Study design

This is a prospective observational IRB approved study, in which written consent was obtained from all enrolled patients. We collected data at our institution from January 2017 to June 2020, on patients who were treated for primary or recurrent HCC with potentially curative treatment including surgical resection or percutaneous ablation (percutaneous ethanol injection, radiofrequency ablation, and microwave ablation), and subsequently underwent alternating secondary surveillance for a minimum of 7 months. This interval of follow-up was chosen to allow ample time for tumor recurrence to occur, and to ensure relative equality in the number of MRI and CEUS exams performed. There were no other specific inclusion criteria. Patients were excluded if they underwent TACE at the time of recruitment, or if they were to have secondary surveillance exclusively with a single modality (i.e., MRI/CT only or CEUS only). Lesions were diagnosed as HCC by meeting criteria for LI-RADS 5 (in the appropriate patient population) or percutaneous biopsy. Patients were not excluded based on prior HCC treatment, nor if a LI-RADS 5 lesion was found to be a non-HCC lesion (i.e., ICC [intrahepatic cholangiocarcinoma] or mixed ICC/HCC) on the pathologic specimen acquired at the time of treatment.

Patients were recruited following multidisciplinary hepatobiliary rounds, where all patients with a new diagnosis of HCC are presented for determination of optimal management. Secondary surveillance was performed as per our local schema (Fig. 1), which entails an MRI performed 1 month after treatment, followed by CEUS and MRI performed at alternating 3-month intervals (i.e., CEUS at month 4, MRI at month 7, etc.), for a total of 2 years. Patients who were recurrence free for two years after treatment were discharged to the community, where they were evaluated with grayscale ultrasound only, at an increased follow-up interval of 6 months. Patients with new or recurrent tumors within two years of treatment, were typically re-treated and included in the data as a new lesion. In accordance with our local practice, an equivocal finding on one imaging modality during secondary surveillance triggered expeditious evaluation with the alternating imaging modality (Fig. 1). All cases interpreted as “equivocal” or “positive for tumor recurrence” were discussed at multidisciplinary hepatobiliary rounds to determine management. In order to evaluate our real-world clinical practice, image interpretation was based on the report of the dictating radiologist (all of whom were abdominal imaging specialists with experience ranging from 2 to 15+ years) unless an alternate decision was rendered at multidisciplinary hepatobiliary rounds. Access to clinical, biochemical, and prior imaging information was available and its use was at the discretion of the multidisciplinary team. The performance of our study did not influence image interpretation or decisions regarding patient management.

Imaging techniques

The CEUS technique for this study was highly standardized, performed at a sole institution, and interpreted by a group of 6 abdominal radiologists. Of the 6 radiologists that interpreted CEUS, 2 solely interpreted CEUS and did the majority of scans, while the remaining 4 interpreted both CEUS and MRI. The MRI technique was also standardized, but by comparison was performed at four University affiliated facilities, and the group of interpreting radiologists was larger (approximately 15–20 radiologists, all of whom were Abdominal Imaging specialists).

CEUS technique

The majority of exams were performed with either an RS80 EVO ultrasound machine (Samsung, Seoul, South Korea) or an ACUSON Sequoia (Siemens, Munich, Germany). A radiologist was typically not present in the room with the technologist for performance of the examination, but images were checked with the radiologist for all cases prior to discharging the patient from the department. However, the radiologist’s involvement increased for more complex cases. For performance of contrast-enhanced ultrasound, the machine must be equipped with the following:

Contrast-specific software to enable:

Production of low mechanical index sound waves (to minimize destruction of the microbubble contrast agent)

Real-time subtraction technique (to create a microbubble only image)

Dual-screen mode (with dual screen calipers) for anatomic correlation between the grayscale image and the subtracted microbubble only image

Clearly visible on-screen timer

We typically use perflutren microspheres (Definity; Lantheus Medical Imaging, Billerica, MA), a purely intravascular microbubble contrast agent, which functionally is indistinguishable from sulfur hexafluoride (Lumason; Bracco Diagnostics Inc., Monroe Township, New Jersey).

Prior to initiating the exam, the patient’s treatment history must be obtained. Vital information includes treatment type, treatment date, treatment site location, pretreatment imaging features of lesion, and any prior images of the treatment site.

