Blog

LONDON — LONDON (AP) — Amazon says a massive outage of its cloud computing service has been resolved as of Monday evening, after a problem disrupted internet use around the world, taking down a broad range of online services, including social media, gaming, food delivery, streaming and financial platforms.

The all-day disruption and the ensuing exasperation it caused served as the latest reminder that 21st century society is increasingly dependent on just a handful of companies for much of its internet technology, which seems to work reliably until it suddenly breaks down.

About three hours after the outage began early Monday morning, Amazon Web Services said it was starting to recover, but it wasn’t until 6 p.m. Eastern that “services returned to normal operations,” Amazon said on its AWS health website, where it tracks outages.

AWS provides behind-the-scenes cloud computing infrastructure to some of the world’s biggest organizations. Its customers include government departments, universities and businesses, including The Associated Press.

Cybersecurity expert Mike Chapple said “a slow and bumpy recovery process” is “entirely normal.”

As engineers roll out fixes across the cloud computing infrastructure, the process could trigger smaller disruptions, he said.

“It’s similar to what happens after a large-scale power outage: While a city’s power is coming back online, neighborhoods may see intermittent glitches as crews finish the repairs,” said Chapple, an information technology professor at the University of Notre Dame’s Mendoza College of Business.

Amazon pinned the outage on issues related to its domain name system that converts web addresses into IP addresses, which are numeric designations that identify locations on the internet. Those addresses allow websites and apps to load on internet-connected devices.

DownDetector, a website that tracks online outages, said in a Facebook post that it received over 11 million user reports of problems at more than 2,500 companies. Users reported trouble with the social media site Snapchat, the Roblox and Fortnite video games, the online broker Robinhood and the McDonald’s app, as well as Netflix, Disney+ and many other services.

The cryptocurrency exchange Coinbase and the Signal chat app both said on X that they were experiencing trouble related to the outage.

Amazon’s own services were also affected. Users of the company’s Ring doorbell cameras and Alexa-powered smart speakers reported that they were not working, while others said they were unable to access the Amazon website or download books to their Kindle.

Many college and K-12 students were unable to submit or access their homework or course materials Monday because the AWS outage knocked out Canvas, a widely used educational platform.

“I currently can’t grade any online assignments, and my students can’t access their online materials” because of the outage’s effect on learning-management systems, said Damien P. Williams, a professor of philosophy and data science at the University of North Carolina at Charlotte.

The exact number of schools impacted was not immediately known, but Canvas says on its website it is used by 50% of college and university students in North America, including all Ivy League schools in the U.S.

At the University of California, Riverside, students couldn’t submit assignments, take quizzes or access course materials, and online instruction was limited, the campus said.

Ohio State University informed its 70,000 students at all six campuses by email Monday morning that online course materials might be inaccessible due to the outage and that “students should connect with their instructors for any alternative plans.” As of 7:10 p.m. Eastern, access was restored, the university told students.

This is not the first time issues with Amazon cloud services have caused widespread disruptions.

Many popular internet services were affected by a brief outage in 2023. AWS’s longest outage in recent history occurred in late 2021, when a wide range of companies — from airlines and auto dealerships to payment apps and video streaming services — were affected for more than five hours. Outages also happened in 2020 and 2017.

The first signs of trouble emerged at around 3:11 a.m. Eastern time, when AWS reported on its “health dashboard” that it was “investigating increased error rates and latencies for multiple AWS services in the US-EAST-1 Region.” Later, the company reported that there were “significant error rates” and that engineers were “actively working” on the problem.

Around 6 a.m. Eastern time, the company reported seeing recovery across most of the affected services and said it was seeking a “full resolution.” As of midday, AWS was still working to resolve the trouble.

Sixty-four internal AWS services were affected, the company said.

Because much of the world now relies on three or four companies to provide the underlying infrastructure of the internet, “when there’s an issue like this, it can be really impactful” across many online services, said Patrick Burgess, a cybersecurity expert at U.K.-based BCS, The Chartered Institute for IT.

“The world now runs on the cloud,” Burgess said.

And because so much of the online world’s plumbing is underpinned by so few companies, when something goes wrong, “it’s very difficult for users to pinpoint what is happening because we don’t see Amazon, we just see Snapchat or Roblox,” Burgess said.

“The good news is that this kind of issue is usually relatively fast” to resolve, and there’s no indication that it was caused by a cyberattack, Burgess said.

“This looks like a good old-fashioned technology issue. Something’s gone wrong, and it will be fixed by Amazon,” he said.

There are “well-established processes” to deal with outages at AWS, as well as rivals Google and Microsoft, Burgess said, adding that such outages are usually over in “hours rather than days.”

___

Ortutay reported from San Francisco. Associated Press videojournalist Mustakim Hasnath in London and Jocelyn Gecker in San Francisco contributed to this report.

Introduction

Pancreatic cancer is one of the most aggressive malignancies, with only 20% of patients eligible for surgical resection at the time of diagnosis.^1–3 These patients often face prolonged hospitalization and significant postoperative challenges, among which pain control remains a major clinical concern. Poorly managed postoperative pain can stimulate catecholamine release, which may suppress natural killer cell activity—a component of innate immunity—and potentially influence anti-tumor responses.⁴ Additionally, it is associated with increased psychological distress and reduced quality of life. Despite its clinical significance, current pain management strategies after pancreatic surgery are often suboptimal, underscoring the need for more effective analgesic approaches and further investigation into their impact on postoperative outcomes.

Patient-controlled intravenous analgesia (PCIA) with opioids is widely used for postoperative pain control.^5–8 Sufentanil, a selective potent μ-receptor agonist, is widely used for its efficacy in postoperative pain management.⁹ Given the moderate to severe pain typically associated with pancreatic surgery, a potent analgesic strategy is essential. However, increasing the dosage of a single analgesic agent to achieve adequate pain relief may also elevate the risk of adverse effects, including respiratory depression, nausea, and vomiting. Dezocine, a partial μ-receptor agonist and κ-receptor antagonist, has emerged as a promising adjunct due to its analgesic and sedative effects, as well as its favorable safety profile compared to pure μ-receptor agonists.^10–12 By acting on κ-receptors in the spinal cord and brain, dezocine provides analgesic and sedative effects without the typical µ-receptor dependence, potentially reducing adverse reactions such as smooth muscle relaxation.¹⁰ Previous studies have demonstrated that dezocine offers significant postoperative antihyperalgesic and analgesic effects, with benefits lasting up to 48 hours in patients undergoing open gastrectomy.¹³

Several studies have demonstrated that dezocine, when combined with morphine, enhances postoperative analgesia and reduces opioid-related side effects, such as nausea and pruritus, making it a valuable option in anesthesia practice.^14–16 At our institution, the combination of sufentanil and dezocine has been used in PCIA for pancreatic cancer patients for several years. However, the efficacy and safety of this combination have not been thoroughly investigated. To address this gap, we conducted a propensity score-matched (PSM) study at a high-volume pancreatic center to evaluate the role of dezocine as an adjunct to sufentanil in PCIA for postoperative pain management following pancreatic surgery, which, to our knowledge, is the first study to investigate the analgesic effects of this combination in PCIA for pancreatic surgery patients.

Materials and Methods

Ethics Approval

This retrospective study was approved by the Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine (Ethics Approval Number: (2023) No. 48), with a waiver of patient written informed consent due to the use of de-identified, archival medical records (no active patient intervention). All patient identifiers were removed, and data were stored securely on encrypted servers accessible only to the research team, adhering to the Declaration of Helsinki (as revised in 2013).

