Deep learning with refined single candidate optimizer for early polyp detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Colorectal cancer (CRC) remains a leading cause of cancer death globally.
• Early detection of precancerous polyps via colonoscopy is crucial for reducing CRC burden.
• Automating polyp detection can enhance the efficiency and accuracy of colonoscopy screening.

Purpose of the Study:

• To develop and evaluate a novel deep learning-based approach for automated polyp detection in colonoscopy images.
• To improve the accuracy of polyp detection by refining feature extraction and classification stages.
• To provide an effective tool for early polyp detection to aid in timely CRC diagnosis.

Main Methods:

• Utilized the CaffeNet architecture for feature extraction and a Support Vector Machine (SVM) for classification.
• Introduced the Refined Single Candidate Optimizer (RSCO) to enhance optimization, balancing exploration and exploitation.
• Evaluated the model on the SUN Colonoscopy Video Database, comparing it against conventional methods.

Main Results:

• The proposed deep learning model demonstrated superior performance compared to conventional methods.
• Achieved significant improvements in precision, recall, and accuracy for polyp detection.
• The RSCO effectively refined feature extraction and classification, enhancing overall model performance.

Conclusions:

• The developed automated polyp detection system shows high efficacy for early detection during routine colonoscopies.
• This approach has the potential to significantly assist clinicians in the timely diagnosis of colorectal cancer.
• The novel optimization technique offers a promising direction for advancing AI in medical diagnostics.