The Past: The Evolution of LI-RADS	Choose Top of pageABSTRACTThe Past: The Evolution o... <<The Present: LI-RADS From...The Future: Not Just an I...ConclusionReferences

LI-RADS was initially developed by a small group of liver imaging experts convened by the American College of Radiology (ACR) to address the need for improved clarity and consistency of liver imaging reporting. The first version of LI-RADS (v1.0) was released in 2011. Modeled after BI-RADS, the well-established standardized reporting system for breast imaging, LI-RADS v1.0 introduced a five-category system for classifying the likelihood of HCC within a focal liver observation, spanning from LR-1 (definitely benign) to LR-5 (definitely HCC). A liver imaging lexicon providing initial definitions for terms such as “arterial phase hyperenhancement” and “late arterial phase” was also introduced [9].

Since v1.0 was released, LI-RADS has been serially updated and refined to incorporate new science, technology, and user feedback [10–15] (Table 1). The scope of LI-RADS has also been considerably expanded since 2011. For example, in the 2017 version (v2017), three new algorithms were introduced: ultrasound (US) LI-RADS for screening and surveillance, contrast-enhanced ultrasound (CEUS) LI-RADS for diagnosis, and LI-RADS CT/MRI Treatment Response for assessment after locoregional therapy. In this section, we focus on the evolution of an important entity, the enhancing “capsule” as a major feature for HCC, which provides an illustrative example of the nature of LI-RADS as a living document that continually improves.

View Larger Version

TABLE 1: Changes Made to LI-RADS Over Time

Enhancing “capsule” (which may or may not represent a true capsule) is one of five major imaging features of HCC in the 2018 version (v2018) of the LI-RADS CT/MRI diagnostic algorithm (Fig. 1). In LI-RADS v1.0, enhancing “capsule” was referred to as “portal venous or delayed phase ring.” It was considered an ancillary feature for HCC diagnosis and a trivial component of the diagnostic algorithm. In 2013, the same year that the Organ Procurement and Transplantation Network (OPTN) and the United Network for Organ Sharing published diagnostic criteria for HCC [16], LI-RADS promoted enhancing “capsule” from an ancillary imaging feature of HCC to a major imaging feature of HCC, although, at that time, it was referred to simply as “capsule.” As of late 2020, most other major HCC imaging systems do not incorporate enhancing “capsule” into their diagnostic criteria; however, the imaging feature remains highly relevant [17].

View larger version (132K)

Fig. 1 —CT/MRI LI-RADS (Liver Imaging Reporting and Data System) version 2018 diagnostic algorithm. This material is reprinted without modification with permission from American College of Radiology (© American College of Radiology; www.acr.org/-/media/ACR/Files/RADS/LI-RADS/LI-RADS-2018-Core.pdf), and pursuant to Creative Commons BY-NC-ND license and terms contained therein (creativecommons.org/licenses/by-nc-nd/4.0/), including disclaimer in Section 5. HCC = hepatocellular carcinoma.

A fibrous tumor capsule is an important feature of nodular progressed HCC that develops during the process of multistep hepatocarcinogenesis and is histologically present in approximately 70% of HCCs [18]. Pathologically, it is a specific finding in progressed HCC and is not seen in cirrhotic nodules, dysplastic nodules, early HCCs, intrahepatic cholangiocarcinomas, adenomas, or focal nodular hyperplasia [19]. Thus, the presence of an enhancing “capsule” at CT or MRI should in theory increase the specificity of HCC diagnosis. When LI-RADS version 2013 (v2013) was released, the high-quality scientific data available supported the change; capsule appearance was known to be moderately sensitive (42–64%) and highly specific (86–96%) for the diagnosis of HCC, and the performance characteristics of this finding improved with increasing lesion size [20, 21]. For this reason, enhancing “capsule” was elevated to a major feature of HCC in the LI-RADS v2013 algorithm.