Patients

A total of 1485 patients who underwent elective open or minimally invasive pancreatic tumor surgery and received patient-controlled intravenous analgesia (PCIA) for postoperative pain management at the Pancreas Center of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, between January 2022 and January 2023 were retrospectively enrolled. The center is one of the largest pancreatic surgery centers in Asia. Among them, 794 were male and 691 were female, with an age range of 18 to 85 years (mean age: 60.55 ± 12.55 years) and American Society of Anesthesiologists (ASA) physical status classification ranging from I to IV. Based on the PCIA regimen, patients were allocated into two groups: the sufentanil group (n = 251) and the sufentanil-dezocine combination group (n = 1234). Surgical approach (Laparotomy/Laparoscopic/Robotic) was documented based on the description of the surgical procedure in the operative notes. All operative notes were reviewed and signed off by the attending surgeon or a senior resident physician to ensure consistency in classification. To minimize confounding and selection bias, PSM was performed using a logistic regression model based on age, sex, BMI, surgical approach (laparotomy, laparoscopic, robotic), surgery type (pancreatoduodenectomy, total pancreatectomy, middle-preserving pancreatectomy, distal pancreatectomy, as different techniques may affect pain severity due to varying tissue trauma), and dexmedetomidine dose. A caliper of 0.02 and nearest-neighbor matching were applied in a 1:3 ratio using R software (v.4.3.1, The R Foundation for Statistical Computing, Vienna, Austria. http://www.r-project.org). Exclusion criteria included: (1) known allergies to study drugs; (2) inability to use patient-controlled intravenous analgesia (PCIA); (3) history of chronic pain or long-term use of analgesic medications; (4) requirement for reoperation due to postoperative bleeding or severe abdominal infection; (5) severe cardiopulmonary or hepatorenal insufficiency and (6) cognitive dysfunction.

Anesthesia Procedure

All patients fasted for 8 hours (solids) and 6 hours (clear liquids) preoperatively and were transferred to the operating room without premedication. Standard monitoring included electrocardiography (ECG), non-invasive blood pressure (BP), respiratory rate (RR), oxygen saturation (SpO₂), end-tidal carbon dioxide pressure (PetCO₂), and bispectral index (BIS). A uniform anesthetic regimen was administered to all patients, with surgeries performed by the same surgical team.

General anesthesia was induced with propofol (2–2.5mg/kg), sufentanil (0.3–0.5 µg/kg), rocuronium (0.6–0.8mg/kg) or cisatracurium (0.2–0.3mg/kg), dexamethasone 5 mg, and dexmedetomidine 0.6 µg/kg. Preoxygenation with 100% oxygen was administered for at least 3 minutes via a face mask. Anesthesia was maintained with sevoflurane (3vol%, 0.8–1.3MAC), remifentanil (0.2–0.4µg/kg/min), supplemental rocuronium (1/3–1/5 of the induction dose), and intermittent sufentanil (0.4 µg/kg). Ventilation was set at a tidal volume of 8 mL/kg, with respiratory frequency adjusted to maintain PetCO₂ at 35–45 mmHg. Anesthesia depth was titrated to maintain a BIS between 40 and 60, ensuring mean arterial pressure (MAP) and heart rate (HR) remained within 20% of baseline values. Patient temperature was maintained above 36°C using infusion heaters and warming blankets. A sufentanil loading dose (0.1 µg/kg) was administered 30 minutes before the end of surgery. Intraoperative fluid balance was defined as the net change in a patient’s total body fluid volume during surgery, calculated as the difference between the total intraoperative fluids inputs and outputs. Postoperatively, patients were transferred to the post-anesthesia care unit (PACU), where residual neuromuscular blockade was reversed with neostigmine (40 µg/kg) and atropine (20 µg/kg).

Postoperative PCIA Regimen

After meeting extubation criteria, patients were extubated and connected to an Artificial Intelligence Patient-Controlled Analgesia (AI-PCA) system (Model ZZB-IB, Nantong AIPU Medical Inc., China). Patients were divided into two groups based on the PCIA solution: the sufentanil group received sufentanil (1.0 µg/mL), and the combination group received sufentanil (1.0 µg/mL) plus dezocine (2.5 mg/mL). Group allocation was guided by clinical judgment of the anesthesiologist considering factors reflected in our dataset such as patient demographics, surgical complexity, intraoperative management details.

The Acute Pain Service team prepared the PCIA solution in 100 mL normal saline bags, containing either sufentanil alone or the combination and monitored patients. If the Numerical Rating Scale (NRS) at rest was ≥4, a 2 mL bolus of PCIA solution was administered at 15-minute intervals until NRS <4. Patients were then encouraged to self-administer PCIA as needed.

The PCIA pump was set to a background infusion rate of 2 mL/h, with a 2 mL bolus dose and a 15-minute lockout interval. PCIA was maintained for 48 hours postoperatively, during which vital signs including respiratory rate, oxygen saturation, and sedation scores were closely monitored.

Outcome Measures

Demographic and intraoperative data, including surgery type, site, anesthetic drug dosages, blood loss, transfusion, and fluid balance, were recorded. Postoperative data included PCIA pump usage duration, total input, cumulative and effective press counts, rescue analgesia, and adverse events (eg, vomiting, pruritus, respiratory depression, hypotension, dizziness, delirium). We assessed Functional Activity Score (FAS) and Level of Sedation (LOS) at 1 and 2 days post-surgery. FAS (1–3 grades) quantifies pain impact on daily functions: 1=no limitation (normal coughing/limb movement despite pain); 2=mild limitation (slight difficulty/slower actions), and 3=severe limitation (struggles with basic activities). LOS (0–3 grades) evaluates consciousness via responsiveness: 0=alert (follows instructions), 1=somnolent (wakes to calls but drifts off), 2=stuporous (brief pain wakefulness), 3=comatose (no response to calls or pain). Both were graded during routine checks to guide pain management and monitor recovery. Pain intensity was evaluated using the NRS at rest (NRSR) and during coughing (NRSC) at 24, 48, and 72 hours post-surgery. The NRS ranges from 0 (no pain) to 10 (worst imaginable pain). Moderate-to-severe pain was defined as NRS ≥4. Mild pain (NRS 1–3) was also recorded in postoperative data. Adverse events were recorded based on routine clinical documentation in the hospital’s electronic medical records (EMR) and nursing care logs.

Primary endpoints were the incidence of moderate-to-severe pain at rest and during coughing within 48 hours post-surgery. Secondary endpoints included the incidence of moderate-severe pain at rest and during coughing at 24 hours and 72 hours post-surgery, LOS, FAS, and adverse events.

Statistical Analysis

Continuous variables were first assessed for normality, those with normal distribution were expressed as mean ± standard deviation (SD) and compared using independent t-tests. Skewed distributed continuous variables were presented as median (Q1, Q3) and analyzed with the Mann–Whitney U-test. Categorical variables were expressed as frequencies and percentage, and compared using Pearson’s chi-square or Fisher’s exact test. Missing data for demographic characteristics, intraoperative and postoperative data were imputed using the expectation-maximization algorithm. Univariate and multivariate logistic regression models, alongside with post-PSM analysis and inverse probability weighting (IPW) analysis were conducted to calculate odds ratios (OR) and 95% confidence intervals (CI). Analyses were performed using SAS (v.9.2, SAS Institute Inc., USA). All tests were two-sided, and statistical significance was set at the 5% level. No adjustments have been made for multiple testing.

