IRv2-Net: A Deep Learning Framework for Enhanced Polyp Segmentation Performance Integrating InceptionResNetV2 and...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Colorectal polyps are precancerous growths that can develop into colorectal cancer.
• Manual segmentation of polyps is time-consuming, error-prone, and leads to missed diagnoses.
• Automated polyp detection systems are needed to improve diagnostic accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated polyp segmentation in medical images.
• To compare the proposed model's performance against state-of-the-art methods.
• To create a user-friendly interface for real-time polyp detection.

Main Methods:

• The IRv2-Net model, utilizing a UNet architecture with an InceptionResNetV2 encoder, was developed.
• Test Time Augmentation (TTA) was employed for enhanced boundary and multi-scale feature extraction.
• The model was evaluated on Kvasir-SEG and CVC-ClinicDB datasets using metrics like accuracy, DSC, IoU, precision, and recall.

Main Results:

• The IRv2-Net model achieved superior performance on unseen data, outperforming SOTA models.
• It demonstrated high accuracy, Dice Similarity Coefficients (DSC), and Intersection over Union (IoU).
• The model successfully detected various polyp types and showed excellent results across datasets, minimizing missed detections.

Conclusions:

• The proposed IRv2-Net model offers an effective automated solution for polyp segmentation, crucial for colorectal cancer diagnosis.
• The system's real-time capabilities and accuracy have significant potential for clinical colonoscopy procedures.
• Further research can build upon this model to enhance early detection and patient outcomes.