Importantly, the presence of rim enhancement around a liver lesion in a patient at risk for HCC does not necessarily confirm the presence of a true fibrous tumor capsule. Not infrequently, a halo of hyperenhancement around a lesion may be seen at CT or MRI in the portal venous or delayed phases because of altered perfusion within compressed sinusoids in conjunction with the collapsed parenchyma, displaced by the expanding malignant mass. Such a finding is often referred to as a pseudocapsule and may mimic a true fibrous tumor capsule on imaging (Fig. 2). Although true capsule is very specific for HCC, a pseudocapsule may be seen with any rapidly growing mass. A reliable distinction between true capsule and pseudocapsule may not be possible based on imaging, but HCCs account for the majority of rapidly growing masses in cirrhotic livers [22]. As a result, LI-RADS incorporated enhancing “capsule” as a major feature of HCC. Furthermore, LI-RADS adopted the use of “capsule appearance” or “capsule,” with quotation marks included to acknowledge that the imaging appearance of a capsule may be due to either a true capsule or a pseudocapsule.

View larger version (360K)

Fig. 2 —55-year-old man with hepatitis C virus–related cirrhosis. Portal venous phase contrast-enhanced T1-weighted MRI with extracellular contrast agent shows enhancing “capsule” associated with hepatocellular carcinoma (arrow), which has smooth, uniform appearance and sharp borders.

Many users felt that the lexicon presented with the CT/MRI diagnostic algorithm of LI-RADS v2013 was ambiguous in its description of “capsule appearance.” At that time, LI-RADS simply stated “observations with ‘capsules' show unequivocal peripheral rim enhancement in portal venous phase or delayed phase” [10]. This definition lacked precision in describing how fibrous HCC capsules should appear at imaging and gave leeway to allow perfusion phenomena or corona enhancement to be characterized a “capsule” (Fig. 3). Such definitional imprecision risks reducing the diagnostic performance. The subsequent version of LI-RADS addressed this lack of clarity and gave an explicit definition for the expected morphology of an enhancing “capsule” as a “smooth, uniform, sharp border around most or all of an observation, unequivocally thicker or more conspicuous than fibrotic tissue around background nodules, and visible as an enhancing rim in the portal venous phase, delayed phase, or transitional phase” [11]. Although slightly verbose, this level of granularity is necessary to accurately characterize the feature and was driven by user feedback.

View larger version (294K)

Fig. 3 —61-year-old man with hepatitis C virus–related cirrhosis. Portal venous phase contrast-enhanced T1-weighted MRI with extracellular contrast agent shows corona enhancement (arrows) around hepatocellular carcinoma (HCC). Vague appearance of its nonuniform thickness is seen only around small part of HCC; this appearance should not be confused with enhancing “capsule” of HCC.

Evolution of enhancing “capsule” provides a compelling example of how the LI-RADS algorithm has evolved to incorporate science, user feedback, and new technology. Importantly, the need to keep LI-RADS up to date with the most recent scientific evidence and technologic advancements must be balanced against the need for stability; sufficient time between updates is necessary for users to build familiarity and for researchers to collect and publish data that can inform future improvements. As a compromise between these competing needs, major updates to LI-RADS algorithms are planned to occur every 3–5 years.

The Present: LI-RADS From 2018 to 2020	Choose Top of pageABSTRACTThe Past: The Evolution o...The Present: LI-RADS From... <<The Future: Not Just an I...ConclusionReferences

In 2018, LI-RADS eschewed its usual 3- to 5-year interlude between updates and released CT/MRI LI-RADS v2018, its most recent version of the diagnostic algorithm. In doing so, LI-RADS and the American Association for the Study of Liver Diseases (AASLD) guidance became unified with respect to LR-5 criteria for HCC diagnosis, and the LI-RADS algorithms were integrated into the AASLD guidance for HCC diagnosis, staging, and management [23]. This unification represented a milestone achievement for LI-RADS but necessitated an early, albeit small, update to the algorithm [24]. Modifications for the observations falling into the split cell were made from LI-RADS v2017 (Fig. 4) to LI-RADS v2018 (Fig. 5). The requirement for antecedent US visibility for 10- to 19-mm observations with nonrim arterial phase hyperenhancement (APHE) and “washout” was removed, and these observations are consequently categorized LR-5. Additionally, the definition of threshold growth was simplified to match that of the OPTN (≥ 50% increase in observation largest dimension in ≤ 6 months) [15] (Fig. 6). As a result of these changes, the -us and -g qualifiers for LR-5 nomenclature were removed, further simplifying the LR-5 category [24]. A large meta-analysis examining the performance characteristics of the LI-RADS v2018 algorithm showed that the newly adjusted LR-5 criteria remain highly specific for HCC in the diagnostic patient population [25].