Results

Patient Characteristics

Before PSM, the sufentanil group comprised 251 patients, while the combination group included 1234 patients. The sufentanil group was older (mean age 63.73 ± 13.69 years vs 59.90 ± 13.44 years, P< 0.05), had a higher proportion of pancreatoduodenectomy (PD) procedures (55.8% vs 44.8%, P< 0.05), and a greater rate of laparotomy (80.5% vs 73.9%, P < 0.05). Additionally, the sufentanil group had a lower BMI (22.25 ± 3.30 vs 22.73 ± 3.31, P < 0.05) and received a lower dexmedetomidine dosage (16.85 ± 15.75 µg vs 22.50 ± 16.32 µg, P < 0.05) compared to the combination group. No significant difference was observed in sex distribution. After PSM, the study included 247 patients in the sufentanil group and 704 in the combination group, with all baseline variables balanced between the two groups (Table 1).

Table 1 Demographic Characteristics and Perioperative Outcomes of Patients Between the Sufentanial Group and the Combination Group

Perioperative Outcomes

After PSM, no significant differences were observed in blood loss, blood transfusion volume, or total PCIA input between the two groups, despite differences before matching. The dosages of sufentanil and rocuronium bromide, as well as effective and cumulative PCIA press counts, showed no significant differences before or after PSM. However, the sufentanil group exhibited greater fluid balance difference and longer pump usage duration, which were statistically significant both before and after PSM (Table 1).

Primary Endpoint

The incidence of moderate-to-severe pain at rest and during coughing within 48 hours post-surgery is summarized in Table 2. After PSM, 19 patients (7.7%) in the sufentanil group experienced moderate-to-severe pain at rest, compared to 20 patients (2.8%) in the combination group (P < 0.05). Similarly, the incidence of pain during coughing was significantly higher in the sufentanil group (74 patients, 30.0%) than in the combination group (166 patients, 23.6%) during the same period (P < 0.05). These differences were also observed before PSM.

Table 2 Moderate-Severe Pain at Rest and During Coughing After Surgery Between the Sufentanial Group and the Combination Group

At 48 hours post-surgery, NRSR was significantly higher in the sufentanil group (1.97 ± 1.26) compared to the combination group (1.77 ± 0.91) (P= 0.018). Similarly, NRSC at 48 hours was higher in the sufentanil group (3.13 ± 1.57) than in the combination group (2.89 ± 1.17) (P= 0.022). All four analytical approaches including univariate and multivariate logistic regression analyses, post- PSM analysis and IPW analysis consistently identified sufentanil monotherapy as an independent predictor of moderate-to-severe pain, with odds ratios (ORs) and 95% confidence intervals (CIs) presented in Table 3.

Table 3 Logistic Regression Results for Moderate-Severe Pain at Rest and During Coughing After Surgery Between the Sufentanial Group and the Combination Group

Secondary Endpoints

Significant differences in the incidence of pain at rest and during coughing were observed at 24 and 72 hours post-surgery before PSM (P < 0.05). After PSM, these differences remained significant, except for pain during coughing at 72 hours (Table 2). No significant inter-group differences were noted in vomiting, hypotension, dizziness, delirium, or rescue analgesia on the first and second postoperative days, either before or after PSM. However, the functional activity scale (FAS) scores on the first and second postoperative days revealed significant differences between the two groups. Additionally, the proportion of fully alert patients on the second postoperative day was significantly higher in the combination group compared to the sufentanil group, both before and after PSM (Table 4).

Table 4 Adverse Events Between the Sufentanial Group and the Combination Group

Discussion

Pancreatic surgery is a critical intervention for pancreatic cancer, yet patients often experience prolonged postoperative pain, which can hinder physical and mental recovery. Effective pain management is therefore essential for improving patient outcomes and has garnered significant clinical attention. Opioid-based analgesia, particularly sufentanil, is widely used in patient-controlled intravenous analgesia (PCIA). However, the adverse effects of opioids, such as addiction, respiratory depression, pruritus, and sedation, have driven the search for alternative strategies to reduce opioid dosages and minimize side effects.¹⁷ Multimodal analgesia has emerged as a promising approach.¹⁸

In this propensity score-matched study, we evaluated the efficacy of combining sufentanil with dezocine in PCIA for postoperative pain management in patients undergoing pancreatic surgery. After matching, baseline characteristics and perioperative outcomes were comparable between the groups. Our findings demonstrated that the sufentanil-dezocine combination significantly reduced the incidence of moderate-to-severe pain at rest and during coughing within the first 48 hours postoperatively, without increasing the risk of clinically relevant side effects such as vomiting, hypotension, dizziness, delirium, or the need for rescue analgesia. Patients in the combination group exhibited significantly lower NRSR and NRSC at 48 hours post-surgery compared to the sufentanil group. Multivariate logistic regression analysis identified sufentanil monotherapy as an independent predictor of postoperative pain, suggesting that the addition of dezocine enhances analgesic efficacy, consistent with previous findings.¹⁶ These findings align with dezocine’s proposed mechanism: by targeting κ-receptors (which modulate pain perception) and partially activating μ-receptors (avoiding overstimulation), the combination may enhance analgesia while mitigating pure μ-agonist-related side effects. Notably, the reduction in pain during coughing–a high-pain activity critical for pancreatic surgery recovery–suggests the combination may be particularly beneficial for patients requiring early mobilization.

A primary concern with combining dezocine and sufentanil in PCIA is the potential for excessive sedation. However, our study found no evidence of increased sedation in the combination group during the 48-hour postoperative period. While sedation levels on the first postoperative day did not differ significantly, the proportion of fully alert patients was significantly higher in the combination group on the second postoperative day. This finding suggests that dezocine may enhance patient alertness while maintaining effective analgesia–Its ability to improve alertness and reduce sedation-related complications supports its value as a “balanced” adjunct in postoperative pain management^19–21 likely due to κ-receptor activation inducing lighter sedation compared to μ-agonists.

Postoperative adverse events, such as vomiting, hypotension, and dizziness, can negatively impact patient satisfaction and prolong hospital stays.¹⁰ Our study found that the addition of dezocine to sufentanil did not exacerbate these side effects. Notably, the combination group had a significantly lower incidence of respiratory depression compared to the sufentanil group, with no significant differences in vomiting, hypotension, dizziness, or delirium. These results align with previous research^22–25 and further support the safety profile of the sufentanil-dezocine combination.

Despite these promising findings, several limitations should be acknowledged. First, the retrospective design of the study introduces potential for selection bias, although this was mitigated by propensity score matching and the uniformity of our surgical team, Sensitivity analyses using alternative matching strategies (eg inverse probability weighting, multivariate logistic regression) yielded consistent results, suggesting no major residual confounding affected our conclusions. Second, generalizability of our findings may be limited. Due to our single-center design, even though our cohort meets high-volume criteria. As emphasized in a recent review on gastric cancer surgery outcomes, institutional factors can create variability in textbook outcomes (TOs) even among high-volume centers, highlighting the need for cross-institutional validation.²⁶ Future multi-center collaborations will compare textbook outcomes across 10+ high-volume centers using a pragmatic, standardized protocol to address this gap. Third, retrospective data precluded optimization of sufentanil/dezocine dosing. Prospective dose-response studies are needed to refine postoperative pain management in high-risk surgical populations.

In conclusion, our study demonstrates that the sufentanil-dezocine combination in PCIA significantly reduces moderate-to-severe pain at rest and during coughing within the first 48 hours after pancreatic surgery, without increasing the incidence of clinically relevant adverse effects, which has not been previously reported in the context of pancreatic surgery, suggesting it as a promising and safe approach for postoperative pain management in pancreatic cancer patients. Future research should focus on optimizing dosing strategies and confirming these results in prospective, multicenter trials.

Data Sharing Statement

The original contributions presented in the study are included in the article; further inquiries can be directed to the corresponding authors.

