AI-Assisted Diagnosis for Stomach Problems

(AI-OD Trial)

Age: 18+

Sex: Any

Trial Phase: Academic

Sponsor: Daniel Von Renteln

Disqualifiers: Inflammatory bowel disease, Active colitis, Hereditary CRC syndrome, Coagulopathy, ASA status >3

No Placebo Group All trial participants will receive the active study treatment (no placebo)

Compensation: Varies

Number of Visits: 2

Duration: Approximately 17 weeks

204 Participants Needed

Trial Summary

Will I have to stop taking my current medications?

The trial information does not specify whether you need to stop taking your current medications. Please consult with the trial coordinators for more details.

What data supports the effectiveness of the AI-assisted diagnosis treatment for stomach problems?

Research shows that AI systems can accurately diagnose various stomach issues, such as dyspepsia and gastric cancer, with high accuracy rates. For example, an AI system achieved a diagnostic accuracy of 87.7% for dyspepsia and demonstrated reliable performance in differentiating gastric cancer types, suggesting its potential effectiveness in assisting with stomach problem diagnoses. ^{1 2 3 4 5}

Is AI-assisted diagnosis for stomach problems safe for humans?

Current research on AI-assisted diagnosis for stomach issues, like dyspepsia and gastric lesions, shows promising results in terms of accuracy and effectiveness, but there is no specific mention of safety concerns in humans. The studies focus on diagnostic accuracy and improving treatment strategies, suggesting that the technology is generally considered safe for use in clinical settings. ^{2 5 6 7 8}

How does the AI-assisted diagnosis treatment for stomach problems differ from other treatments?

This AI-assisted diagnosis treatment is unique because it uses artificial intelligence to automatically detect and classify stomach lesions from endoscopy images, improving diagnostic accuracy and reducing variability between different doctors' interpretations. ^{1 2 6 8 9}

What is the purpose of this trial?

This is a prospective study that is the first to implement resect and discard and diagnose and leave strategies in real-time practice using stringent documentation and adjudication by 2 expert endoscopists as the gold standard. The primary aim of this study is to show the accuracy of intracolonoscopy AI-assisted optical diagnosis (CADx; autonomous or with human input) when the AI-assisted optical diagnosis made by the expert endoscopists is used as the reference standard. The specific aims are: 1. To evaluate the accuracy of intracolonoscopy AI-assisted optical polyp diagnosis (autonomous or with human input) by comparing it to the obtained optical histology diagnoses provided by two independent expert endoscopists as the reference standard. 2. To evaluate the agreement between the intracolonoscopy AI-assisted optical polyp diagnosis (autonomous or with human input) and the AI-assisted optical diagnosis performed by two independent expert endoscopists. 3. To determine whether AI-assisted optical polyp diagnosis for diminutive (1-5 mm) polyps can be implemented in routine clinical practice by demonstrating that at least 70% of the approached patients are interested in undergoing AI-assisted optical diagnosis (autonomous or with human input). 4. To evaluate the cost savings resulting from replacing pathology with AI-assisted optical diagnosis.

Research Team

Daniel von Renteln, MD

Principal Investigator

Centre hospitalier de l'Université de Montréal (CHUM)

Eligibility Criteria

This trial is for people aged 45-80 who are having an outpatient colonoscopy at the Centre Hospitalier de l'Université de Montréal. They must understand and agree to the study by signing a consent form. It's not for those with severe health risks (ASA status >3), inflammatory bowel diseases, active colitis, blood clotting disorders, or inherited colorectal cancer syndromes.

Inclusion Criteria

You are between 45-80 years old.

You have provided a signed consent document.

You are in the process of having a colonoscopy procedure at CHUM.

