In vivo evaluation of complex polyps with endoscopic optical coherence tomography and deep learning during routine...

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Standard-of-care imaging for colorectal polyps lacks accuracy in assessing invasion depth.
• Current methods like white-light colonoscopy (WLC) and narrow-band imaging (NBI) are insufficient for complex polyps.

Purpose of the Study:

• To evaluate the feasibility and safety of a custom endoscopic optical coherence tomography (OCT) probe for colorectal polyp assessment.
• To assess the accuracy of OCT combined with a deep learning model in differentiating polyp subtypes and invasion depth.

Main Methods:

• A pilot study enrolled patients undergoing endoscopic treatment for large colorectal polyps.
• A side-viewing OCT catheter was used during colonoscopy to capture subsurface images before treatment.
• A deep learning model was trained to classify lesions based on OCT images, comparing results to histopathology.

Main Results:

• OCT imaging added an average of 3 minutes 40 seconds to procedure time, with no observed complications.
• OCT successfully visualized distinct subsurface structures correlating with histological findings (adenomas, serrated polyps, cancer).
• The deep learning model achieved an AUROC of 0.984 and Cohen's kappa of 0.845 for lesion classification.

Conclusions:

• Endoscopic OCT is a feasible and safe addition to routine colonoscopy for polyp assessment.
• The combination of OCT and deep learning provides high accuracy in differentiating benign from malignant colorectal lesions.
• This technology has the potential to increase clinician confidence and improve decision-making in managing colorectal polyps.