A Multi-center Study on the Efficacy and Safety of AI-assisted Navigation System for Biliopancreatic EUS

Not Applicable

• Conditions: Pancreatic Cancer
• Interventions: Other: Artificial intelligence assistant system

• Registration Number: NCT04892329
• Lead Sponsor: Renmin Hospital of Wuhan University

Brief Summary

Pancreatic cancer is one of the most lethal cancers. Survival rates vary largely depending on the stage at which it is diagnosed. EUS is considered one of the most sensitive modalities for pancreatic cancer detection. To avoid a missed diagnosis of the pancreatic cancer, the continuity and integrity of EUS needs to be ensured as much as possible. The station approach in pancreatic EUS has been established as the standard scanning procedure. Complete anatomical scanning is helpful for the identification of standard stations, and its imaging findings can assist in the diagnosis of pancreatic lesions and guide patient treatment and prognosis. But EUS is highly operator-dependent and the learning curve is steep. In this study, we constructed a deep learning-based pancreatic scanning navigation system in EUS, which can assist in identifying important anatomical structures adjacent to the pancreas in real time. In order to improve the quality of EUS and reduce the missed diagnosis of pancreatic lesions.

Detailed Description

Pancreatic cancer is a malignant tumor of the digestive system with insidious onset, rapid progression and very poor prognosis. According to the latest cancer data in China in 2020 released by the International Agency for Research on Cancer (IARC) of the World Health Organization, there are about 120,000 new cases of pancreatic cancer in China, with a mortality rate close to 100%, which seriously endangers the national health. Early diagnosis of pancreatic cancer can be achieved by surgical resection with a 5-year survival rate of 58%, Once advanced pancreatic cancer develops, patient survival is 7.2%. As a rapidly developing deadly cancer, missed diagnosis of pancreatic cancer may have extremely serious consequences for patients.How to improve the diagnostic rate of early pancreatic cancer is an urgent problem to be solved.

EUS(endoscopic ultrasonography) is considered one of the most sensitive modalities for pancreatic cancer detection. It has a much higher diagnostic accuracy than MRI and CT for the diagnosis of pancreatic cancer, especially early pancreatic cancer \< 1 cm in diameter (EUS-FNA 95.6% vs CT 77.4%, MRI 76.2%). EUS is the modality of choice for the early diagnosis of pancreatic tumors. To avoid a missed diagnosis of the pancreatic cancer, the continuity and integrity of EUS needs to be ensured as much as possible. But EUS is highly operator-dependent and the learning curve is steep, and the quality of the examination is highly dependent on the operator's technique. Therefore, it is necessary to develop a system that can effectively assist the full scanning of EUS.

The station approach in pancreatic EUS has been established as the standard scanning procedure. The principle of completing the station approach is to find the anatomical landmarks of this station, Such as organs (kidney, spleen), blood vessels (such as splenic artery, splenic vein, portal vein), ducts (pancreatic duct, bile duct), etc.The scanning of these anatomical landmarks is the basis for an accurate assessment of the entire pancreas。 At the same time, the type of pancreatic lesions and the development of the course have abnormal imaging findings of different anatomical structure. For example, ultrasound images of pancreatic cancer will show vascular invasion, deformation of the biliopancreatic duct, and metastasis of adjacent organs. The guidelines clearly require that the choice of surgical approach for pancreatic cancer needs to be based on the degree of invasion of the cancer to adjacent important anatomical structures, to maximize the volume sparing of functional pancreatic parenchyma. Complete anatomical scanning can assist in the diagnosis of pancreatic lesions and guide patient treatment and prognosis.

In recent years, artificial intelligence (AI) has been successfully applied in multiple medical fields. At present, there have been studies of AI-based endoscopic ultrasonography for the identification of pancreatic lesions, However, there are no studies of AI-based navigation system for pancreatic endoscopic ultrasonography. Previously, we have successfully developed a standard station scanning navigation system for the pancreas and bile ducts. This system can improve the recognition accuracy of endoscopists for standard stations and enhance the cognitive ability of endoscopic ultrasonography images.