Funding

This work was supported by the National Natural Science Foundation of China (No: T2293734).

Disclosure

The authors declare that they have no conflicts of interest in this work.

References

1. Zhang Q, Zeng L, Chen Y, et al. Pancreatic cancer epidemiology, detection, and management. Gastroenterol Res Pract. 2016;2016:10. doi:10.1155/2016/8962321

2. van Dijk SM, Heerkens HD, Tseng DSJ, et al. Systematic review on the impact of pancreatoduodenectomy on quality of life in patients with pancreatic cancer. HPB. 2018;20(3):204–215. doi:10.1016/j.hpb.2017.11.002

3. Karamarković AR, Juloski JT. Current surgical concepts and future perspectives in the treatment of borderline resectable and potentially resectable locally advanced pancreatic cancer. Chirurgia. 2022;117(4):385–398. doi:10.21614/chirurgia.2770

4. Min EK, Chong JU, Hwang HK. Negative oncologic impact of poor postoperative pain control in left-sided pancreatic cancer. World J Gastroenterol. 2017;23(4):676–686. doi:10.3748/wjg.v23.i4.676

5. Song JW, Shim JK, Song Y, et al. Effect of ketamine as an adjunct to intravenous patient-controlled analgesia, in patients at high risk of postoperative nausea and vomiting undergoing lumbar spinal surgery. Br J Anaesth. 2013;111:630–635. doi:10.1093/bja/aet192

6. Klotz R, Larmann J, Klose C, et al. Gastrointestinal complications after pancreatoduodenectomy with epidural vs patient-controlled intravenous analgesia: a randomized clinical trial. JAMA Surg. 2020;155(7):e200794. doi:10.1001/jamasurg.2020.0794

7. Gostian M, Loeser J, Bentley T, et al. Analgesia after tonsillectomy with controlled intravenous morphine – overdue or exaggerated? Braz J Otorhinolaryngol. 2023;89(1):48–53. doi:10.1016/j.bjorl.2021.08.002

8. Liu F, Li TT, Yin L, et al. Analgesic effects of sufentanil in combination with flurbiprofen axetil and dexmedetomidine after open gastrointestinal tumor surgery: a retrospective study. BMC Anesthesiol. 2022;22(1):130. doi:10.1186/s12871-022-01670-0

9. Lindemann C, Strube P, Fisahn C, et al. Patient-controlled sublingual sufentanil tablet system versus intravenous opioid analgesia for postoperative pain management after lumbar spinal fusion surgery. Eur Spine J. 2023;32(1):321–328. doi:10.1007/s00586-022-07462-x

10. Zhu H, Chen YB, Huang SQ, et al. Interaction of analgesic effects of dezocine and sufentanil for relief of postoperative pain: a pilot study. Drug Des Devel Ther. 2020;14:4717–4724. doi:10.2147/DDDT.S270478

11. Xia Y, Sun Y, Liu J. Effects of dezocine on PAED scale and Ramsay sedation scores in patients undergoing NUSS procedure. Am J Transl Res. 2021;13(5):5468–5475.

12. Ye RR, Jiang S, Xu X, et al. Dezocine as a potent analgesic: overview of its pharmacological characterization. Acta Pharmacol Sin. 2022;43(7):1646–1657. doi:10.1038/s41401-021-00790-6

13. Yu F, Zhou J, Xia S, et al. Dezocine prevents postoperative hyperalgesia in patients undergoing open abdominal surgery. Evid Based Complement Alternat Med. 2015;2015:946194. doi:10.1155/2015/946194

14. Sun ZT, Yang CY, Cui Z, et al. Effect of intravenous dezocine on fentanyl-induced cough during general anesthesia induction: a double-blinded, prospective, randomized, controlled trial. J Anesth. 2011;25:860–863. doi:10.1007/s00540-011-1237-x

15. Zhu Y, Jing G, Yuan W. Preoperative administration of intramuscular dezocine reduces postoperative pain for laparoscopic cholecystectomy. J Biomed Res. 2011;25:356–361. doi:10.1016/S1674-8301(11)60047-X

16. Wu L, Dong YP, Sun L, Sun L. Low concentration of dezocine in combination with morphine enhance the postoperative analgesia for thoracotomy. J Cardiothorac Vasc Anesth. 2015;29(4):950–954. doi:10.1053/j.jvca.2014.08.012

17. Li QZ, Yao HX, Xu MY, et al. Dedetomidine combined with sufentanil and dezocine-based patient controlled intravenous analgesia increases female patients’ global satisfaction degree after thoracoscopic surgery. J Cardiothorac Surg. 2021;16(1):102. doi:10.1186/s13019-021-01472-4

18. Gritsenko K, Khelemsky Y, Kaye AD, et al. Multimodal therapy in perioperative analgesia. Best Pract Res Clin Anaesthesiol. 2014;28(1):59–79. doi:10.1016/j.bpa.2014.03.001

19. Barr GA, Schmidt HD, Thakrar AP, Kranzler HR, Liu R. Revisiting dezocine for opioid use disorder: a narrative review of its potential abuse liability. CNS Neurosci Ther. 2024;30(9):e70034. doi:10.1111/cns.70034

20. Schmidt HD, Zhang Y, Xi J, et al. A new formulation of dezocine, Cycdezocine, reduces oxycodone self-administration in female and male rats. Neurosci Lett. 2023;815:137479. doi:10.1016/j.neulet.2023.137479

21. Grothusen J, Lin W, Xi J, et al. Dezocine is a biased ligand without significant beta-arrestin activation of the mu opioid receptor. Transl Perioper Pain Med. 2022;9(1):424–429.

22. Wang CY, Li LZ, Shen BX, et al. A multicenter randomized double-blind prospective study of the postoperative patient controlled intravenous analgesia effects of dezocine in elderly patients. Int J Clin Exp Med. 2014;7(3):530–539.

23. He LX, Yao YT, Shao K, et al. Efficacy of dezocine on preventing opioid-induced cough during general anaesthesia induction: a PRISMA-compliant systematic review and meta-analysis. BMJ Open. 2022;12(4):e052142. doi:10.1136/bmjopen-2021-052142

24. Zhang L, Li C, Zhao C, et al. Analgesic comparison of dezocine plus propofol versus fentanyl plus propofol for gastrointestinal endoscopy: a meta-analysis. Medicine. 2021;100(15):e25531. doi:10.1097/MD.0000000000025531

25. Gui YK, Zeng XH, Xiao R, et al. The Effect of dezocine on the median effective dose of sufentanil-induced respiratory depression in patients undergoing spinal anesthesia combined with low-dose dexmedetomidine. Drug Des Devel Ther. 2023;17:3687–3696. doi:10.2147/DDDT.S429752

26. Marano L, Verre L, Carbone L, et al. Current trends in volume and surgical outcomes in gastric cancer. J Clin Med. 2023;12(7):2708. doi:10.3390/jcm12072708

Introduction

Hepatocellular carcinoma (HCC) is one of the most prevalent and lethal malignancies worldwide, with hepatectomy remaining a primary curative treatment option for early-stage disease.^1,2 However, postoperative recurrence poses a significant challenge, with 5-year recurrence rate as high as 60%.³ Recurrence can be categorized as early recurrence (<2 years), caused by micro metastases after resection, or late recurrence (>2 years), caused by new tumors arising in a microenvironment prone to carcinogenesis.⁴ Research has shown that HCC patients who experience early recurrence post-resection have a worse prognosis compared to those with late recurrence.⁵ Therefore, accurate prediction of early recurrence is critical for personalized postoperative surveillance and adjuvant therapy strategies.