Next, the patient’s liver is thoroughly evaluated with grayscale ultrasound. This very important step serves multiple purposes. Firstly, all known treatment sites are identified, measured, and correlated with known treatment history. It is essential to assess the treatment site for any juxtaposed areas of nodularity on grayscale. Secondly, any new or growing nodules elsewhere in the liver are identified. Lastly, the grayscale ultrasound is used to plan the contrast-enhanced ultrasound. Planning consists of identifying the best possible acoustic window, best patient position, and practicing the required breathing technique with the patient to facilitate optimal evaluation of the treatment site. We prefer assessment in the long-axis to reduce in-plane and out-of-plane respiratory motion, but this is not always possible. During performance of the contrast-enhanced scan the transducer is maintained over the central portion of the treatment site during the arterial phase, with careful cranial and caudal sweeps to assess the entirety of the treatment site.

The set-up for CEUS involves obtaining intravenous access with a 20–24 gauge needle, onto which a three-way stopcock is attached. A 1.2 mL vial of perflutren is activated in a vial mixer (Vialmix; Lantheus Medical Imaging, Billerica, MA), which increases the content volume to 1.8 mL. Perflutren is then drawn into a 1.0 mL syringe and attached to the parallel/straight port of the three-way stopcock. A 10 mL saline flush is attached to the perpendicular/side port. Next, 0.2 mL of perflutren is injected by hand, immediately followed by a 10 mL saline flush (which ensures optimal delivery of the contrast bolus). Occasionally 0.3 mL of perflutren is used for severely cirrhotic livers that enhance poorly.

Timer initiation is simultaneous to initiation of the saline flush. The treatment site is monitored continually in anticipation of the first microbubble, at which time the cine loop recording is initiated. Ongoing continuous monitoring is performed from initiation of the timer (not of the cine loop recording) to just beyond the peak of arterial phase enhancement. Subsequently images are taken at 30–60 s intervals, and including at 1 min, to assess for washout up to 5–6 min, with particular attention paid to any suspicious foci of APHE to identify corresponding washout.

Repeat contrast injections are performed, as needed (up to a dose of 20 uL/kg), to re-evaluate any suspicious findings or if the initial assessment is unsatisfactory. Note that a maximal dose of 20uL/kg would permit up to 1.4 mL of total contrast agent (7 injections of 0.2 mL) in a 70 kg patient. Additionally, a technique we refer to as an “on top injection” is particularly useful if a discrete region of washout is observed. In such a circumstance, an additional 0.2 mL of perflutren is injected (followed by 10 mL saline flush) during the portal venous or late phase on top of the identified washout. This valuable technique facilitates careful observation for corresponding APHE in a known region of washout, which may not have been conspicuous on the initial injection.

MRI technique

MRI technique varied slightly between institutions and scanners, but all protocols met the requirements detailed in the CT/MRI LI-RADS v2018 manual. All scans were performed on either a 1.5T or 3T magnet. Universally acquired sequences included unenhanced T1-weighted in-phase and out-of-phase images, unenhanced T2-weighted images with and without fat suppression, and fat suppressed multiphase 3D T1-weighted images (pre-contrast, late arterial phase, portal venous phase, and delayed phase) using the macrocylic extracellular contrast agent gadobutrol (Gadavist; Bayer Healthcare Pharmaceuticals, Leverkusen, Germany), and post-processed subtraction images of the fat suppressed multiphase T1-weighted images.

Image interpretation: MRI and CEUS

The algorithm for interpretation of imaging findings is outlined in Fig. 2 and detailed in the following text. These definitions were designated based on our clinical experience, as the onset of our study (2017) predated inclusion of a “Treatment Response” section in the CT/MRI LI-RADS Manual, which was first included in the 2018 version [10].

Fig. 2

Flowchart summarizing interpretation of imaging findings. An avascular treatment site, or the presence of treatment specific enhancement, was defined as negative for tumor recurrence on both modalities (equivalent to LR-TR Non-Viable). The presence of a nodular region of arterial phase hyperenhancement (APHE) that demonstrates washout on portal venous or delayed phase images was defined as positive for tumor recurrence on both modalities (equivalent to LR-TR Viable). An isolated finding of APHE only or washout only, on either modality, was defined as equivocal for tumor recurrence (equivalent to LR-TR Equivocal). Equivocal findings were further evaluated with the alternative modality. An avascular treatment site or treatment specific enhancement on the alternate modality was then defined as negative for tumor recurrence. APHE with washout, APHE only, or washout only, seen on the alternate modality (and corresponding to the equivocal finding on the first modality) was then defined as positive for tumor recurrence