Exclusion Criteria

Inflammatory Bowel Disease

American Society of Anesthesiologists (ASA) status >3

Active colitis

Timeline

Screening

Participants are screened for eligibility to participate in the trial

2-4 weeks

1 visit (in-person)

Treatment

Participants undergo standard colonoscopy procedures with AI-assisted optical diagnosis for diminutive colorectal polyps

Approximately 17 weeks

1 visit (in-person) per patient

Follow-up

Participants are monitored for safety and effectiveness after treatment

4 weeks

Treatment Details

Interventions

• Artificial intelligence-assisted classification (CADx)

Trial Overview The study tests if artificial intelligence can help doctors decide how often patients need colonoscopies based on real-time analysis during the procedure. This AI-assisted method will be compared to decisions made by two expert endoscopists following established guidelines.

Participant Groups

Group I: Autonomous AI-assisted classification Experimental Treatment 1 Intervention

AI-assisted classification for diminutive polyps during a colonoscopy procedure using the CAD-eye detection and classification system, with no input from the endoscopist, for patients who agree to undergo optical diagnosis of diminutive colorectal polyps.

Group II: AI-assisted classification with endoscopist's input Experimental Treatment 1 Intervention

AI-assisted classification for diminutive polyps during a colonoscopy procedure using the CAD-eye detection and classification system, with input from the endoscopist in the case of serrated polyps, for patients who agree to undergo optical diagnosis of diminutive colorectal polyps.

Who Is Running the Clinical Trial?

Daniel Von Renteln

Lead Sponsor

Trials

1

Recruited

200+

Centre hospitalier de l'Université de Montréal (CHUM)

Lead Sponsor

Trials

389

Recruited

143,000+

Findings from Research

The computer-aided diagnosis system demonstrated a high accuracy rate, correctly diagnosing 78.8% of acute abdominal pain cases and 70% of dyspepsia cases in a study of 205 patients from Sherbrooke, Quebec.

The system was particularly effective for diagnosing appendicitis (97% accuracy) and cholecystitis (91% accuracy), but it struggled with pancreatitis, achieving only a 25% detection rate.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Computer-aided diagnosis of gastroenterologic diseases in Sherbrooke: preliminary report. Horrocks, JC., Devroede, G., de Dombal, FT.[2007]

In a study of 212 patients with dyspepsia, the computer-aided diagnosis achieved a high diagnostic accuracy of 92.6% after full investigation, indicating its effectiveness in identifying conditions before surgery.

The computer system demonstrated an overall diagnostic accuracy of 87.7% based on initial interviews, and it was particularly successful in distinguishing organic from functional dyspepsia, correctly categorizing 22 out of 23 patients with organic disease.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Computer-aided diagnosis of "dyspepsia". Horrocks, JC., de Dombal, FT.[2022]

The proposed artificial intelligence-based decision support system can effectively classify five subtypes of gastric cancer, achieving a class-average sensitivity of over 0.85, which enhances the accuracy of treatment planning.

AI-assisted pathologists showed significantly improved diagnostic sensitivity and reduced screening time compared to traditional human pathologists, indicating the system's potential to improve clinical outcomes in gastric cancer management.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Multi-Scale Hybrid Vision Transformer for Learning Gastric Histology: AI-Based Decision Support System for Gastric Cancer Treatment. Oh, Y., Bae, GE., Kim, KH., et al.[2023]

AI has the potential to significantly enhance the diagnosis and treatment of gastrointestinal conditions by improving image analysis and integrating complex data, which can lead to more accurate assessments and personalized therapies.

In managing diseases like inflammatory bowel disease and peptic ulcer bleeding, AI can predict disease outcomes, potentially increasing treatment efficacy, but there are important considerations regarding the responsibility and liability of AI in medical practice.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Artificial intelligence in gastroenterology: where are we heading? Sung, JJ., Poon, NC.[2021]

The AI diagnostic tool 'Tango' outperformed specialist endoscopists in sensitivity for detecting early gastric cancers (EGCs), achieving 84.7% sensitivity compared to 65.8% for specialists, indicating its potential to enhance diagnostic accuracy.