Certain clinicopathological features, including the Barcelona Clinic Liver Cancer (BCLC) staging system, alpha-fetoprotein (AFP), microvascular invasion (MVI), and satellite nodules (SNs), have been linked to postoperative prognosis.^1,6 However, conventional predictive assessments based on these biomarkers lack sufficient predictive accuracy, and information on MVI and SNs is not available preoperatively. Several recent studies have used artificial intelligence models based on CT/MRI data to effectively predict the risk of postsurgical HCC recurrence risk.^7–10 Compared with CT and MRI, ultrasound possesses the advantages of real-time, radiation-free, and cost-effective, and can also be used for intraoperative evaluation and guide the operation pathway. Additionally, contrast-enhanced ultrasound (CEUS) has emerged as a valuable imaging modality for HCC, offering real-time dynamic assessment of tumor vascularity and perfusion characteristics.¹¹

Previous studies have reported that certain CEUS characteristics such as rapid wash-in, early wash-out phase, rapid portal phase signal regression, and high LI-RADS grading are risk factors associated with early recurrence after radical HCC resection.^12,13 However, a critical gap remains: the interpretation of these characteristics is highly subjective and difficult to quantify, which limits their consistent application in prognostication. To address this gap in objectivity, deep learning (DL) emerges as a powerful tool. DL algorithms can automatically extract massive, quantitative features from medical images to model disease diagnosis and prognosis.¹⁴ The application of DL to liver CEUS is already established for tasks such as focal liver lesion classification and predicting HCC biological behavior,^15,16 demonstrating its potential in this domain. However, these existing applications primarily focus on classification limited to single-phase analysis or static images, potentially overlooking the critical time-dependent kinetic information. The specific knowledge gap our study aims to fill is the use of DL to predict long-term clinical outcomes directly from pre-operative dynamic CEUS cines. Therefore, we hypothesize that a DL-based framework, by leveraging the rich temporal data in multiphase CEUS, can overcome the limitations of subjective interpretation. This study aims to develop and validate such a framework, integrating CEUS data with clinicopathological variables to provide a noninvasive and efficient preoperative tool for predicting early recurrence in patients with early-stage HCC after surgical resection.

Methods

Study Population and Design

This study was performed in accordance with the principles of the Declaration of Helsinki. This retrospective study was approved by the Ethics Committee of the Second Affiliated Hospital of Nanchang University (Approval No: IIT-O-2025-251). The requirement for informed consent was waived by the Ethics Committee due to the retrospective and minimal-risk nature of the study. All patient data were kept strictly confidential and de-identified prior to analysis. From January 2010 to January 2023, 155 patients with HCC who underwent CEUS within two weeks before hepatectomy were enrolled. The diagnosis of HCC was based on the diagnostic criteria established by the European Association for the Study of the Liver.² Inclusion criteria were: BCLC stage 0 or A; liver function with Child-Pugh class A or B; performance status Eastern Cooperative Oncology Group score 0 or 1; R0 resection;¹⁷ regular follow-up within two years after hepatectomy. The exclusion criteria were poor CEUS imaging quality, recurrent HCC or primary HCC combined with other primary tumors, and a history of antitumor treatment before surgery, such as radiofrequency ablation, microwave ablation, transcatheter arterial chemoembolization, or chemotherapy. Pretreatment baseline clinical characteristics, including demographic data, laboratory test results, and clinical diagnoses, were collected from the Institutional Picture Archiving and Communication System (PACS^®; Carestream Health, Toronto, Canada). Finally, 115 patients’ preoperative CEUS cines and clinical variables were retrospectively analyzed. We developed four CEUS-based AI models to preoperatively predict early recurrence: CEUS-AP, CEUS-PP, CEUS-LP, and CEUS-MP. We further integrated clinical variables into the CEUS AI model to construct an individualized nomogram for preoperative prediction of early recurrence. The patient enrollment workflow and study design were shown in Figure 1.

Figure 1 Study workflow flowchart. Abbreviations: CEUS, contrast-enhanced ultrasound; HCC, hepatocellular carcinoma; ROI annotation illustration; AP, arterial phase; PP, portal venous phase; LP, late phase; MP, multiple phases.

CEUS Data Acquisition

CEUS was performed by one of two radiologists (ZH and YC) with over 10 years of experience in liver CEUS. Two types of ultrasound instruments were used in this study, Toshiba Aplio (n = 98) and SuperSonic Imagine (n = 17). The system parameters, including gain, dynamic range, mechanical index, output power, and focal zone, were fine-tuned to ensure effective tissue signal reduction without compromising penetration capability. A volume of 2.4 mL of the second-generation contrast agent (SonoVue^®; Bracco Imaging, Milan, Italy) was injected within one second via the elbow, followed by a 5-mL saline flush. The transducer was maintained in a stable position to continuously observe perfusion of the lesion. Three-phase contrast enhancement was dynamically monitored, consisting of arterial phase (AP, 0–30 s), portal venous phase (PP, 31–120 s), and late phase (LP, 121–360 s). The CEUS digital video recordings were stored as DICOM format.

Hepatectomy Procedure and Follow-Up Protocol

Anatomical partial hepatectomy was performed with a resection margin of ≥ 10 mm. All enrolled tumors were safely and completely removed, with no residual tumors following resection (R0 classification). The patients underwent regular follow-up assessments at 1, 3, 6, 9, and 12 months postoperatively, followed by evaluations every 3–6 months thereafter. During these visits, contrast-enhanced imaging (CEUS, CECT, or CEMRI) and serum biomarker testing (AFP and liver function) were performed. In this study, early recurrence was defined as a new lesion with typical imaging features of HCC that appeared in imaging examinations within 2 years after hepatectomy or as confirmed by pathological results from percutaneous liver biopsy or a second surgery.

Developing Prediction Model Based on CEUS

This study employed CEUS cines from three phases (arterial phase, portal venous phase, and late phase) to construct the corresponding single-phase models (CEUS-AP, CEUS-PP, and CEUS-LP). We then developed a comprehensive CEUS-MP (multi-phase) model that integrated all three phases. In each patient, three separate regions of interest (ROIs) were outlined by a radiologist with 5-year experience in CEUS to enable quantitative assessment of the CEUS videos across the arterial, portal venous, and late phases. The annotation tool was ITK-SNAP, which is an open-source software.¹⁸ Specifically, the doctor first annotated the contour of the tumor in the CEUS frame when the tumor displayed clearly. A rectangle to cover the tumor contour and an approximately 0.5 cm wide liver parenchyma surrounding the tumor were created (Figure 2). All frames with the same ROI location determined in the annotated frame were cropped and manually adjusted if necessary. To assess inter-observer variability in ROI delineation, two doctors (one with more than 10 years of experience in CEUS and the other with more than 5 years of experience) independently contoured the ROIs for a randomly selected cohort of 30 patients, blinded to patient outcomes and clinical characteristics. Radiomic features extracted from both sets of ROIs were compared using the intraclass correlation coefficient (ICC) to determine feature reproducibility.

Figure 2 ROI annotation illustration. (a–c) An example of the red/green/blue rectangle ROI segmented in one AP/PP/LP CEUS frame when the tumor displayed clearly. Abbreviations: ROI, region of interest; AP, arterial phase; PP, portal venous phase; LP, late phase; CEUS, contrast-enhanced ultrasound.

