Optimizing Pix2Pix GAN With Attention Mechanisms for AI-Driven Polyp Segmentation in IoMT-Enabled Smart Healthcare

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computer Vision

Background:

• Automated polyp segmentation in colonoscopy is crucial for early colorectal cancer detection.
• Conventional methods face challenges with variable polyp appearances, image quality, and limited data.
• Advanced AI techniques are needed to enhance segmentation precision and reliability.

Purpose of the Study:

• To introduce an enhanced Pix2Pix Generative Adversarial Network (GAN) with an attention mechanism for automated polyp segmentation.
• To improve the accuracy and robustness of polyp segmentation in colonoscopy images.
• To facilitate early detection of colorectal cancer through advanced AI in smart healthcare.

Main Methods:

• Developed an enhanced Pix2Pix GAN incorporating an attention mechanism in the discriminator.
• Employed a hybrid training strategy using both real and synthetic colonoscopy data.
• Validated the model on multiple public colonoscopy image datasets.

Main Results:

• The attention-enhanced Pix2Pix GAN demonstrated significantly improved polyp segmentation performance.
• The model achieved higher precision and reliability compared to existing state-of-the-art methods.
• Enhanced focus on intricate polyp features led to improved segmentation accuracy.

Conclusions:

• The developed AI model offers efficient and reliable automated polyp segmentation.
• The model shows potential for integration into remote health monitoring and smart healthcare systems.
• This work highlights the efficacy of AI in advancing Internet of Medical Things (IoMT)-enabled healthcare.