Recognition of esophagitis in endoscopic images using transfer learning

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Gastroenterology

Background:

• Esophagitis involves esophageal mucosa inflammation and damage, potentially leading to severe complications like stenosis or perforation.
• Endoscopic visualization is key for diagnosing esophagitis, but objective analysis can be challenging.

Purpose of the Study:

• To develop a deep neural network model using transfer learning for recognizing esophagitis in endoscopic images.
• To leverage artificial intelligence for improved diagnostic accuracy in esophageal conditions.

Main Methods:

• A binary deep learning classifier (DenseNet-201 architecture) was developed using a large dataset of endoscopic images.
• Transfer learning with ImageNet pre-trained weights was employed and fine-tuned for esophagitis detection.
• Model performance was rigorously evaluated using accuracy, sensitivity, specificity, and AUC metrics.

Main Results:

• The classifier achieved high performance metrics: 93.32% accuracy, 93.18% sensitivity, 93.46% specificity, and 0.96 AUC on the test set.
• Heatmaps were generated to visualize the model's decision-making process, enhancing explainability.
• The study demonstrated the effectiveness of deep convolutional neural networks for esophagitis recognition.

Conclusions:

• Transfer learning offers a promising strategy for analyzing endoscopic images in diagnosing esophagitis.
• Further research is recommended to validate and implement AI-driven diagnostic tools in clinical practice.
• AI integration can enhance the accuracy and efficiency of diagnosing esophageal diseases.