View larger version (24K)

Fig. 4 —LI-RADS (Liver Imaging Reporting and Data System) v2017 CT/MRI split box algorithm for observations measuring 10–19 mm with nonrim arterial phase hyperenhancement. This material is reprinted without modification with permission from American College of Radiology (© American College of Radiology; www.acr.org/-/media/ACR/Files/RADS/LI-RADS/LIRADS_2017_Core.pdf), and pursuant to Creative Commons BY-NC-ND license and terms contained therein (creativecommons. org/licenses/by-nc-nd/4.0/), including disclaimer in Section 5. OPTN = Organ Procurement and Transplantation Network, AASLD = American Association for the Study of Liver Diseases, HCC = hepatocellular carcinoma.

View larger version (17K)

Fig. 5 —LI-RADS Liver Imaging Reporting and Data System) v2018 CT/MRI split box algorithm for observations measuring 10–19 mm with nonrim arterial phase hyperenhancement. This material is reprinted without modification with permission from American College of Radiology (© American College of Radiology; www.acr.org/-/media/ACR/Files/RADS/LI-RADS/LI-RADS-2018-Core.pdf), and pursuant to Creative Commons BY-NC-ND license and terms contained therein (creativecommons.org/licenses/by-nc-nd/4.0/), including disclaimer in Section 5.

View larger version (56K)

Fig. 6 —Comparison of definitions of threshold growth between LI-RADS (Liver Imaging Reporting and Data System) versions 2017 (v2017) and 2018 (v2018). This material is reprinted without modification with permission from American College of Radiology (© American College of Radiology; www.acr.org/-/media/ACR/Files/RADS/LI-RADS/LI-RADS-2018-Core.pdf), and pursuant to Creative Commons BY-NC-ND license and terms contained therein (creativecommons.org/licenses/by-ncnd/4.0/), including disclaimer in Section 5.

Two recent accomplishments of LI-RADS are free to download on the ACR website: the LI-RADS manual [26], and the LI-RADS lexicon [27]. The LI-RADS manual provides elaborations on the various LI-RADS algorithms and is a cogent source of information on various topics important to liver imaging such as cirrhosis, hepatocarcinogenesis, liver anatomy, and treatment response [26]. The manual serves as a comprehensive reference for imaging of patients with cirrhosis.

The LI-RADS lexicon was created to standardize the terminology used to describe liver imaging findings and is recommended for use in all research and educational materials relating to liver imaging (Table 2). Adoption of a universal liver imaging lexicon is important for a variety of reasons. Variability in terminology and definitions challenges the significance of the published literature, slowing progress. The recently recommended terminology for nodules without APHE that are hypointense in the hepatobiliary phase of a gadoxetate-enhanced MRI examination provides a good example [28]. These nodules represent an interesting and active area of scientific investigation; most reflect early HCC or high-grade dysplastic nodules, and a substantial minority transform to progressed HCC within 2 years of identification [29–31] (Fig. 7). However, whether detection of these nodules leads to improved overall survival in patients at risk for HCC is not yet known. One barrier to compiling the evidence regarding these nodules are the various terms by which they are referred to in scientific articles, such as “nonhypervascular hypointense nodules” or “hypovascular high-risk borderline lesions” [28]. In addition to variable terminology, many scientific works on such nodules use slight definitional variations. To address this issue, the LI-RADS Hepatobiliary Agent Working Group recommends standardized terminology for these nodules: “hepatobiliary phase hypointense nodules without APHE” [28]. A single term with a uniform definition improves the ability of scientific investigators to interpret and pool data relating to these nodules and assists radiologists and referring clinicians in interpreting their clinical significance.

View Larger Version

TABLE 2: LI-RADS Lexicon Key Points

View larger version (297K)

Fig. 7A —64-year-old man with history of cryptogenic cirrhosis. A, Contrast-enhanced T1-weighted MR images reveal observation in posterior right hepatic lobe (arrows) that is invisible in arterial (A) and portal venous (B) phases but appears as nodular region of hypointensity in hepatobiliary phase (C). This observation would be considered “hepatobiliary phase hypointense nodule without APHE” and would appropriately be classified LR-3.

