Objective:

To classify the texture, colour, and vessel features of polyps according to the NICE classification using a deep neural network, enhancing diagnostic accuracy.

Key Findings:

• The NICE classification categorizes polyps into three types based on vessels, surface patterns, and colour, which aids in clinical decision-making.
• Deep learning techniques, particularly CNNs, have shown promise in improving polyp classification accuracy by automating feature extraction.
• The introduction of an 'indistinguishable' label addresses low-quality frames in classification, enhancing model reliability.

Interpretation:

The proposed framework aims to enhance diagnostic decision-making by providing clinicians with interpretable features of polyps, potentially leading to better patient outcomes.

Limitations:

• The study focuses primarily on adenomas and hyperplastic polyps, limiting the scope of classification and its applicability to other polyp types.
• Reliance on private datasets may hinder broader applicability and comparison with other studies, potentially affecting the generalizability of the findings.

Conclusion:

This study presents a novel approach to classifying polyp features using the NICE classification, potentially improving clinical decision-making in colonoscopy screenings.