Quest for the best endoscopic imaging modality for computer-assisted colonic polyp staging.

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Advanced endoscopic imaging technologies like high-definition endoscopy, high magnification endoscopy, and image enhancement (e.g., chromoendoscopy, narrow-band imaging [NBI], i-Scan) aid in colonic polyp detection and classification.
• Previous studies highlight the utility of these technologies, but a comprehensive analysis of their impact on automated polyp classification is lacking.

Purpose of the Study:

• To evaluate which endoscopic imaging modalities are most effective for computer-assisted staging of colonic polyps.
• To determine the influence of different imaging techniques on the performance of computer-assisted diagnosis systems for colonic polyp classification.

Main Methods:

• Utilized twelve state-of-the-art feature extraction methods for colonic polyp classification on five endoscopic image databases.
• Employed a controlled experimental setup to mitigate bias from varying image database sizes.
• Image databases were acquired using high-definition endoscopy with and without i-Scan, and high-magnification endoscopy with narrow-band imaging (NBI) or chromoendoscopy.

Main Results:

• Classification accuracy is feature-dependent, with varying performance across imaging modalities.
• High-definition endoscopy achieved up to 88.9% classification rate without chromoendoscopy and 79.2% with chromoendoscopy.
• High-magnification endoscopy with NBI yielded the highest rates, reaching up to 97.4%, while chromoendoscopy resulted in up to 81.4%.
• Image recording conditions significantly impacted automated diagnosis system performance, often more than the imaging modality itself.

Conclusions:

• Chromoendoscopy negatively affects classification results compared to other methods.
• Narrow-band imaging (NBI) is more suitable for computer-assisted polyp staging than chromoendoscopy.
• High-definition and high-magnification endoscopy demonstrate comparable suitability for automated colonic polyp staging.