View larger version (104K)

Fig. 7B —64-year-old man with history of cryptogenic cirrhosis. B, Contrast-enhanced T1-weighted MR images reveal observation in posterior right hepatic lobe (arrows) that is invisible in arterial (A) and portal venous (B) phases but appears as nodular region of hypointensity in hepatobiliary phase (C). This observation would be considered “hepatobiliary phase hypointense nodule without APHE” and would appropriately be classified LR-3.

View larger version (279K)

Fig. 7C —64-year-old man with history of cryptogenic cirrhosis. C, Contrast-enhanced T1-weighted MR images reveal observation in posterior right hepatic lobe (arrows) that is invisible in arterial (A) and portal venous (B) phases but appears as nodular region of hypointensity in hepatobiliary phase (C). This observation would be considered “hepatobiliary phase hypointense nodule without APHE” and would appropriately be classified LR-3.

View larger version (249K)

Fig. 7D —64-year-old man with history of cryptogenic cirrhosis. D, Within 18 months, nodule had increased in size and transformed into LR-5 hepatocellular carcinoma. Contrast-enhanced T1-weighted MR images show that lesion exhibits arterial phase hyperenhancement on arterial phase image (arrow, D), “washout”, and “capsule” on delayed phase image (arrow, E).

View larger version (264K)

Fig. 7E —64-year-old man with history of cryptogenic cirrhosis. E, Within 18 months, nodule had increased in size and transformed into LR-5 hepatocellular carcinoma. Contrast-enhanced T1-weighted MR images show that lesion exhibits arterial phase hyperenhancement on arterial phase image (arrow, D), “washout”, and “capsule” on delayed phase image (arrow, E).

The Future: Not Just an Imaging System	Choose Top of pageABSTRACTThe Past: The Evolution o...The Present: LI-RADS From...The Future: Not Just an I... <<ConclusionReferences

LI-RADS has well-articulated near-term goals and long-term objectives to ensure continued success and sustained growth. Continued updates to the LI-RADS diagnostic algorithms are planned to take place every 3–5 years to incorporate new knowledge, technology, and user feedback. Such updates require a step-wise expansion of high-quality data related to liver imaging. With numerous contributors across the world, LI-RADS is poised to play an active role in the facilitation of important research to progress the field of liver imaging. The new knowledge incorporated into LI-RADS is not limited to scientific discovery but will also take into account global shifts and geographic variations in the epidemiology of chronic liver disease and HCC. Consider liver transplantation in the United States: in 2010, hepatitis C virus infection was the most common cause of chronic liver disease among adults registered to liver transplant waiting lists [32]. In the decade since then, the number of patients in the United States with chronic hepatitis C virus infection with or without HCC awaiting liver transplant has precipitously declined because of effective antiviral therapy [32, 33]. During the same time period, the number of patients on liver transplant waiting lists with chronic liver disease related to nonalcoholic steatohepatitis sharply increased [34]. Nonalcoholic steatohepatitis represents the most rapidly rising cause of HCC in the United States and soon will become the most common cause of chronic liver disease for patients with HCC awaiting liver transplant [35].

Interestingly, in a substantial minority (25–46%) of patients with nonalcoholic fatty liver disease–associated HCC, liver malignancy develops without cirrhosis [36]. This situation poses a future diagnostic conundrum: these patients do not meet criteria for LI-RADS application given that the current LI-RADS CT/MRI diagnostic algorithm does not apply to all patients at increased risk for HCC. Rather, it applies to a precisely defined population of patients (i.e., patients with cirrhosis from a nonvascular cause, patients with chronic hepatitis B virus infection, and patients with known prior HCC) for whom the pretest probability of HCC is sufficiently high and the pretest probability of lesions mimicking HCC is sufficiently low, so that the LR-5 criteria reach near 100% specificity for HCC [37]. In the United States, high specificity is required because patients with LR-5 observations frequently undergo treatment, including liver transplant, without biopsy confirmation of HCC [23]. As high-quality evidence accrues, the LI-RADS diagnostic population could be expanded to include additional patient populations at increased risk for HCC, so long as the LR-5 criteria remain highly specific for HCC diagnosis. The performance of the LI-RADS diagnostic algorithm for diagnosis of HCC in patients without cirrhosis at increased risk for HCC (e.g., patients with chronic hepatitis C virus infection or nonalcoholic steatohepatitis with advanced fibrosis) is an example of an important knowledge gap warranting further research that has been identified and must be addressed.