Based on the previous, we constructed a deep learning-based pancreatic scanning navigation system in EUS, which can assist in identifying important anatomical structures adjacent to the pancreas in real time. and verify its auxiliary performance for endoscopists in clinical practice. In order to improve the quality of EUS and reduce the missed diagnosis of pancreatic lesions.

Study Timeline

• Study First Submitted: 2021-05-10
• Study Start: 2021-05-12
• Study First Submitted QC: 2021-05-13
• Study First Posted: 2021-05-19
• Status Verified: 2022-06
• Last Update Submitted: 2022-06-29
• Last Update Posted: 2022-07-05
• Primary Completion: 2022-11-29
• Study Completion: 2022-11-30

Recruitment & Eligibility

• Status: UNKNOWN
• Sex: All
• Target Recruitment: 285

Inclusion Criteria

• All patients meeting the following criteria will be considered for participation in the study:

  1. Male or female aged 18 or above;
  2. EUS is needed to further clarify the characteristics of biliopancreatic diseases;
  3. Patients able to give informed consent were eligible to participate.
  4. Able and willing to comply with all study process.

Exclusion Criteria

• All patients meeting the following criteria will not be considered for participation in the study:

  1. Has participated in other clinical trials, signed informed consent and was in the follow-up period of other clinical trials.
  2. Has participated in clinical trials of the drug and is in the elution period of the experimental drug or control drug.
  3. Drug or alcohol abuse or psychological disorder in the last 5 years.
  4. Patients in pregnancy or lactation.
  5. A history of Upper Gastrointestinal surgery.
  6. Patients with anatomical abnormalities of the upper gastrointestinal tract due to advanced neoplasia
  7. Patients in whom the presence of clearly defined vital anatomical structures cannot be observed
  8. Researchers believe that the patient is not suitable to participate in the trial.

Study & Design

• Study Type: INTERVENTIONAL
• Study Design: PARALLEL

Arm && Interventions

Group	Intervention	Description
EUS AI navigation system augmentation	Artificial intelligence assistant system	The endoscopists in the experimental group will be assisted by EndoAngel, which can assist in identifying important anatomical structures adjacent to the pancreas in real time. The system is an non-invasive AI system .

Primary Outcome Measures

Name	Time	Method
Missed scanning rate of adjacent important anatomical structures in pancreatic endoscopic ultrasonography	Six month	It was calculated by dividing the number of important anatomy that is not scanned in the actual EUS pancreas by the number of EUS.

Secondary Outcome Measures

Name	Time	Method
Pancreatic lesions detection rate	Six month	It was calculated by dividing the total number of patients being detected pancreatic lesions by the number of EUS
Cholangiopancreatic duct lesions detection rate	Six month	It was calculated by dividing the total number of patients being detected cholangiopancreatic duct lesions by the number of EUS.
Detection rate of lesions in different pancreatic standard stations of endoscopic ultrasonography	Six month	It was calculated by dividing the number of patients with pancreatic lesions and Cholangiopancreatic duct lesions in the different standard stations by the number of EUS.
Mean scanning time in different pancreatic standard stations of endoscopic ultrasonography	Six month	It was calculated by dividing the total scanning time by the number of EUS.
The average number of scanning in the pancreatic standard station of endoscopic ultrasonography	Six month	It was calculated by dividing the total number of scanning in the pancreatic standard station by the number of EUS.
Missed scanning rate of adjacent vital anatomical structures in different pancreatic standard stations of endoscopic ultrasonography	Six month	It was calculated by dividing the number of important anatomy that is not scanned in the different standard stations by the number of EUS.

Trial Locations

Locations (2)

Union Hospital, Tongji Medical College, Huazhong University of Science and Technology; 🇨🇳 Wuhan, Hubei, China

Renmin Hospital of Wuhan University; 🇨🇳 Wuhan, China