Endoscopic Ultrasound (EUS) Artificial Intelligence Model for Normal Mediastinal and Abdominal Strictures Assessment
Study Details
Study Description
Brief Summary
Therefore, a high number of procedures is necessary to achieve EUS competency, but interobserver agreement still varies widely. Artificial intelligence (AI) aided recognition of anatomical structures may improve the training process and inter-observer agreement. Robles-Medranda et al. developed an AI model that recognizes normal anatomical structures during linear and radial EUS evaluations. We pursue to design an external validation of our developed AI model, considering an endoscopist expert as the gold standard.
| Condition or Disease | Intervention/Treatment | Phase |
|---|---|---|
|
Detailed Description
Endoscopic ultrasound (EUS) is a high-skilled procedure with a limited number of facilities available for training. Therefore, a high number of procedures is necessary to achieve competency. However, the agreement between observers varies widely. Artificial intelligence (AI) aided recognition and characterization of anatomical structures may improve the training process while improving the agreement between observers. However, developed EUS-AI models have been explicitly trained or only with disease samples or for detecting abdominal anatomical features.
In other fields as Radiation Oncology, developed AI models have been widely used. They must recognize in unison healthy and disease strictures throughout any part of the human body during the contouring. It avoids unnecessary irradiation of normal tissue. EUS-AI models not trained with healthy samples can cause an increase in false-positive cases during real-life practice. It implies potential overdiagnosis of abnormal/disease strictures. EUS-AI models not trained with samples outside
Using an automated machine learning software, Robles-Medranda et al. have previously developed a convolutional neuronal networks (CNN) AI model that recognizes the anatomical structures during linear and radial EUS evaluations (AI Works, MD Consulting group, Ecuador). To the best of our knowledge, this EUS-AI model is the first trained with EUS videos from patients without pathologies and, thus, with normal mediastinal and abdominal organ/anatomic strictures. In this second stage, we pursue to design an external validation of our developed AI model, considering an endoscopist expert as the gold standard.
Study Design
Arms and Interventions
| Arm | Intervention/Treatment |
|---|---|
| Patients with normal mediastinal and abdominal organ/anatomic strictures Adult patients with normal mediastinal and abdominal organ/anatomic strictures after imaging test and EUS assessment due to chronic dyspepsia. |
Diagnostic Test: Identification or discharge visualization of mediastinal and abdominal organ/anatomic strictures through Endoscopic ultrasound (EUS) videos by an expert endoscopist
An expert endoscopist will select a dataset of mediastinal and abdominal EUS videos (one per patient). An expert endoscopist will identify or discharge visualization of the following organs correctly: aorta, vertebral spine, aortic arch, trachea, AP window, left kidney, liver, spleen, pancreas body, pancreas tail, coeliac trunk, splenic artery, splenic vein, inferior vena cava, adrenal gland, right kidney, gallbladder, common bile duct, ampulla of Vater, portal vein.
Diagnostic Test: Recognition of mediastinal and abdominal organ/anatomic strictures through Endoscopic ultrasound (EUS) videos using artificial intelligence (AI)
Using the same previous dataset of mediastinal and abdominal EUS videos, the EUS-AI model will recognize the following organs: aorta, vertebral spine, aortic arch, trachea, AP window, left kidney, liver, spleen, pancreas body, pancreas tail, coeliac trunk, splenic artery, splenic vein, inferior vena cava, adrenal gland, right kidney, gallbladder, common bile duct, ampulla of Vater, portal vein. Considering each patient (and not data frame videos) as the study unit, a contingency table per each mediastinal and abdominal organ/anatomic stricture will be designed.
|
Outcome Measures
Primary Outcome Measures
- Overall accuracy of Endoscopic ultrasound (EUS) artificial intelligence (AI) model for identifying normal mediastinal and abdominal organ/anatomic strictures [Three months]
Overall accuracy features will be calculated: sensitivity, specificity, positive predictive value, negative predictive value, diagnostic accuracy, and observed agreement. In addition, there will be defined the following probabilities: True-positive (TP): mediastinal/abdominal organ/anatomic stricture recognized by the EUS-AI model. The expert endoscopist previously correctly identified it. False-positive (FP): mediastinal/abdominal organ/anatomic stricture recognized by the EUS-AI model. The expert endoscopist previously correctly discharged its visualization. False-negative (FN): mediastinal/abdominal organ/anatomic stricture not recognized by the EUS-AI model. The expert endoscopist previously correctly identified it. True-negative (TN): mediastinal/abdominal organ/anatomic stricture not recognized by the EUS-AI model. The expert endoscopist previously correctly discharged its visualization.
Eligibility Criteria
Criteria
Inclusion Criteria:
-
Patients with no history of the thorax and abdominal abnormalities confirmed through an imaging test requested for healthcare purposes in the last twelve months (e.g., thorax X-ray and abdominal ultrasound or thorax and abdominal CT)
-
Patients who undergo EUS assessment due to chronic dyspepsia.
