Good afternoon, everybody. Welcome to our CNA and thank you so much for coming to our session. My name is Victoria Chernek and I'm going to take a minute to kind of introduce our session and then I'll go into my talk. So as some of you know, 2021 is a big year for Lyraids because 10 years ago, the first version of Lyraids was released. And in the past 10 years, Lyraids grew into this large expanse ecosystem with multiple components. And the correlation of components to each other can be well demonstrated by Russian nesting doll. And if we think of the outside, the largest doll is our lexicon. Inside of it, the four algorithms, which then are supported by things like core materials and a manual. All of these components allow us to create standardized reporting, will eventually allow us to create large data registries. And then the core of all this ecosystem is a very important pretty doll, which signifies improved patient's care and outcome. And we can demonstrate again, this ecosystem by concentric rings, where the outside rings encompasses the more internal rings. And you can see that there's a lot to cover about Lyraids. And obviously we don't have time to cover everything. So what we're going to do today is we're going to discuss Lyraids lexicon, which I will do. Dr. Claude Serlin will follow me and discuss CTMR Lyraids. Dr. David Fetzer will then follow and discuss TUS Lyraids. And Professor Jianmin Li will discuss treatment response algorithm. And at the end, we will take your questions. So without further ado, I'll go into my talk, which is on Lyraids lexicon. And as I mentioned before, Lyraids is 10 years old. And you can see that in the past 10 years, there have been quite a number of updates and transformations and nuances that have happened. But no matter what changes have been introduced, the core approach to categorizing live observations in patients at risk of HTC stayed the same. And what we have is we have building blocks, which are imaging features. And then we use a CTMR diagnostic table to put together these building blocks. And once we put them together, that will inform us of the category. And then the category will give us information on probability of malignancy and HTC assigned to this lesion. Now, you may be wondering, when we were creating Lyraids, how did we come up with these combinations? And how did we come up with this combination means this category, and this category means this probability? And the answer is there were some opinions involved, but really we tried to use as much data as possible. We know that data is very important. But according to the founder of Lyraids, Dr. Serlin, data was the biggest challenge that we were facing when we were creating Lyraids. And I'm gonna editorialize this and say that inconsistent data was the biggest challenge. So let's take a look at the two of our most important building blocks, namely arterial phase hyperenhancement and washout appearance. These are all the terms available in literature that describe how HTC appears on arterial phase. And you can see that the terms are quite variable, even though they all describe the same thing. And this variability is acceptable to the point where some papers interchangeably use different terms, again, to refer to the same process. Washout has a slightly more consistent term name, but if we look at the definitions provided by different papers, the definitions vary slightly. But importantly, some include what the lesion should appear like on arterial phase, some just focus on appearance on portal venous and delayed phases. Why does it matter? Well, if we look at this particular observation, which is 26 millimeters, it enhances similar to the liver on arterial phase and then becomes hypoenhancing to the liver on portal venous and delayed phases. And then we ask ourselves, what does it look like on arterial phase? Well, if we look at all these terms, it does have arterial enhancement, right? So at least two terms, it meets the criteria for. At least three terms, it doesn't, because it doesn't seem like it is hypervascularized. And then these three terms, it's unclear. Does it have arterial wash in? Maybe, maybe not. What about washout? Again, if we look at all the different definitions, depending on the definition we choose, the lesion can be characterized as either having washout or not, or not sure. What does it matter? Well, the matter is when we assign the category, where are we gonna fall? Well, if we choose the term and definitions that say that this observation has both arterial phase hyperenhancement and washout appearance, well, that will give us Lyraids 5 definite HCC category. If we choose the term and definition that has arterial phase hyperenhancement and no washout, well, then the category is LR4, probable HCC. If we choose the term and definition that has no arterial phase hyperenhancement, but does have washout, we still have LR4, but now notice that the observation will fall into a different cell. Finally, if we choose the term and definitions that characterizes both of these as absent, well, then the category is LR3. So you can see that the category really depends on the choice of terms and definitions, which means that if this observation is included in multiple studies, we cannot be certain how this observation is characterized, and probably it's not characterized