Enrolled patients were randomly divided into training (n = 75) and validation (n = 40) cohorts. To minimize model-induced bias, convolutional neural networks (CNN) were employed for the four models based on CEUS cines in different phases consisting of a two-dimensional (2D) convolution layer, 2D max-pooling layer, ReLU activation function, vector of locally aggregated descriptors (VLAD), and a fully connecting layer (FCL) (Figure 3). The inputs of the CEUS-AP, CEUS-PP, CEUS-LP, and CEUS-MP models were different CEUS cines in the arterial, portal venous, late, and the above-mentioned three phases, respectively. In the four CEUS AI models, convolution layers automatically learn the imaging features from each frame of CEUS, and VLAD is responsible for the quantitative analysis of enhancement patterns over time for the CEUS cines. The FCL guide model predicted the probability of early recurrence for each patient. A ten-fold validation method was adopted in the training phase to optimize the four CEUS AI models. To improve the prediction performance and reduce the risk of overfitting, CEUS cines were split over time into multiple independent samples at two frames per second; in this way, the training population was augmented for better learning. Data augmentation techniques, including random rotations, flips, and intensity variations, were employed to augment the diversity of the training dataset. Moreover, depth-wise convolution was selected to reduce the number of parameters in the training phase. Both L1 and L2 regularization were employed to combat overfitting. L1 regularization promoted sparsity and can perform feature selection by driving less important weights to zero, while L2 regularization penalized large weights to ensure the model was less sensitive to small fluctuations in the input data. This combined approach was chosen to create a more robust and generalizable model given the high-dimensional feature space relative to our sample size. We implemented early stopping based on the validation loss to halt training once performance on the validation set ceased to improve. In this study, the convolutional layers employed a depthwise separable convolution structure to replace the conventional 3D convolution operations, separating the processes of feature convolution filtering and the integration of cross-channel features. A global pooling layer was incorporated before the fully connected layers, aggregating features across temporal and spatial dimensions to reduce the number of neural nodes that the fully connected layers need to process. Through a series of specialized network design strategies, the parameter count of R-DLCEUS was reduced to approximately 20,000. During the training phase, the model used cross-entropy as the objective function and optimized the parameters using the stochastic gradient descent method. Each batch consisted of 16 samples, and the model was trained for 70 epochs. The initial learning rate was set to 0.001 and was reduced by a factor of 0.5 at the 15th, 30th, and 55th epochs. A momentum coefficient of 0.9 was applied during training. For the CEUS-MP model, the single-channel cine loops from the AP, PP, and VP were concatenated along the channel dimension to form a multi-channel input, which was then processed by the subsequent 2D convolutional network. The area under the receiver operating characteristic curve (AUC) was adopted to quantitatively measure the prediction performance of the CEUS AI models.

Figure 3 The overall network structure of CEUS AI models. CNN were employed for the four models based on CEUS cines in different phases. Input step: The CEUS cines of AP/ PP/LP/MP were inputted into the CNN model, respectively. Feature extraction step: The features extracted from three phases were then integrated into one feature collection which demonstrated the characteristics of the entire CEUS cine. 2D Conv x included 2D convolution layer, 2D max-pooling layer, and ReLU activation function. Output step: probability was calculated to estimate risk of early recurrence after hepatectomy. Abbreviations: CEUS, contrast-enhanced ultrasound; AI, artificial intelligence; CNN, convolutional neural network; AP, arterial phase; PP, portal phase; LP, late phase; MP, multiple phases; 2D Conv, two-dimensional convolution; VLAD, vector of locally aggregated descriptors; FCL, full connected layer.

Building Combined Individualized Nomogram

To incorporate information from clinical variables with CEUS cines, an individualized model was developed to preoperatively predict early recurrence after hepatectomy. Specifically, four models developed from CEUS video data (CEUS-AP, CEUS-PP, CEUS-LP, and CEUS-MP) were evaluated to identify the optimal predictive model. Subsequently, the predictive probability of the selected model was used as a Radiomics signature and integrated with clinical variables to build a combined nomogram prediction model based on the multi-variable logistic regression analysis. In order to demonstrate the incremental value of DL-CEUS-based nomogram, univariate and multivariate logistic regression analyses were performed to build the Clinical-only model. The analyzed clinical variables included 16 basic clinical characteristics (Table 1). Variable selection followed a two-stage process. First, all parameters showing marginal significance (P < 0.10) in univariate analyses were included; subsequently, these were incorporated into a stepwise multivariable regression model. Only variables demonstrating statistical significance (P < 0.05) in the final multivariable analysis were retained as independent predictors and were used for combined nomogram development. The prediction performance of the nomogram was assessed using AUC, calibration curves, and decision curves.

Table 1 Baseline Characteristics of Patients

Visualization of the CEUS-MP Model

To enhance the interpretability of the DL model’s predictions regarding early recurrence, we employed Selvaraju R.’s method to transform the DL feature maps into pseudo-colored visualization maps.¹⁹ In these maps, warm-colored (red) pixels represent regions with stronger predictive relevance, indicating high-weight features that significantly contribute to the model’s output. Conversely, the cool-colored (blue) pixels denote areas of weaker correlation corresponding to the low-weight factors in the prediction. This visualization approach effectively highlights the critical image regions that influence the model’s decision-making process.

Statistical Analysis

The statistical software and packages used were Python (version 3.11), PyTorch (version 2.0.1), R (version 3.4.4), and computeC. The chi-square test was used to compare categorical variables. Student’s t-test or the Mann–Whitney test, as appropriate, was used to compare continuous variables. All statistical tests were two-sided, and differences were considered significant at P < 0.05.

Results

Clinical Characteristics

Among 115 patients, there were 93 males (93/115, 80.9%) and 22 females (22/115, 19.1%). The mean tumor size was 3.1 ± 0.9 cm (range, 1.0–5.0 cm). Postoperative follow-up results revealed that among the115 patients, 30 (30/115, 26.1%) experienced early recurrence, whereas 85 (85/115, 73.9%) showed no evidence of early recurrence. There were no significant differences in the baseline characteristics between the training (n = 75) and validation (n = 40) cohorts. Detailed baseline characteristics of the two cohorts are shown in Table 1.

Predictive Performance of CEUS AI Models

The features demonstrated high reproducibility, with 91% achieving excellent ICC values, indicating strong inter-observer agreement in ROI segmentation. In the training cohort, AUCs of CEUS-MP, CEUS-AP, CEUS-PP, and CEUS-LP reached 0.922 (95% CI: 0.816–0.971), 0.829 (95% CI: 0.712–0.954), 0.816 (95% CI: 0.697–0.941) and 0.808 (95% CI: 0.691–0.938), respectively. The DL-based AI model using multiple CEUS cines (CEUS-MP) achieved the best prediction performance compared with the other three models based on single-phase CEUS cine (CEUS-AP, CEUS-PP, and CEUS-LP) (Figure 4a). Similar results were observed in the validation cohort. In the validation cohort, the AUCs of CEUS-MP, CEUS-AP, CEUS-PP, and CEUS-LP were 0.840 (95% CI: 0.701–0.989), 0.741 (95% CI: 0.670–0.921), 0.719 (95% CI: 0.637–0.909), and 0.703 (95% CI: 0.661–0.889), respectively (Figure 4b). Table 2 provides an overview of the quantitative assessment of the four models. CEUS-MP offered significantly higher AUCs than CEUS-AP, CEUS-PP, and CEUS-LP in both training (P = 0.026, P = 0.014, P = 0.001) and validation (P = 0.029, P = 0.016, P = 0.001) cohorts. However, no significant differences in AUCs were observed between CEUS-AP, CEUS-PP, and CEUS-LP in either the training or validation cohorts (all P > 0.05).