Positive for recurrence

MRI scans were categorized as positive tumor recurrence in the presence of APHE with washout appearance (in portal venous or late phase), or enhancement similar to the initial lesion prior to treatment. We defined recurrence on CEUS in the same way (i.e. APHE with washout, or enhancement similar to pretreatment). This is a deviation from the definition in the “Treatment Response” section of the LI-RADS CT/MRI manual (v2018), which suggests APHE, or washout appearance, or enhancement similar to pre-treatment is sufficient to diagnose tumor recurrence (i.e. “LR-TR Viable”) [10].

Equivocal for recurrence

Cases in which APHE or washout occurred as a solitary finding on either modality, or enhancement was atypical for post-treatment appearance, were categorized as equivocal. Following expeditious assessment with the alternate modality the equivocal result was either confirmed or disproven (Fig. 2). The following findings on the alternate modality were considered confirmatory for tumor recurrence: APHE with washout, or APHE only, or washout only. An avascular treatment site or treatment-specific enhancement were considered negative for tumor recurrence. The LI-RADS manual defines an equivocal result (LR-TR Equivocal) as “enhancement atypical for treatment-specific expected enhancement pattern and not meeting criteria for probably or definitely viable” [10].

Negative for recurrence

A scan was categorized as negative for tumor recurrence in the setting of an avascular treatment site or treatment-specific enhancement pattern. This is identical to the definition used by LI-RADS (LR-TR “Non-Viable”) [10].

Geographic categorization of recurrent tumor

Recurrent tumors were categorized based on their location relative to the treatment site as intrazonal or extrazonal (Fig. 3) [9]. Newly identified nodules were categorized as segmental (within the same segment as a treatment site), or as remote (elsewhere in the liver). Geographically describing tumor recurrence has over time become integral to interactions with our Interventional Radiology colleagues, as it contributes to decision-making at multidisciplinary rounds and treatment planning.

Fig. 3

Schematic representation shows geographic HCC recurrence patterns (orange). Reproduced from reference 9 with permission from The Radiologic Society of North America (RSNA). Intrazonal tumor occurs within a treatment site. Extrazonal tumor describes recurrent tumor that is juxtaposed to, but outside of, the treatment site. Segmental tumor occurs within the same segment as a previously treated lesion, but with normal intervening liver tissue, and may represent either a satellite nodule or de novo HCC. Remote recurrence is a new nodule, in a segment separate from any treatment site. Intrazonal and extrazonal tumors are considered true recurrence (i.e. LR-TR Viable), while segmental and remote tumors are considered new tumors (i.e. LR-4 or LR-5 lesion)

As compared to standard liver ultrasound, where nodules are measured on grayscale images, post-treatment scans are better viewed in conjunction with the contrast-enhanced image. This is because the exact margins of both the treatment site and recurrent tumor are often not clearly delineated on the grayscale images alone. The size of recurrent tumors was measured based on the longest dimension, in accordance with the LI-RADS CT/MRI Manual [10].

Statistical analysis

Patients with incomplete treatment were not included in the analysis, until they were re-treated and complete treatment was achieved. Additionally, the first follow up was not included in the test comparisons, as it was constrained to MRI only. Subsequent alternating total number of MRI and CEUS examinations were analyzed using classical test diagnostics.

Point and 95% confidence interval (CI) estimates were calculated for sensitivity, specificity, and the receiver operator curve (ROC). A scan was denoted as true positive (TP) if it met requirements for “positive for recurrence” (defined in previous section), or if it was “equivocal for recurrence” and subsequently confirmed as “positive for recurrence” on the expedited scan. If a scan was “equivocal for recurrence” and disproven (i.e. “negative for recurrence”) on the expedited scan, it was denoted as false positive (FP). When a scan was “positive for recurrence”, the preceding scan was re-examined. If, in retrospect, tumor recurrence was clearly visible on the preceding scan the scan was denoted as false negative (FN). If tumor could not be clearly identified on the preceding scan, it was denoted as true negative (TN). True negative (TN) denotation was also applied to scans that were “negative for recurrence”, and remained “negative for recurrence” on at least 2 follow-up scans. All statistical analyses were undertaken using Stata ® version 16.1 (StataCorp, TX, USA) with an alpha of 0.05.