consistently. So what does that mean? When we go and we synthesize the data for meta-analysis to try to inform how Lyraids should develop and should be modified, we have a much podge of studies. They're all looking at the same thing, right? We're not looking at apples and oranges, we are looking at apples, but all apples are slightly different. So when we put them together and try to get a 3D understanding of what the ultimate truth is, instead of having a nice understanding of an apple, we have this very unusual kind of patchy collection that doesn't quite represent the true state of affairs. So we do need to do better. Because in order for us to have a consistent data-driven algorithm, really we need a good pillar of consistent and widely utilized lexicon. So in order to achieve that, in 2019, we formed a lexicon and writing group. You can see the members here. And what the group did was reviewed all the supporting materials, manuals, scores. We extracted the terms and definitions which were available in these documents. And then through really multiple, multiple iteration, discussions, rewriting, discussions with lexicon writing group and the steering committee, finally we came up with a final lexicon, which includes terms, definitions, context of use and comments. Now this one slide is a summary of a process that took about 18 months and over 1,000 hours to complete. Where do you find the result of this hard work? If you go to a lending page for LIARDS at the ACR, notice there's gonna be a link that says Access LIARDS Lexicon. If you click on it, you will get free access to all the terms and definitions. The lexicon contains name of terms, the final definition as of 2021, context of use, the applicable modality for each term, and then comments, which are also specific to LIARDS and can be applicable to general population. And then things like synonyms, type of term and approval date. Now you may be wondering what is this context of use I mentioned? Well, for each term, we have two possibilities. Broad, it means that the term is applicable to any person who comes in for liver imaging and LIARDS specific means it applies only to LIARDS population. And majority of the terms in LIARDS Lexicon do have broad context of use. So let's take a look. For example, targetoid appearance on transitional or hepatobiliary phase images. This term is applicable to both patients who are LIARDS patients and patients who fall into general population. But if we look at the threshold growth, which is defined a size increase of a mass by at least 50% within six months, what this term with this definition really makes sense only in under LIARDS conditions. And therefore this is LIARDS specific context of use. What about context of use of Lexicon itself? Well, we intend it to be used for research, for educational materials and clinical care. If we use it for research, meaning that all of us who produce scientific publications, liver imaging, we use the same terms and definitions. No matter which study this observation will fall in, if you use a Lexicon, you will arrive that this lesion does not have arterial phase hyperenhancement. What about washout? Well, it does have washout appearance based on the definition. And therefore you will consistently arrive at LIARDS for observation. We'll go as far as to say that we should use LIARDS Lexicon even if you're not really assessing LIARDS categorization. This is a recent study published in JMRI. And the goal of the study was to create a separate diagnostic system specific for patients with hepatitis B. But the study used the definitions provided by LIARDS version 2018 Lexicon. So all the imaging features they've assessed were assessed using our Lexicon. This is the table with a scoring system that they proposed. And they said if you achieve score of 12 or above, you have pretty good performance for diagnosis of HTC. So even though this study didn't really go into details and didn't assess LIARDS per se, just knowing that they use our imaging feature definitions, we can actually directly compare how their system is comparable to LIARDS 5 definitions. And their system, for example, can diagnose hypovascular HTC over two centimeters with provision of other features present. But their system cannot diagnose 10 to 19 millimeter HTC based only on imaging features of AFI and washout. So now you can see that we can use that information to really further improve and refine the algorithm. So when we go and we synthesize the data, now we're looking at consistent appearance of the same apple. And now when we put it all together, we have a nice 3D representation of what we are of the ground truth that we are seeking. What about education and clinical care? Well, it's very hard for me to provide the data that says if we use standardized language for education and clinical care, it has positive impact. But the importance of speaking the same language and how it affects our ability to cooperate and really achieve our goals, that importance is known for thousands and thousands of years. Any of you know the story of Tower of Babel, when people decided that they're going to build a tower that will reach the sky and get to the Garden of Eden. And the only way to stop this process was to mix up the languages that people stop understanding each other. And this thousands of years old quote