Tango also showed superior sensitivity and accuracy compared to nonspecialist endoscopists, suggesting that this AI tool could significantly reduce misdiagnosis rates in early gastric cancer detection.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Performance of an artificial intelligence-based diagnostic support tool for early gastric cancers: Retrospective study. Ishioka, M., Osawa, H., Hirasawa, T., et al.[2023]

A novel AI system called MMP-AI was developed to automatically detect and classify protruded gastric lesions, achieving high diagnostic accuracy with AUC scores of 0.896 for GISTs, 0.890 for GILs, and 0.999 for GEPs based on data from 1,366 patients.

MMP-AI outperformed experienced endoscopists in recognizing submucosal tumors, indicating its potential to improve diagnostic accuracy and reduce variability in routine clinical practice.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

A Multimodal Multipath Artificial Intelligence System for Diagnosing Gastric Protruded Lesions on Endoscopy and Endoscopic Ultrasonography Images. Zhu, C., Hua, Y., Zhang, M., et al.[2023]

Computer-assisted systems in esogastroduodenoscopy show promising results for quality monitoring and accurately detecting gastric lesions, often matching or surpassing the performance of experienced endoscopists.

These AI systems can help determine the best treatment strategies for gastric cancer, potentially reducing unnecessary surgeries and improving patient outcomes, while ongoing clinical trials aim to enhance their real-world application.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

The role of computer-assisted systems for upper-endoscopy quality monitoring and assessment of gastric lesions. Lazăr, DC., Avram, MF., Faur, AC., et al.[2023]

In a study of 165 patients at Airedale District General Hospital, a computer-aided diagnosis system accurately predicted 83% of positive lesions found during endoscopy, including nearly all cases of gastric cancer.

The findings suggest that the computer-aided system can effectively transfer between locations and may help identify 'high-risk' patients who require more thorough investigation.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Transfer of computer-aided diagnosis of dyspepsia from one geographical area to another. Horrocks, JC., Lambert, DE., McAdam, WA., et al.[2019]

An integrated and automated system has been developed to collect and analyze clinical and instrumental data from dyspeptic patients, which may enhance the accuracy of diagnosing dyspepsia.

The system not only focuses on dyspepsia but also shows potential for application in diagnosing absorption and hepatic problems, indicating its versatility in gastrointestinal diagnostics.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Breath Multi Analysis: a database to collect data on gastric related non invasive analysis. Giacomini, M., Peragallo, I., Bertone, S., et al.[2004]

References

1. Canada pubmed.ncbi.nlm.nih.gov

Computer-aided diagnosis of gastroenterologic diseases in Sherbrooke: preliminary report. [2007]

2. United States pubmed.ncbi.nlm.nih.gov

Computer-aided diagnosis of "dyspepsia". [2022]

3. United States pubmed.ncbi.nlm.nih.gov

Multi-Scale Hybrid Vision Transformer for Learning Gastric Histology: AI-Based Decision Support System for Gastric Cancer Treatment. [2023]

4. China pubmed.ncbi.nlm.nih.gov

Artificial intelligence in gastroenterology: where are we heading? [2021]

5. Australia pubmed.ncbi.nlm.nih.gov

Performance of an artificial intelligence-based diagnostic support tool for early gastric cancers: Retrospective study. [2023]

6. United States pubmed.ncbi.nlm.nih.gov

A Multimodal Multipath Artificial Intelligence System for Diagnosing Gastric Protruded Lesions on Endoscopy and Endoscopic Ultrasonography Images. [2023]

7. England pubmed.ncbi.nlm.nih.gov

The role of computer-assisted systems for upper-endoscopy quality monitoring and assessment of gastric lesions. [2023]

8. England pubmed.ncbi.nlm.nih.gov

Transfer of computer-aided diagnosis of dyspepsia from one geographical area to another. [2019]

9. Netherlands pubmed.ncbi.nlm.nih.gov

Breath Multi Analysis: a database to collect data on gastric related non invasive analysis. [2004]