: Systems using this technology, such as those discussed in research from the National Cancer Center in Tokyo , are designed to alert endoscopists to the presence of colorectal polyps or adenomas in real-time.
In the academic community, such video files are often part of larger datasets, such as the (a public cohort of multimodal colonoscopy videos), which are used to train machine learning classifiers. These datasets require high-quality, focused images where polyps occupy at least 10% of the frame and are free from artifacts like blood or blurriness. EVIS-436.mp4
: The video data (like an .mp4 file) is typically fed into a computer with high-end GPUs. The software analyzes every frame—often at 30 frames per second—to flag potential lesions with a visual box or an audio alert. : Systems using this technology, such as those