Since 2011, the scope of LI-RADS has been considerably expanded. At inception, LI-RADS v1.0 included two imaging modalities (CT and MRI with extracellular agents) and a single clinical indication (HCC diagnosis and staging). In its current state, LI-RADS encompasses five imaging modalities (the original two plus MRI with hepatobiliary contrast agents, ultrasound, and contrast-enhanced ultrasound) for three clinical indications (the original indication as well as HCC screening and surveillance and HCC treatment response assessment). Although HCC assessment accounts for a substantial proportion of indications for undergoing liver imaging, numerous other causes necessitate liver imaging. LI-RADS is planning a continued expansion in its scope to address indications for liver imaging that are not related to HCC. LI-RADS has convened or is convening working groups to address topics such as benign liver lesions, pediatric liver lesions, and quantitative liver imaging. The LI-RADS Pediatric Working Group recently released consensus imaging recommendations for hepatoblastoma and pediatric HCC [38]. This work includes the formulation of guidelines related to imaging modality selection, imaging technique and protocols, appropriate contrast agent use, and standardized liver imaging reporting for pediatric patients.

LI-RADS is broadly accepted at academic centers and is a common subject of research and scientific literature, but global adoption in nonacademic clinical settings is lagging. Known barriers to adoption include radiologist refusal, inconsistent system use, failure of system integration into patient management paradigms, preference for other HCC imaging systems (e.g., OPTN or specific regional guidelines), terminology complexity, and the high number of LI-RADS categories [6].

Simplification of the algorithm may improve efficiency with minimal effect on the system's specificity and accuracy [39]. For instance, using three size cut-offs (< 10 mm, 10–19 mm, and ≥ 20 mm) for observations with nonrim APHE may not be necessary; one size cutoff (i.e., < 10 mm vs ≥ 10 mm) may be appropriate. A meta-analysis that included 19 publications reported the specificity of CT and MRI with an extracellular contrast agent to be similar for HCCs measuring 10–20 mm (88% for CT and 87% for MRI) and those that were larger than 20 mm (90% for CT and 87% for MRI) [40]. On the other hand, the specificity for HCCs < 10 mm with an extracellular contrast agent was considerably lower (69% for CT and 46% for MRI) [40]. Thus, data suggest that the 20-mm threshold for observations with nonrim APHE could be removed.

Another potential area for algorithm simplification is changes to ancillary feature application [39]. The use of these features was made optional at the release of LI-RADS v2017 to decrease algorithm complexity. However, this change increased interreader variability for the final LI-RADS category. Eliminating ancillary features that are redundant or rarely encountered in practice may help in system simplification [39]. Additionally, in most practices, ancillary features are used commonly to adjust categories between LR-3 and LR-4 but rarely to adjust categories for observations that are scored as definitely or probably benign (LR-1 and LR-2) or definitely HCC (LR-5) on the basis of major features [39]. Restricting use of ancillary features only to adjustments between LR-2, LR-3, and LR-4 categories might be practical. If rarely used ancillary features are eliminated and ancillary feature application is limited to a subset of LI-RADS categories, use of ancillary features could be modified from optional to mandatory. Such changes may improve interreader agreement without further complicating the system [39].

Since its initial release, LI-RADS has provided a CT/MRI diagnostic algorithm that is equally applicable to both modalities. However, the evidence shows that LI-RADS categories assigned by either modality are different in 36–71% of observations [41–44]. When observations are imaged with both MRI and CT, 49% are seen only on MRI, whereas only 2% are seen only on CT [42]. Additionally, observations are categorized as LR-1 less frequently with CT than with MRI [42, 44]. Exclusive of the LR-1 category, the LI-RADS category on MRI is higher than that assigned on CT in 25–31% of observations [42, 44]. Conversely, the category assigned on CT is higher than the category assigned on MRI in 5–12% of observations [42, 44]. LR-5 observations seen on MRI also meet LR-5 criteria on CT 42–59% of the time [42, 44] (Fig. 8). On the other hand, LR-5 observations seen on CT also meet LR-5 criteria on MRI 92–100% of the time [42, 44].