Table 2 Predictive Performance of CEUS-MP, CEUS-AP, CEUS-PP, and CEUS-LP in Training and Validation Cohorts

Figure 4 ROC curves of four CEUS AI models. (a) ROC curves of CEUS AI models in training cohort. (b) ROC curves of CEUS AI models in validation cohort. CEUS-MP achieved the best prediction performance compared with CEUS-AP, CEUS-PP, and CEUS-LP in both cohorts.

The Individualized Predictive Value of Combined Nomogram

Multivariable regression analysis revealed that serum albumin (< 35 g/L) (HR = 1.146, P = 0.034), AFP (> 1000 ng/mL) (HR = 1.908, P = 0.028), and the CEUS-MP signature (HR = 5.141, P < 0.001) were independent preoperative predictors of early recurrence (Table 3). An individualized nomogram was built based on these three significant values (Figure 5a). In the training and validation cohorts, the AUCs of the nomogram for early recurrence was 0.945 (95% CI: 0.861–0.965) and 0.871 (95% CI: 0.751–0.970) for early recurrence, respectively. (Figure 5b). The nomogram achieved a sensitivity of 88.9% (95% CI: 75.9–96.1%) and specificity of 98.1% (95% CI: 81.6–99.1%) in the training cohort. In the validation cohort, the sensitivity and specificity of the nomogram were 83.0% (95% CI: 70.4–90.8%) and 82.5% (95% CI: 70.8–91.5%), respectively. The nomogram showed good calibration for early recurrence prediction (Figure 5c). The Hosmer-Lemeshow test showed that there were no significant differences in early recurrence prediction between the nomogram and the ideal reference curve (P = 0.415). The decision curve analysis of the nomogram is presented in Figure 5d. AFP and serum albumin were selected to construct a Clinical-only model. The Clinical-only model had an AUC of 0.573 (95% CI: 0.501–0.675) in the training cohort with a 0.554 (95% CI: 0.512–0.685) accuracy and an AUC of 0.528 (95% CI: 0.500–0.618) in the validation cohort with a 0.545 (95% CI: 0.509–0.678) accuracy. The nomogram achieved a significantly higher predictive performance compared to the Clinical-only model in both training and validation cohorts (Both P < 0.001). The decision curves showed that if the threshold probability of a patient was > 30%, integrating CEUS-MP signatures and clinical variables (AFP and serum albumin) to predict the patient’s early recurrence added more net benefit than using only clinical variables. In addition, the net benefit of the CEUS-MP signature was similar to that of the nomograms.

Table 3 Prognostic Factors to Predict Early Recurrence of HCC Patients After Hepatectomy

Figure 5 Prediction performance of individualized nomogram, ROC curves, calibration curves and Decision curve analysis. (a) Nomogram for predicting the probability of a patient occur early recurrence after hepatectomy. A patient who obtained high total scores of the three factors tends to have high probability of early recurrence after hepatectomy. ALB, 0 means < 35 g/L, 1 means ≥ 35 g/L; AFP, 0 means < 200ng/ mL, 1 means ≥ 200ng/mL. (b) ROC curves of nomogram in training and validation cohort. (c) Calibration curves of the nomogram for predicting early recurrence in training and validation cohorts. (d) Decision curve analysis for nomogram.

Understanding the R-DLCEUS Model

From the visualization maps of the CEUS-LP model, it can be observed that the model focuses more on the early period AP, suggesting its potential ability to capture the “fast-in” feature, which may be closely associated with the early recurrence of HCC. Additionally, we observed that for some early cases, the attention of the model exhibited a “patchy” pattern (Figure 6). We hypothesize that these heterogeneous attention regions may correspond to areas of necrotic or fibrotic tissue, features often associated with aggressive tumor biology. However, this requires further pathological correlation.

Figure 6 Visualization of the CEUS-LP model. (a–c) Representative time-point images (9s, 12s, 15s; early period AP) from dynamic CEUS in a post-hepatectomy patient with early recurrence. Each set of images displayed, from left to right, the original monochrome ultrasound image, the pseudo-color heatmap, and the heatmap overlaid on the original ultrasound image.

Discussion

This study developed a DL-based model using CEUS to preoperatively predict early recurrence in patients with early-stage HCC patients following hepatectomy. Our results demonstrated that the multiphase CEUS model (CEUS-MP) achieved superior predictive performance compared with single-phase models, with AUC of 0.922 and 0.840 in the training and validation cohorts, respectively. Furthermore, the integration of CEUS-MP signatures with clinical variables (serum albumin and AFP) into a nomogram further enhanced predictive accuracy, yielding AUCs of 0.945 and 0.871 in the training and validation cohorts, respectively. These findings highlighted the potential of CEUS combined with DL as an auxiliary method for preoperative risk stratification in patients with HCC.

While conventional ultrasound remains the first-line imaging tool for liver screening, its inability to assess the tumor vasculature limits its diagnostic and prognostic utility. Since CEUS cines overcome these limitations by providing dynamic, contrast-enhanced visualization of tumor microvasculature, leading to more accurate HCC diagnosis, recurrence prediction, and treatment monitoring.²⁰ Given these advantages, we utilized CEUS cines, rather than conventional ultrasound images, to develop our prediction models in this study, with the help of advanced quantification techniques for more precise analysis. The superior performance of the CEUS-MP model highlighted the importance of leveraging multiphase CEUS data, which capture dynamic tumor perfusion characteristics across the arterial, portal venous, and late phases. This aligns with previous studies emphasizing the prognostic value of CEUS features, such as rapid wash-in and early wash-out, in HCC recurrence prediction.¹² However, traditional CEUS interpretation is limited by subjectivity and inter-observer variability. Our DL approach addressed this limitation by automatically extracting quantitative features from CEUS cines, thereby standardizing the analysis and improving the reproducibility. Huang et al conducted a quantitative analysis of CEUS images of patients with hepatocellular carcinoma (HCC) following radical resection to predict early recurrence.²¹ However, their model achieved an AUC of only 0.57 in the testing cohort, which was significantly lower than that of our proposed model. A potential limitation of their approach was the reliance on a single-frame analysis, the peak contrast intensity of the lesion on CEUS, which may have decreased predictive accuracy. By contrast, our study leveraged continuous multiphase CEUS cine imaging, enabling a more comprehensive assessment and significantly improving the precision of early recurrence prediction.

The integration of CEUS-MP signatures with clinical variables into a nomogram represented a significant advancement in personalized HCC management. Multivariate analysis identified serum albumin and AFP as the independent clinical prognostic predictors. Notably, low ALB levels served as a reliable indicator of advanced hepatic dysfunction, reflecting both the progression of liver carcinogenesis and deteriorating hepatic synthetic capacity.²² Concurrently, elevated AFP served as a biological marker indicating high invasiveness in HCC.²³ In such patients, some microlesions may escape detection by imaging examinations during the initial HCC diagnosis, contributing to early recurrence after surgical resection. This study demonstrated that higher AFP levels were correlated with an increased risk of early tumor recurrence. Therefore, the AFP level held a significant value in predicting the prognosis of therapy. The high sensitivity and specificity of the nomogram in both cohorts suggested its clinical utility in identifying high-risk patients who may benefit from other treatment selections or intensified surveillance. Notably, the decision curve analysis indicated that the nomogram provided a greater net benefit than the clinical variables alone when the threshold probability exceeded 30%, supporting its practical application in clinical decision-making. The visualization maps further enhanced interpretability, revealing that the model focused on regions with heterogeneous enhancement patterns, which may correlate with aggressive tumor biology. The heterogeneous or peripheral enhancement patterns highlighted by the model were often associated with active tumor regions, while the lack of enhancement in certain areas may correlate with central necrosis or fibrosis commonly found in larger lesions.