says, if they have begun to do this as one people speaking the same language, that nothing they devise will be beyond them. And if we use a more modern translation, the people are united and they all speak the same language. After this, nothing they set out to do will be impossible for them. So our ability to collaborate and achieve our goals is vastly improved if we all speak the same language. So again, we intend to use Lyra's lexicon in research, education and clinical care. So in summary, lexicon is a collection of precisely defined term pertaining to liver imaging. It is intended to be used for scientific publication, even those that do not use Lyra's characterization, educational material and clinical care. And use of this unified language will allow comprehensive data synthesis, clear communication, improved collaboration and ultimately improve patient care and outcomes. Thank you very much for your attention. Thank you very much. And it is my pleasure to be here today. I thank the program committee and the opportunity to be on the same stage with not only my mentors and heroes on this topic, but my friends as well. So my task today is to talk about one of the newer systems within Lyrads and that's the contrast enhanced ultrasound diagnostic algorithm. So we'll briefly review contrast enhanced ultrasound for those hopefully few in the audience that are unfamiliar with this technique. Talk about how the CEUS Lyrads algorithm fits within the entire system. Go into some of the details of the algorithm itself with a few case examples, but of course there is a lot online and in the literature to help you get started if you are not started already. Here in the United States, there are several approved agents with category one CPT codes. So this is not research. This is now standard of care, clinical care at many sites in the United States. There are some interesting agents around the world, including Sonozoid not available in the United States that has some very unique pharmacokinetics. This particular agent has a Kupfer cell phase that functions a lot like the hepatobiliary agents we use in MRI. Unfortunately, we do not have time to go into that today. So ultrasound contrast agents are made up of micro bubbles, about five microns in size. At this size, they function as a pure blood pool agent. They are cleared by the lungs, not by the kidneys and they have a very short half-life which allows you to administer multiple injections in a single examination. The CUS Lyrads algorithm was published online in 2017 and has lots of great information, including the core and essentials documents and multiple translations. But it begs the question to probably many in the audience, why have another diagnostic system when we have two fantastic modalities for this purpose already? Well, we should all be humble enough to know that each of the modalities we have in our tool belt has strengths and weaknesses and not every patient and every center is going to fit in one mold and one model and be best served by only one tool. And there's always going to be disruptive technologies that come in that completely disrupt how we practice medicine and how we take care of these patients. So we should always keep our minds open and I hope I can open up your mind with the benefits of contrast enhanced ultrasound. Some of which are inherent to the modality itself, which I don't have to go into today, other than to say that contrast enhanced ultrasound provides very high spatial and temporal resolution. The specific advantages of the contrast agent are the safety profile and the very high contrast sensitivity, meaning a high signal to background and the ability to perform real time tissue subtraction, which provides a lot of advantages as we're doing these studies in our patients. So this is how the Lyrads system looks and has multiple components now. Many of you know of the CT and MRI diagnostic algorithm, which Dr. Serlin provided some information on before my talk. There's a treatment response algorithm, which I'm sure many of you are using as well. We'll hear some great information about that to follow. We don't have time to talk about their surveillance ultrasound algorithm. I'll have to talk to the program committee about leaving out another ultrasound system. But today we'll be talking about the CEUS diagnostic algorithm and look, it has the same prominence in the whole system as the CT and MRI diagnostic algorithm to where if you find an observation suspicious for HCC, CEUS Lyrads can be used for the definitive diagnosis of HCC. The table looks, the core document looks a lot like the CT and MRI algorithm. It has the decision tree and the diagnostic table and the decision tree is much like Dr. Serlin was pointing out that there's several checks you need to make. For instance, is there tumor in vein, which contrast enhanced ultrasound can be used for this purpose, showing enhancing thrombus that washes out, which can be easily differentiated from bland thrombus. The next two checks are, is this a LR1 or LR2 observation? So we provide really some definition or some examples as opposed to strict definitions of a benign observation on ultrasound. Of course, these for LR1 would be simple cysts and classic hemangiomas. We rarely use the LR2, probably benign category. Conceptually, this would be close to 100% benign, but there