View larger version (322K)

Fig. 8A —72-year-old man with history of hepatitis C cirrhosis. A, Axial contrast-enhanced CT images in arterial (A) and delayed (B) phases show 23-mm observation with no arterial phase hyperenhancement (arrow, A) and with “washout” (arrow, B). Combination of major features results in LR-4 (probable hepatocellular carcinoma) category.

View larger version (366K)

Fig. 8B —72-year-old man with history of hepatitis C cirrhosis. B, Axial contrast-enhanced CT images in arterial (A) and delayed (B) phases show 23-mm observation with no arterial phase hyperenhancement (arrow, A) and with “washout” (arrow, B). Combination of major features results in LR-4 (probable hepatocellular carcinoma) category.

View larger version (298K)

Fig. 8C —72-year-old man with history of hepatitis C cirrhosis. C, Arterial (C) and delayed (D) phase contrast-enhanced 1.5-T MR images with extracellular contrast agent performed 10 days later shows nonrim arterial phase hyperenhancement (arrow, C), delayed phase “washout” (long arrow, D) and enhancing “capsule” (short arrow, D). Combination of major features results in LR-5 (definite hepatocellular carcinoma) category.

View larger version (278K)

Fig. 8D —72-year-old man with history of hepatitis C cirrhosis. D, Arterial (C) and delayed (D) phase contrast-enhanced 1.5-T MR images with extracellular contrast agent performed 10 days later shows nonrim arterial phase hyperenhancement (arrow, C), delayed phase “washout” (long arrow, D) and enhancing “capsule” (short arrow, D). Combination of major features results in LR-5 (definite hepatocellular carcinoma) category.

Multiple reasons have been postulated to cause the discrepancies between the two modalities. First, CT and MRI differ in their ability to depict the major features of HCC [40, 45]. Intermodality agreement is low, with reported kappa values of 0.20 for APHE, 0.38 for “washout,” and 0.26 for “capsule” [41, 44, 46]. Second, many ancillary feature used for category adjustment are only assessable with MRI; three features of LR-M (targetoid hepatobiliary phase appearance and targetoid restriction for targetoid LR-M and marked restriction for nontargetoid LR-M), and six ancillary features (restricted diffusion, mild-moderate T2 hyperintensity, transitional phase hypointensity, hepatobiliary phase hypointensity, marked T2 hyperintensity, and hepatobiliary phase isointensity) are specific to MRI. Furthermore, three of the noted ancillary features apply to hepatobiliary agents only [15]. Inherent differences between the two modalities in diagnostic performance and imaging feature assessment make resolving differences in final LI-RADS category difficult [39]. As the body of evidence continues to grow, the exact probabilities of malignancy and HCC may be accepted for both CT and MRI and their combined major and ancillary features. Reporting the probability of HCC instead of LI-RADS categories might be the preferred answer to discordance between CT and MRI.

The discrepancy between CT and MRI in their ability to detect major imaging features offers a chance to explore integration of the two modalities. Combining major and ancillary features between CT and MRI may allow a more accurate LI-RADS category assignment than what can be achieved with the modalities independently [39]. For instance, if “capsule” and “APHE” are seen on MRI for a 19-mm observation, it would be categorized as LR-4. If “washout” seen on CT, performed in a short time interval, is added to the MRI findings for the same observation, it would now be categorized as LR-5. Although complementary information integration from CT and MRI may seem appropriate, no empirical evidence currently supports it. Further research is necessary to validate cross-modality integration and establish the maximal allowable interval between performance of the two studies [39].

Continued worldwide adoption is an aim of LI-RADS that is of paramount importance. The LI-RADS International Working Group has translated CT/MRI LI-RADS v2018 into 12 languages to increase its availability for use across the globe. Translations are available in Spanish, Portuguese, Farsi, German, French, Italian, Korean, Vietnamese, Japanese, and Chinese (traditional and simplified) [13–15]. The US and CEUS algorithms are actively being translated. LI-RADS plans to continue this translational work and steadily expand its geographic footprint.