Exclusion Criteria:
-
Morphological alteration on at least one mediastinal and abdominal organ/anatomic strictures documented through any imaging test or EUS.
-
Uncontrolled coagulopathy, kidney/liver failure, or any comorbidity with a meaningful impact on cardiac risk assessment (NHYA III/IV);
-
Refuse to participate in the study or to sign corresponding informed consent.
Contacts and Locations
Locations
| Site | City | State | Country | Postal Code | |
|---|---|---|---|---|---|
| 1 | Ecuadorian Institute of Digestive Diseases | Guayaquil | Guayas | Ecuador | 090505 |
Sponsors and Collaborators
- Instituto Ecuatoriano de Enfermedades Digestivas
Investigators
- Principal Investigator: Carlos Robles-Medranda, Ecuadorian Institute of Digestive Diseases
Study Documents (Full-Text)
None provided.More Information
Publications
- Cazacu IM, Udristoiu A, Gruionu LG, Iacob A, Gruionu G, Saftoiu A. Artificial intelligence in pancreatic cancer: Toward precision diagnosis. Endosc Ultrasound. 2019 Nov-Dec;8(6):357-359. doi: 10.4103/eus.eus_76_19.
- Kuwahara T, Hara K, Mizuno N, Haba S, Okuno N, Koda H, Miyano A, Fumihara D. Current status of artificial intelligence analysis for endoscopic ultrasonography. Dig Endosc. 2021 Jan;33(2):298-305. doi: 10.1111/den.13880. Epub 2020 Dec 5. Review.
- Marya NB, Powers PD, Chari ST, Gleeson FC, Leggett CL, Abu Dayyeh BK, Chandrasekhara V, Iyer PG, Majumder S, Pearson RK, Petersen BT, Rajan E, Sawas T, Storm AC, Vege SS, Chen S, Long Z, Hough DM, Mara K, Levy MJ. Utilisation of artificial intelligence for the development of an EUS-convolutional neural network model trained to enhance the diagnosis of autoimmune pancreatitis. Gut. 2021 Jul;70(7):1335-1344. doi: 10.1136/gutjnl-2020-322821. Epub 2020 Oct 7.
- Minoda Y, Ihara E, Komori K, Ogino H, Otsuka Y, Chinen T, Tsuda Y, Ando K, Yamamoto H, Ogawa Y. Efficacy of endoscopic ultrasound with artificial intelligence for the diagnosis of gastrointestinal stromal tumors. J Gastroenterol. 2020 Dec;55(12):1119-1126. doi: 10.1007/s00535-020-01725-4. Epub 2020 Sep 11.
- Robles-Medranda C, Oleas R, Del Valle R, Mendez JC, Alcívar-Vásquez JM, Puga-Tejada M, Lukashok H. Intelligence for real-time anatomical recognition during endoscopic ultrasound evaluation: a pilot study. Gastrointestinal Endoscopy. 2021; 93(6), AB221. https://doi.org/10.1016/J.GIE.2021.03.491
- Tonozuka R, Mukai S, Itoi T. The Role of Artificial Intelligence in Endoscopic Ultrasound for Pancreatic Disorders. Diagnostics (Basel). 2020 Dec 24;11(1). pii: E18. doi: 10.3390/diagnostics11010018. Review.
- Udriștoiu AL, Cazacu IM, Gruionu LG, Gruionu G, Iacob AV, Burtea DE, Ungureanu BS, Costache MI, Constantin A, Popescu CF, Udriștoiu Ș, Săftoiu A. Real-time computer-aided diagnosis of focal pancreatic masses from endoscopic ultrasound imaging based on a hybrid convolutional and long short-term memory neural network model. PLoS One. 2021 Jun 28;16(6):e0251701. doi: 10.1371/journal.pone.0251701. eCollection 2021.
- Yao L, Zhang J, Liu J, Zhu L, Ding X, Chen D, Wu H, Lu Z, Zhou W, Zhang L, Xu B, Hu S, Zheng B, Yang Y, Yu H. A deep learning-based system for bile duct annotation and station recognition in linear endoscopic ultrasound. EBioMedicine. 2021 Mar;65:103238. doi: 10.1016/j.ebiom.2021.103238. Epub 2021 Feb 24. Erratum in: EBioMedicine. 2021 Nov;73:103650.
- Zhang J, Zhu L, Yao L, Ding X, Chen D, Wu H, Lu Z, Zhou W, Zhang L, An P, Xu B, Tan W, Hu S, Cheng F, Yu H. Deep learning-based pancreas segmentation and station recognition system in EUS: development and validation of a useful training tool (with video). Gastrointest Endosc. 2020 Oct;92(4):874-885.e3. doi: 10.1016/j.gie.2020.04.071. Epub 2020 May 6. Erratum in: Gastrointest Endosc. 2021 Mar;93(3):781.
- IECED-26112021