may be some features that may not make us completely comfortable to call an LR1 and again, there's some caveats and some descriptions here which we don't have time to go into. Before I get to the LRM category, I'd like to talk about the CUSLR major features. And there are two, which is non-RIM arterial phase hyperenhancement or AFI and washout. In the context of CUS, this is late and mild washout. So for arterial phase hyperenhancement, this is non-RIM AFI, much like in CT or MRI. And interestingly, literature is showing that a defined nodule on grayscale ultrasound with AFI has a high likelihood of representing HCC. And again, this is in high-risk patients. I didn't mention that, but Dr. Serlin provided that same graphical table of when and when not to apply LIRADs and CUS LIRADs is similarly applied. Also, CUS is not nearly as confounded by these small little perfusional variants in the periphery of the liver that often end up being LR3s at CT or MRI. Now, if you have RIM AFI, this is a feature of LRM, so not HCC-specific. And if you have the classic peripheral discontinuous nodular enhancement of a hemangioma, this is LR1. So the second major feature is gonna be washout. And this is where CUS is slightly different than CT or MRI, in which case, in that we assess both the onset as either early or late, as defined by less than 60 seconds or 60 seconds or greater, respectively. Early washout is very concerning for malignancy, but not HCC-specific, whereas a major feature of HCC is late washout. We also assess the washout degree as either mild or marked. Major feature of HCC is mild washout, which means it does become less enhancing than liver but does not become devoid of contrast which is considered marked but that assessment is defined less than two minutes. Again concerning for malignancy but not HCC specific. So let's look at an example. Here we have a 3.4 centimeter observation showing arterial phase hyper enhancement. Mild washout first detected at 90 seconds and does not become marked by two minutes. So that arterial phase hyper enhancement in an observation greater than 10 millimeters with late mild washout defines CUSLR5. So let's go back to those LRM features. They're very similar to what I've been talking about before. So rim enhancement and then again early washout. So washout detected less than 60 seconds or marked washout becoming punched out in appearance within two minutes. So here's an observation in a patient with HCC, 70-year-old male with HBV, not HCC, with HBV cirrhosis and an observation here that shows enhancement in washout but we should know not to apply the diagnostic table in this case because it's rim AFI, early and marked washout, all features of LRM which can be seen here also on this MRI and this happens to be a combined hepatocellular intrahepatic cholangiocarcinoma. CUSLIRADS also has ancillary features, some favoring malignancy in general, some favoring HCC in particular, as well as ancillary features favoring validity and unfortunately we do not have time to go over those today. I'm going to wrap up here with a case example from our clinic. This is a patient coming in. He has hepatitis C and B and there's a 3-centimeter isochoccal observation in the left lobe of liver and this is the surveillance ultrasound algorithm with categories 1, 2, and 3, so this is assigned a 3 positive examination with recommendation of a multi-phase contrast enhanced CT MRI or ultrasound. In our clinic we have the ability to do same day contrast enhanced ultrasound. We send the patient down the hall, they get an IV, they come back and at that same visit we show arterial phase hyperenhancement and mild late washout, CUSLR5, definitively HCC. Dr. Serlin pointed out the intent of the LIRADS system is to maintain a high specificity for CUSLR5 category and with CUS you can see that multiple studies have shown that we achieve the expected and needed high specificity over 95% for HCC when assigning this category. So in conclusion, CUS may be used as a safe and effective problem solving tool when CT or MRI are either contraindicated or suboptimal and in our case we often use it as a primary diagnostic modality. It has a high accuracy for differentiating benign from malignant lesions and has a high specificity for HCC and again it may be utilized for same day characterization of nodules detected at screening or surveillance ultrasound. I wish to thank all of the members of the CUS LIRADS working group, keeping in mind that we are just a few members of the large LIRADS team and all working toward the same goal of improving patient care and with that I thank you. The last talk will be about the LIRAD TRA, treatment response assessment algorithm. During the next 12 minutes I will briefly introduce local regional therapies for HCC and explain LIRAD treatment response algorithm with representative cases. Then I will discuss diagnostic performance of LRTRA and will briefly explain possible role of ancillary features to improve diagnostic performance of LRTRA. Finally I will address challenges, current challenges of LRTRA after TARE and external beam radiation therapy. There are many kinds of HCC treatment options, surgery, local regional therapies including chemo ablation, energy based ablation, transcatheter therapy, radiation therapy and systemic therapy. In terms of population, LRTRA can be