Gastrointestinal image classification with GIDNet CNN model and non-linear Tansh activation function.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Gastroenterology

Background:

• Gastrointestinal diseases present substantial health risks, necessitating improved diagnostic tools.
• Current methods for GI disease detection can be enhanced through advanced image processing and AI.
• A gap exists in specialized AI models and activation functions for GI disease classification and explainability.

Purpose of the Study:

• To develop a highly accurate AI model for classifying gastrointestinal diseases.
• To introduce and evaluate a novel convolutional neural network (CNN) model, GIDNet.
• To propose and assess a new activation function, Tansh, for enhanced classification performance.

Main Methods:

• Implementation of the novel Tansh activation function within the GIDNet CNN architecture.
• Training and evaluation of GIDNet using the Kvasir dataset for GI disease classification.
• Comparative analysis of the Tansh activation function against existing functions for accuracy and explainability.

Main Results:

• The GIDNet model with the Tansh activation function achieved a classification accuracy of 98.75% on the Kvasir dataset.
• The proposed model demonstrated superior performance compared to existing state-of-the-art methods.
• Enhanced model explainability was observed, addressing a key research gap.

Conclusions:

• The novel GIDNet model and Tansh activation function significantly advance GI disease classification accuracy and interpretability.
• This AI-driven approach offers potential for automated diagnosis and treatment planning in gastroenterology.
• The findings pave the way for more reliable and explainable AI solutions in healthcare.