RETRACTED: Gastrointestinal tract disorders classification using ensemble of InceptionNet and proposed GITNet based...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Gastroenterology

Background:

• Computer-aided classification of gastrointestinal tract (GIT) diseases is vital for early diagnosis.
• Manual screening of endoscopic frames is time-consuming and prone to errors.
• AI and deep learning offer potential for automated and accurate disease detection.

Purpose of the Study:

• To propose an automatic deep feature learning-based system for GIT disease classification.
• To enhance the accuracy and efficiency of diagnosing GIT diseases from endoscopic images.
• To reduce the missed rate of GIT diseases through automated analysis.

Main Methods:

• Adaptive Gamma Correction and Weighting Distribution (AGCWD) for image preprocessing.
• Deep feature extraction using InceptionNetV3 and GITNet models.
• Feature optimization with Ant Colony Optimization (ACO) and serial fusion.
• Classification using Support Vector Machine (SVM) variants (Cubic, Coarse Gaussian, Quadratic, Linear).

Main Results:

• The proposed model achieved 99.32% accuracy on the KVASIR dataset (8 classes).
• The model achieved 99.89% accuracy on the NERTHUS dataset (4 classes).
• Outperformed existing methods in GIT disease classification accuracy.

Conclusions:

• The developed deep learning system demonstrates high efficacy for automated GIT disease classification.
• The system offers a promising tool for improving diagnostic accuracy and efficiency in gastroenterology.
• Early and accurate detection of GIT diseases can significantly improve patient outcomes.