Application of Structural Entropy and Spatial Filling Factor in Colonoscopy Image Classification

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Artificial Intelligence

Background:

• Colonoscopy is standard for colorectal polyp detection, relying on expert interpretation.
• Automatic polyp segmentation is crucial for computer-aided diagnosis (CAD) systems.
• Limited public polyp image datasets necessitate advanced detection and classification methods.

Purpose of the Study:

• To develop and validate a fuzzy inference method for polyp detection and classification.
• To enhance CAD systems using metaheuristic and deep learning approaches.
• To improve the accuracy of polyp identification in colonoscopy images.

Main Methods:

• Generation and validation of a fuzzy rule set using a statistical approach and histograms.
• Development of a method for selecting relevant antecedent variables based on histogram comparison.
• Assessment of including Rényi-entropy-based structural entropy and spatial filling factor as input variables.

Main Results:

• The fuzzy rule set generation and validation process was completed.
• A method for selecting relevant input variables was successfully presented.
• Inclusion of structural entropy from hue and saturation channels improved classification rates.

Conclusions:

• Fuzzy inference methods, combined with deep learning, can enhance polyp detection and classification.
• Structural entropy is a beneficial feature for improving classification accuracy in HSV images.
• The proposed methods contribute to the development of more effective computer-aided diagnosis systems for colorectal polyps.