Limitations

Our study has several limitations. First, it is a retrospective study conducted at a single center, which introduces potential selection bias. Our patient cohort may not fully represent the more heterogeneous patient populations seen across different healthcare institutions. Second, the uniform high-standard scanning protocols and specialized operators may have inflated the performance estimates of our model. Moreover, the relatively small cohort, particularly the limited number of early recurrence events, is a recognized limitation that warrants caution in interpreting the findings and underscores the need for future validation in larger populations. Furthermore, the use of two different ultrasound systems for CEUS acquisition, despite our efforts at intensity normalization, represents a potential technical confounder. Therefore, the generalizability of our findings needs to be further validated in real-world, multi-center settings, and should also explore incorporating additional imaging modalities (eg, MRI or CT) and molecular biomarkers to refine predictive accuracy.

Conclusion

In conclusion, our study demonstrates that a DL-based CEUS framework, particularly when combined with clinical variables, shows strong potential for the preoperative prediction of early recurrence in patients with HCC undergoing hepatectomy. While these initial results are promising, further validation in large-scale, multi-center, prospective cohorts is essential to confirm its generalizability and establish clinical utility. If validated, this approach could facilitate personalized postoperative management and potentially improve long-term patient outcomes. Future work should also focus on integrating this model with other imaging modalities, such as CT or MRI, to further enhance predictive accuracy.

Ethics Approval and Consent to Participate

This study was approved by the Institutional Review Board of The Second Affiliated Hospital of Nanchang University.

Acknowledgments

The authors thank the data collectors for their efforts and interest in participating in data collection. We would like to thank the patients who willingly provided all the necessary information without any reservation.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Funding

This study was supported by grants from the National Natural Science Foundation of China (Grant No. 82360348), the Jiangxi Provincial Natural Science Foundation (Grant No. 20232BAB216096), and the Jiangxi Ganpo Outstanding Talent Support Program–Academic and Technical Discipline Leader Development Project (Grant No. 20243BCE51175).

Disclosure

All authors have no conflicts of interest to declare in this work.

References

1. Forner A, Reig M, Bruix J. Hepatocellular carcinoma. Lancet. 2018;391(10127):1301–1314. doi:10.1016/S0140-6736(18)30010-2

2. European Association for the Study of the Liver. EASL clinical practice guidelines: management of hepatocellular carcinoma. J Hepatol. 2018;69(1):182–236. doi:10.1016/j.jhep.2018.03.019

3. Lim KC, Chow PK, Allen JC, et al. Systematic review of outcomes of liver resection for early hepatocellular carcinoma within the Milan criteria. Br J Surg. 2012;99:1622–1629. doi:10.1002/bjs.8915

4. Llovet JM, Kelley RK, Villanueva A, et al. Hepatocellular carcinoma. Nat Rev Dis Primers. 2021;7(1):6. doi:10.1038/s41572-020-00240-3

5. Poon RT, Fan ST, Ng IO, et al. Different risk factors and prognosis for early and late intrahepatic recurrence after resection of hepatocellular carcinoma. Cancer. 2000;89(3):500–507.

6. Villanueva A. Hepatocellular carcinoma. N Engl J Med. 2019;380(15):1450–1462. doi:10.1056/NEJMra1713263

7. Lee IC, Huang JY, Chen TC, et al. Evolutionary learning-derived clinical-radiomic models for predicting early recurrence of hepatocellular carcinoma after resection. Liver Cancer. 2021;10(6):572–582. doi:10.1159/000518728

8. Kucukkaya AS, Zeevi T, Chai NX, et al. Predicting tumor recurrence on baseline MR imaging in patients with early-stage hepatocellular carcinoma using deep machine learning. Sci Rep. 2023;13(1):7579. doi:10.1038/s41598-023-34439-7

9. Yuan C, Wang Z, Gu D, et al. Prediction early recurrence of hepatocellular carcinoma eligible for curative ablation using a Radiomics nomogram. Cancer Imaging. 2019;19(1):21. doi:10.1186/s40644-019-0207-7

10. Hui TCH, Chuah TK, Low HM, et al. Predicting early recurrence of hepatocellular carcinoma with texture analysis of preoperative MRI: a radiomics study. Clin Radiol. 2018;73(12):1056.e11–1056.e16. doi:10.1016/j.crad.2018.07.109

11. Alzaraa A, Gravante G, Chung WY, et al. Contrast-enhanced ultrasound in the preoperative, intraoperative and postoperative assessment of liver lesions. Hepatol Res. 2013;43(8):809–819. doi:10.1111/hepr.12044

12. Cao K, Wu L, Wang X, et al. Risk factors for early recurrence after radical resection of hepatocellular carcinoma based on preoperative contrast-enhanced ultrasound and clinical features. Technol Cancer Res Treat. 2024;23:15330338241281327. doi:10.1177/15330338241281327

13. Terzi E, Iavarone M, Pompili M, et al. Contrast ultrasound LI-RADS LR-5 identifies hepatocellular carcinoma in cirrhosis in a multicenter restropective study of 1,006 nodules. J Hepatol. 2018;68(3):485–492. doi:10.1016/j.jhep.2017.11.007

14. Wang K, Lu X, Zhou H, et al. Deep learning radiomics of shear wave elastography significantly improved diagnostic performance for assessing liver fibrosis in chronic hepatitis B: a prospective multicenter study. Gut. 2019;68(4):729–741. doi:10.1136/gutjnl-2018-316204

15. Ding W, Meng Y, Ma J, et al. Contrast-enhanced ultrasound-based AI model for multi-classification of focal liver lesions. J Hepatol. 2025;83(2):426–439. doi:10.1016/j.jhep.2025.01.011

16. Xu W, Zhang H, Zhang R, et al. Deep learning model based on contrast-enhanced ultrasound for predicting vessels encapsulating tumor clusters in hepatocellular carcinoma. Eur Radiol. 2025;35(2):989–1000. doi:10.1007/s00330-024-10985-0

17. Hermanek P, Wittekind C. Residual tumor (R) classification and prognosis. Semin Surg Oncol. 1994;10(1):12–20. doi:10.1002/ssu.2980100105

18. Yushkevich PA, Piven J, Hazlett HC, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. 2006;31(3):1116–1128. doi:10.1016/j.neuroimage.2006.01.015

19. Selvaraju RR, Cogswell M, Das A. Grad-CAM: visual explanations from deep networks via gradient-based localization. In: Proceedings of the IEEE International Conference on Computer Vision. 2017:618–626.

20. Kudo M, Ueshima K, Osaki Y, et al. B-Mode ultrasonography versus contrast-enhanced ultrasonography for surveillance of hepatocellular carcinoma: a prospective multicenter randomized controlled trial. Liver Cancer. 2019;8(4):271–280. doi:10.1159/000501082

21. Huang Z, Shu Z, Zhu RH, et al. Deep learning-based radiomics based on contrast-enhanced ultrasound predicts early recurrence and survival outcome in hepatocellular carcinoma. World J Gastrointest Oncol. 2022;14(12):2380–2392. doi:10.4251/wjgo.v14.i12.2380

22. Huang YH, Wu JC, Chen CH, et al. Comparison of recurrence after hepatic resection in patients with hepatitis B vs. hepatitis C-related small hepatocellular carcinoma in hepatitis B virus endemic area. Liver Int. 2005;25(2):236–241. doi:10.1111/j.1478-3231.2005.01081.x

23. Yang L, Gu D, Wei J, et al. A radiomics nomogram for preoperative prediction of microvascular invasion in hepatocellular carcinoma. Liver Cancer. 2019;8(5):373–386. doi:10.1159/000494099