An Ensemble-Based Deep Convolutional Neural Network for Computer-Aided Polyps Identification From Colonoscopy.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Colorectal cancer (CRC) is a leading global cause of cancer mortality.
• Early detection and removal of precancerous polyps are crucial for reducing CRC-related deaths.
• Current polyp detection relies heavily on endoscopist expertise, leading to potential missed diagnoses.

Purpose of the Study:

• To develop a more effective and powerful machine learning classifier for polyp identification in colonoscopy videos.
• To improve upon existing polyp detection models that have not yet matched expert endoscopist performance.

Main Methods:

• Proposed a multiple classifier consultation strategy, termed an Ensemble classifier.
• Integrated ResNet's residual connections for feature extraction and Xception's depth-wise separable convolutions to handle occlusions.
• Applied the strategy to still frames from colonoscopy videos.

Main Results:

• The Ensemble classifier achieved a performance measure exceeding 95% across all algorithmic parameters.
• Outperformed existing state-of-the-art techniques in polyp identification.
• Demonstrated the efficacy of combining different classification models for enhanced diagnostic power.

Conclusions:

• The proposed Ensemble classifier offers a significant advancement in computational-guided colonoscopy screening.
• This method holds promise for developing clinically applicable tools for early polyp detection.
• The strategy may be adaptable to other image-based clinical diagnostic applications.