Artificial Intelligence and Hepatocellular Carcinoma
- Conditions
- HCC
- Interventions
- Diagnostic Test: CT Scan Radiomics Features Extraction
- Registration Number
- NCT05637788
- Lead Sponsor
- Humanitas Clinical and Research Center
- Brief Summary
To identify new relevant biomarkers for HCC patients and their risk of recurrence. Radiomics data and computer-vision data will be explored for their ability to predict the presence of particular pathological signs of aggressiveness (microvascular invasion and satellitosis), and the prognosis after surgery.
- Detailed Description
Hepatocellular carcinoma (HCC) is 1 of the 5 most common malignancies worldwide and the third most common cause of cancer related mortality of 500,000 deaths globally every year.
Although more common in East Asia, the incidence of HCC is increasing in the Western world. Hepatic resection is the first-line therapeutic option, and it is accepted as a safe treatment with a proven impact on prognosis, with a low operative mortality as the result of advances in surgical techniques and perioperative management. Nevertheless, surgical resection is applicable in only about 20% to 30% of patients with HCC, since most have poor hepatic reserve function caused by underlying chronic liver disease and multi focal hepatic distributions of HCC. Although hepatic resection is one of the curative treatments for hepatocellular carcinoma, the recurrence rate of HCC even after curative resection is quite high, estimated to be approximately 50 % during the first 3 years and more than 70 % during the first 5 years after curative resection, and so the postoperative long term results remain unsatisfactory. In this scenario the role of liver transplantation has been, in the last years, predominant, due to the ability of transplant to reduce disease recurrence, because of the treatment of liver cirrhosis associate to HCC which represent the most important driver to recurrence. Otherwise, the scarcity of organ source has been a boost to the spread of liver resection, not only confined in the boundary taken into account in the BCLC algorithm (guidelines endorsed by EASL and AASLD), but even in patients considered not suitable for curative treatment as well as liver resection. Although surgical treatment has been adopted in the last years in more patients outside the Guidelines with satisfactory results in terms of mortality, morbidity and Short term oncological outcomes, the limits of this approach remain the long term disease free survival.
Risk factor for recurrence has been yet identified in the last years as hcc dimension, grading, microvascular invasion and satellitosis. The evidence that these two prognostic factors could negatively impact on the long term prognosis enhancing the risk of recurrence, has led many Author to propose anatomical resection (segmental resection) as the ideal surgical treatment to reduce these risks in HCC patients. Otherwise, literature results are in conflict regarding the real benefit of this approach. In fact in many patients with HCC and underlying cirrhosis the anatomical approach is not feasible due to the risk of postoperative liver failure. So a parenchyma-sparing technique has been developed and compared to anatomical resection in terms of oncological outcomes. At the present, all these risk factors are not predictable, and the staging systems are based only on crude radiological features as the number and the size of the nodules. In the recent years, several authors proposed new approaches to increase our ability to extract data from the radiological imaging: by the analysis of the measurements and numbers obtained during the radiological acquisition (by CT or MRI scans), thousands of other information are obtainable, overcoming the ability of human eyes. Those techniques go under the names of "Radiomics", which is a very promising branch when merged with the novel machine learning algorithms (e.g. Deep Learning, Neural Networks, etc). Moreover, nowadays, novel data can be obtained also by simple intraoperative photo obtained during the surgical procedure, for example of the liver cut surface: by the "computer-vision analysis", another powerful machine-learning algorithm, other data can be produced to predict short and long term outcomes. Those potentialities rely on the modern field of "artificial intelligence", where a machine is trained to recognize different recurrent patterns to create prediction models with a very powerful accuracy. On these data is based the proposal to create the present multicentric study with the aim to develop a prediction model for post-operative complications and HCC recurrence, based on the analysis of CT-radiomics features, liver cut surface photos and machine learning analysis.
Recruitment & Eligibility
- Status
- ACTIVE_NOT_RECRUITING
- Sex
- All
- Target Recruitment
- 535
- Age >= 18 years old.
- Hepatocarcinoma diagnosis confirmed at histological specimen
- Being at the first HCC diagnosis or with a recurrence/persistence disease evaluated and treated for the first time with surgery at the participating center.
- Available contrast-enhanced CT Scan obtained no more than 1 month prior to surgery.
- Surgery as a downstaging therapy for transplant
- Patients treated with surgery in case of not-curative intent (palliation, best supportive care, etc).
- Histopathological specimen of combined liver primary neoplasms (e.g. 'epatocolangiocarcinoma').
Study & Design
- Study Type
- OBSERVATIONAL
- Study Design
- Not specified
- Arm && Interventions
Group Intervention Description HCC Patients Submitted Surgery CT Scan Radiomics Features Extraction For the retrospective data collection, the planned number of subjects that will be enrolled will be almost 150/year. Considering the study-period (2010-2020), it is estimated a total of 1500 patients. For the prospective observational data collection, the estimation of patients'enrolment is based on the number of patients treated per year in the participating centers (globally 150/year). Since the observational nature, patients will be evaluated for their enrolment consecutively. The prospective data collection will be prosecuted for 2 years, leading to a prospective cohort of 300 patients. The whole study is planned to be ended in December 2023. Inclusion and exclusion criteria will be the same among the retrospective and the prospective parts of the study.
- Primary Outcome Measures
Name Time Method Evaluate the association between different radiomics and computer-vision features, and the survival after surgery 1 Year and 4 Months Evaluate the association between different radiomics (obtained by the pre-operative CT scans) and computer-vision (obtained by the photos of the remnant liver after surgery) features, and the survival after surgery (in terms of Recurrence-Free Survival and Overall Survival). Thus, to develop a prediction algorithm based on that features.
- Secondary Outcome Measures
Name Time Method Evaluate the association between different radiomics and computer-vision features, and the short-terms results after surgery 1 Year and 4 Months Evaluate the association between different radiomics (obtained by the pre-operative CT scans) and computer-vision (obtained by the photos of the remnant liver after surgery) features, and the short-terms results after surgery (complications, biliary fistula, severity, 90-day mortality, presence of microvascular invasion and satellitosis).
Trial Locations
- Locations (1)
IRCCS Istituto Clinico Humanitas
🇮🇹Rozzano, Lombardy, Italy