Advanced endoscopic imaging for prediction of neoplasia in colorectal polyps

Abstract

Aims: I hypothesize that characterizing polyps as neoplastic or non-neoplastic is possible using current and emerging technologies, including image-enhanced endoscopy, artificial intelligence (AI), computer-aided diagnosis, and photonics microscopy. I further hypothesize that these methods have sufficient diagnostic accuracy to support resect-and-discard strategy for managing small colorectal polyps. Methods: The research scope was defined by reviewing the current role of colonoscopy in preventing colorectal cancer, particularly focusing on the polypectomy of small adenomas. A meta-analysis of the current diagnostic technologies, such as i-scan and AI-based computer-aided diagnosis, was conducted. A systematic review of emerging technologies like photonics endomicroscopy was also performed. Prototype microscopy platforms for optical coherence tomography (OCT) and multiphoton microscopy (MPT) were developed, with pre-clinical results presented. Results: Current image enhancement technologies enable endoscopists to achieve diagnostic accuracy that meets ASGE standards for the optical diagnosis of colorectal polyps. Specifically, the Pentax i-scan technology demonstrated a sensitivity of 0.92, specificity of 0.90, and negative predictive value of 89.9% for detecting adenomatous pathology in small polyps. Emerging technologies such as OCT, MPT, and hyperspectral imaging show promise, as does AI and computer-aided diagnosis, which achieved a sensitivity of 0.94, specificity of 0.82, and accuracy of 0.90 in meta-analysis. Pre-clinical trials of an integrated OCT/MPT endomicroscope did not meet expectations, requiring separate OCT and MPT subsystems. While optical signals were generated from biological tissue and polyp features were observed in rat colon, the results were not clinically relevant. Conclusions: Colonoscopy with polypectomy of detected neoplastic polyps reduces colorectal cancer incidence. Resect-and-discard or diagnose-and-leave strategies can lower healthcare costs and procedural risks. Image-enhanced endoscopy performed by trained operators, provides adequate diagnostic accuracy for these strategies. AI-based diagnosis shows comparable accuracy. Future technologies like OCT and MPT need further development, as current prototypes, while technically feasible, are not clinically useful or endoscopically deployable.