm?#
Specificity. Fibrotic pseudocapsules form around benign nodules in a cirrhotic liver, so a capsule alone at that size does not carry the confidence LR-5 requires — it gives LR-4. Nonperipheral washout or threshold growth alone does reach LR-5. It is the only cell in the table where the identity of the feature matters rather than just the count, and it is the one most often misremembered.
What is threshold growth in version 2018?#
An increase in size of 50% or more within 6 months or less. Version 2018 simplified this to a single rule; version 2017 had several branches depending on the imaging interval, which made it harder to apply consistently.
What does LR-M mean and why is it checked first?#
LR-M means probably or definitely malignant but not specific for hepatocellular carcinoma — the main considerations are intrahepatic cholangiocarcinoma and combined HCC-cholangiocarcinoma. Its features include rim arterial hyperenhancement, peripheral washout, targetoid appearance, infiltrative appearance and marked diffusion restriction. It is checked before the diagnostic table because those tumours can otherwise produce an LR-5 appearance, and treating them as hepatocellular carcinoma without a biopsy would be a serious error.
Can LI-RADS be used in a patient without cirrhosis?#
Not meaningfully. LI-RADS is defined for patients with cirrhosis, chronic hepatitis B, or current or prior hepatocellular carcinoma, and explicitly excludes those under 18 and those whose cirrhosis is due to congenital hepatic fibrosis or vascular disorders. Outside that population the categories are not validated, and a category implying a high probability of hepatocellular carcinoma would mislead. Report the observation descriptively instead.
Why might the same lesion get different categories on CT and MRI?#
Chiefly because nonperipheral washout is detected less often at CT than at MRI. Since washout is one of the three additional major features, missing it can drop an observation a category — most consequentially at 10–19 mm, where washout alone is the difference between LR-4 and LR-5. This is also why changing modality or contrast agent is a reasonable next step for an LR-3 observation.
Does an LR-5 observation need a biopsy?#
No. That is the point of the category. LR-5 is constructed for very high specificity so that a non-invasive diagnosis can be made, and version 2018 was harmonised with AASLD guidance so that an LR-5 observation satisfies the imaging criteria for diagnosing hepatocellular carcinoma in an at-risk patient. Confirm that no LR-M or tumour-in-vein feature was overlooked, then refer for staging and treatment.
References
Original / primary reference
Clinical practice guidelines
- Marrero JA, Kulik LM, Sirlin CB, Zhu AX, Finn RS, Abecassis MM, et al. Diagnosis, Staging, and Management of Hepatocellular Carcinoma: 2018 Practice Guidance by the American Association for the Study of Liver Diseases. Hepatology. 2018;68(2):723-750.
- European Association for the Study of the Liver. EASL Clinical Practice Guidelines: Management of hepatocellular carcinoma. J Hepatol. 2018;69(1):182-236.