applied in patients to assess response for path proven or presumed malignancies like LL4, LL5 or LLM after local regional therapies. Until now, LRTRA is not applicable for systemic therapy. In terms of imaging modality, LRTRA applies for multi-phasic CT and MRI, CEUS is developing now. Several image based treatment response systems have been developed to improve accuracy in response assessment. RESIST, WHO criteria are size based classification, modified RESIST, ESIL and LIRAD TRA are enhancements based classifications. The advantage of M-RESIST and ESIL are HCC specific by focusing on the enhancing component. Overall response by M-RESIST and ESIL criteria have been better associated with the survival than RESIST and are therefore preferable to RESIST. However, these M-RESIST and ESIL criteria assess overall patient response rather than to assess individual tumors. Therefore LRTRA was developed to provide the standardized terminology as well as comprehensive but simple system to assess individual tumor treatment response which is more suitable for routine clinical practice. This is the outline of LRTRA. If the images are not adequate to evaluate, the observation is categorized into LRTRA non-evaluable. Otherwise, treatment response, treated observations would be classified into non-viable, equivocal or viable. When conventional TAC is used on follow-up CT scan, you sometimes to encounter this kind of compact uptake of the lipidol which is a good prognostic sign. However, it interferes interpretation of AFI and washout. Therefore this can be categorizable as LRTRA non-evaluable. In this situation, MRI can be better using modality to evaluate treatment response. SMR doesn't show any intralesional enhancement, we can categorize this as LRTRA non-viable. And later on surgery, the tumor shows 100% complete necrosis. LRTRA adapted the concept from MResist for assessment of viability and tumor size measurement following treatment. However, LRTRA allows measurement of tumors lacking AFI but adds features of washout and enhancement similar to pre-treatment. There are three categories, let's look into each category and corresponding imaging features. For LRTRA non-viable, there are two imaging features, no regional enhancement, treatment-specific expected enhancement, which means expected temporal and special pattern of a post-treatment enhancement attributable to treatment-related changes in parenchymal perfusion. Enhancement atypical for treatment-specific expected enhancement pattern should be regarded as LRTL equivocal. As an example, in this patient's having LR5 lesion in right posterior segment showing AFI and portal washout and hepatobiliary hypo-enhancement, tumor size was 1.7 centimeters. After ablation, CT shows non-enhancing tumor surrounded by uniform, thin, smooth peripheral enhancement along ablative margin. This rim represent treatment-specific expected enhancement pattern, which is quite frequently observed in early first months after treatment. Therefore, this should be categorized as LRTL non-viable. In addition, on nine months follow-up CT and 15 months follow-up MRI, the lesion does not show any internal enhancement within the lesion, therefore categorizable as TL non-viable. For LRTL viable, there are three imaging features that suggest the tumor viability. Nodular, mass-like, or thick and irregular-shaped post-treatment AFI, washout, post-treatment enhancement similar to pretreatment. Any enhancement not meeting criteria for these three imaging features should be regarded as LRTL equivocal. Let me show a typical example. Here is a large HCC measuring 6.9 centimeters in the right lobe of the liver, which was treated with Deptase. On one month follow-up CT, you can see a nodular-shaped arterial face hyper-enhancing lesion and also washout on delayed face. Therefore, this can be regarded as LRTL viable. These categories suggest the presence of a residual viable tumor, which was later confirmed on surgical pathology. When MRI is used for evaluation of a treatment response after local regional therapy, sometimes the treated lesion shows a high signal intensity on pre-contrast T1-weighted image. With this high signal intensity, arterial face image fails to show residual tumor enhancement. Therefore, based on this ordinary MRI image, this was regarded as LRTL non-viable, but on subtraction image, you can see intralesional enhancement. According to these papers, using subtraction images may improve sensitivity while maintaining specificity. Furthermore, it may increase reader confidence level. But this caveat is misregistration artifacts. The transcatheter treatment, which is increasingly used for treating HCC, is TARE, T-A-R-E, radioembolization. As the embolic effect of TARE is less prominent compared to other transcatheter therapies, tumor enhancement can persist after treatment, even in nodular pattern on early follow-up imaging, which may not indicate residual viable tumor. On one month follow-up, the lesion still shows intralesional enhancement, although the tumor sizes are decreasing. On six months follow-up, still you can see intralesional enhancement or peripheral enhancement. The tumor sizes are much less compared to the original tumor size. On 15 months follow-up, now you can see a new nodular development from the adjacent liver parenchyma. So these two images could be categorizable