Grant: CIHR Fellowship
Direct Payment
$75,000.00
Improving prediction of colorectal cancer prior to endoscopic resection of large colorectal polyps using Artificial Intelligence
FY 2024-25
Summary
Large flat tumors in the colon, known as laterally spreading tumors (LSTs), are precursors to colorectal cancer (CRC). Accurate assessment of these tumors is crucial because they often contain areas of early cancer. LSTs are typically removed through two techniques: piecemeal resection, which involves cutting the tumor into multiple pieces but may require additional surgery if cancer is found, and en-bloc resection, which removes the tumor in one piece and can be curative for early cancers. However, en-bloc resection is technically challenging and carries a higher risk of complications, such as perforation.
Choosing the right technique depends on accurately predicting the risk of cancer within these tumors. Currently, endoscopists rely on visual inspection to make this decision. Unfortunately, this method is not very accurate and often results in unnecessary or inappropriate treatments. Artificial intelligence (AI) could offer a more accurate way to assess these tumors.
Our study aims to develop an AI model to predict the presence of early-stage CRC in LSTs that can be cured with en-bloc resection. Using data from a large French database of patients who underwent endoscopic removal of LSTs, we will train and test this AI model. The model will be evaluated against the current standard optical diagnosis to see if it can better identify cases where en-bloc resection is needed.
The primary outcome is the model's ability to correctly identify early cancers that can be treated endoscopically. We will also look at other factors like accuracy, specificity, and predictive values to assess the model's overall performance. Our goal is to develop a tool that can help endoscopists choose the best treatment approach, reducing unnecessary procedures and improving patient outcomes.
Department
Program
CIHR Fellowship
Recipient
Location
France
Type
Grant
Source
CIHR Funding Decisions Database
As published
Id
513282
Name
Djinbachian Roupen
Country
France
Orgtype
Research Institution
Instname2
Nutrition, Metabolism and Diabetes
Theme2
Clinical
Prcname2
Health Research Training B - HP
Projecttitle
Improving prediction of colorectal cancer prior to endoscopic resection of large colorectal polyps using Artificial Intelligence
Programtype2
Fellowship Programs
Orgnameinp2
Direct Payment
Cihramount2
$75,000
Cihrequipment2
$0
Programname2
CIHR Fellowship
Keyworddelim
Artificial Intelligence; Colonoscopy; Colorectal Cancer Treatment; Complex Polypectomy; Endoscopic Resection
Abstract
Large flat tumors in the colon, known as laterally spreading tumors (LSTs), are precursors to colorectal cancer (CRC). Accurate assessment of these tumors is crucial because they often contain areas of early cancer. LSTs are typically removed through two techniques: piecemeal resection, which involves cutting the tumor into multiple pieces but may require additional surgery if cancer is found, and en-bloc resection, which removes the tumor in one piece and can be curative for early cancers. However, en-bloc resection is technically challenging and carries a higher risk of complications, such as perforation.
Choosing the right technique depends on accurately predicting the risk of cancer within these tumors. Currently, endoscopists rely on visual inspection to make this decision. Unfortunately, this method is not very accurate and often results in unnecessary or inappropriate treatments. Artificial intelligence (AI) could offer a more accurate way to assess these tumors.
Our study aims to develop an AI model to predict the presence of early-stage CRC in LSTs that can be cured with en-bloc resection. Using data from a large French database of patients who underwent endoscopic removal of LSTs, we will train and test this AI model. The model will be evaluated against the current standard optical diagnosis to see if it can better identify cases where en-bloc resection is needed.
The primary outcome is the model's ability to correctly identify early cancers that can be treated endoscopically. We will also look at other factors like accuracy, specificity, and predictive values to assess the model's overall performance. Our goal is to develop a tool that can help endoscopists choose the best treatment approach, reducing unnecessary procedures and improving patient outcomes.
Competitiondate
202410
Approvedterm2
1 yr 0 mth
Pinamesdelim
Djinbachian, Roupen
Approvedterm
1 an 0 mois
Cihrcontribution2
~ $75,000~ $0~