As described, integration of LI-RADS into the AASLD HCC guidance was a major achievement [23]. However, global unification of diagnostic systems of HCC is still a major ultimate goal. Many other diagnostic systems exist in the United States (e.g., OPTN) and worldwide [17, 47–49]. These other systems are similar, but they have differences stemming from their target populations and clinical, cultural, and socioeconomic contexts [17]. Integration of the various systems is challenging, partly because of regional priorities of HCC treatment. In North America, liver transplant, when feasible, is frequently pursued as a form of curative therapy for HCC. Given this, transplant-focused countries desire high specificity in an HCC diagnostic algorithm to ensure that allocation of liver allografts, a scarce resource, is fair and appropriate [39]. Conversely, in many Asian countries, hepatic resection and locoregional ablative therapies are often preferred to liver transplant for treatment of HCC. Because of a higher prevalence of chronic hepatitis B virus infection, a larger proportion of patients with HCC in Asian countries develop cancer before advanced medical liver disease and may tolerate large hepatic re-sections that may be curative; given this different management paradigm, maximal sensitivity is preferred when it comes to an HCC diagnostic algorithm [17]. In light of these regional differences, creating a unified system with region-specific management guidelines could be a practical approach for global unification [39]. For instance, in regions prioritizing sensitivity for HCC diagnosis, the same management strategy could be applied to both LR-4 and LR-5 observations such that both may undergo therapy without requiring histologic confirmation; combining these categories has been shown to improve the sensitivity of the LI-RADS algorithm for HCC diagnosis [50–52]. On the other hand, in regions that prioritize orthotopic liver transplant for treatment of HCC, using only the LR-5 criteria to diagnose HCC without a requirement of histologic confirmation would be preferred.

As advanced technologies emerge, they will undoubtedly improve the clinical application of LI-RADS. For instance, the standardized algorithm and lexicon of LI-RADS allow easy integration into a clinical decision support framework with common data elements. The LI-RADS Assist module developed by ACR accurately translates the LI-RADS algorithms into a common data elements schema, which can be implemented into commercial dictation software [38]. This capability allows the interpreting radiologist to input the key data for the observation that they would in a typical LI-RADS report, and the module will output the correct LI-RADS category. As natural language processing techniques are refined, the key verbal inputs may be extracted from free dictation. Widespread implementation of common data elements will allow collection of regionally diverse data for practice improvement, registry development, and multiinstitutional studies.

Continued development and widespread integration of artificial intelligence in radiology practice could eventually allow automated lesion tracking over time [39]. This advantage is particularly useful for patients at risk for HCC because many undergo serial examinations, which may detect numerous observations with various LI-RADS categories. Locoregional therapy further complicates serial follow-up; various observations may be treated using different treatment modalities and may show variable response at different time points. Use of artificial intelligence to track LI-RADS observations may improve the accuracy of radiologists' interpretation of liver studies. As deep learning techniques continue to advance, they can be implemented to detect new observations and assist with imaging feature characterization and category assignment.

Finally, in the distant future, as artificial intelligence systems evolve, in conjunction with worldwide adoption of a standardized diagnostic system for the diagnosis of HCC, large registries and datasets in global cloud systems may allow integration of clinical and imaging data that could determine the probability of HCC, or another malignancy, for a specific individual on the basis of demographics, geographic location, laboratory tests, and clinical history.

Conclusion	Choose Top of pageABSTRACTThe Past: The Evolution o...The Present: LI-RADS From...The Future: Not Just an I...Conclusion <<References

In 2011, the newly released LI-RADS v1.0 represented a callow imaging algorithm to be used for a solitary imaging indication. Since that time, it has enjoyed substantial growth in scope and utilization, unified with important societies such as the AASLD, and improved the care of patients with liver disease undergoing liver imaging. The evolution of the LI-RADS algorithms has largely been driven by high-quality scientific data and feedback from users, and the continued incorporation of both will drive future LI-RADS iterations. Aside from the continued refinement of its algorithms, LI-RADS aspires to progress the field of liver imaging by facilitating important future research and the unification of liver imaging systems, two objectives that go hand in hand.