as TL equivocal, and then this one should be regarded as TL viable. Nowadays, our interventional colleagues use selective high-dose TARE. After selective TARE with the higher radiation dose, the treatment-specific enhancement pattern is somewhat different from standard TARE. We can frequently see more complete necrosis of the treated tumor, which is surrounded by thin enhancing rim. So it's easier to categorize TR non-viable. And furthermore, on pre-contrast CT scan, the treated tumor frequently show microscopic classifications as well. So you need to know high-dose TARE may show different treatment-specific expected enhancement pattern from standard TARE. After external beam RT, we can see similar phenomenon to standard TARE. In these patients who had recurrence from TAC, we used SBRT for treating this recurrent tumor because of its proximity to adjacent portal vein and bile duct. On three months follow-up after SBRT, you can see interval growth of arterially enhancing portion at the radiation field, so-called the pseudoprogression. But on six months follow-up, the degree of enhancement is much less. And also, you cannot see nodular enhancing pattern within this radiation field. So now you can categorize this as LR-TR non-viable. But at three months follow-up, even if the lesion size is bigger, you may categorize it as TR equivocal, considering the radiation-induced hepatitis. In contrast to other per-patient criteria, such as MRIST or ESL, LR-TR algorithm is per-lesion criteria. Therefore, one lesion may be TR viable. The other lesion is TR non-viable, mixed. So however, per-patient, per-lobe, per-segment assessment may be performed later. Then how about diagnostic performance of LR-TRA? One thing you should know is that LR-TRA is an imaging definition. It may not be concordant with pathology viability. The reported diagnostic performance of LR-TR with MRI is like that. According to these papers, the sensitivity is around 50% to 60%. Specificity is in range of 90% to 94%. More recently, there have been an attempt to reinforce the LR-TR by using ancillary features of LIRAT. In fact, LR-TR viable and LR-TR non-viable categories offer clear feedback to physicians. However, LR-TR equivocal category generates some ambiguity. Furthermore, most LR-TR equivocal lesions, more than 80%, contain viable tumors at pathology. Therefore, according to these papers, adding ancillary features like T2 high signal intensity, restricted diffusion, or HBP hypo intensity may improve sensitivity and reduce LR-TR equivocal category while keeping specificity. Here according to this paper, a 10% increase of sensitivity and there was a slight decrease of specificity but statistically insignificant. Here is an example. Here is a 2.5 LR5 lesion in segment 4. After TAC and RFA, on six months follow, you can see faint arterial enhancement adjacent to the treatment lesion but no washout is shown on portal venous phase. According to the major feature, it should be regarded as LR-TR equivocal but if you check the transitional phase, hepatobiliary phase, T2, and high B-value diffusion-weighted imaging, we can increase the diagnostic confidence for presence of a viable tumor. So if we use ancillary imaging features, maybe we can decrease the category of LR-TR. Finally, I will address challenges of LR-TRA after TAER. Here is a poorly differentiated HCC on combium CT interval growth was shown and intralesional hypervascularity was confirmed. On one month follow-up and four month follow-up, there is an interval size decrease although there are some intralesional enhancement. So we had to categorize these as TR-equivocal but clinician and patients wanted to have surgical resection. On pathology, there was a viable tumor. However, as the lesion has been quite stable for four months after TAER with a favorable size decrease, we still don't know when additional treatment is essentially required yet. Here is another case of HCC which was treated with TAER. You can see a huge mass in the right loop and then until 12 months follow-up, you can see interval size decrease although the tumor shows a strong arterial hyperenhancement. So we can say this could be TR-equivocal. On 24 months follow-up, now tumor seems to get bigger. So now we can categorize it as TR-viable. As there is a huge discrepancy between enhancement and size, LRTRA, next version of LRTRA will use different term from TR-equivocal. Maybe they will propose TR-evolving category. The real clinical dilemma would be viable but still not progressed after TAER. So when should we treat the tumor? We don't know the exact ideal timing to do re-treatment for this kind of tumor. Further study will be necessary. In summary, accurate interpretation of post-treatment imaging is essential for guiding further management decisions. LIRED-TRA conveys high degrees of inter-observable improvement and showed high predictive value of viable HCC after local-regional therapy. Radiation-based therapies show different post-treatment imaging findings and requires continuous application of LRTRA, cautious application of LRTRA. Adoption of ancillary imaging features might be helpful for improving diagnostic performance of LRTRA. Thank you very much for your kind attention.

LI-RADS

liver imaging

standardized lexicon

CEUS

HCC diagnosis

treatment response

diagnostic accuracy

imaging algorithms