MCSF-Net: a multi-scale channel spatial fusion network for real-time polyp segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Colorectal cancer (CRC) is a significant global health concern, necessitating effective screening methods like colonoscopy.
• Missed colorectal polyps during colonoscopy remain a challenge, potentially leading to delayed cancer diagnosis and increased mortality.
• Accurate polyp segmentation from colonoscopy images is crucial for diagnosis and surgical planning.

Purpose of the Study:

• To develop and evaluate a novel deep learning framework, MCSF-Net, for real-time automatic segmentation of colorectal polyps.
• To improve the accuracy and efficiency of polyp detection in colonoscopy, thereby assisting gastroenterologists.
• To address the limitations of current computer-aided diagnostic systems in detecting missed polyps.

Main Methods:

• Introduction of MCSF-Net, a multi-scale channel space fusion network for automatic polyp segmentation.
• Utilizing multi-scale fusion, spatial and channel attention mechanisms, and a feature complementation module to enhance feature representation.
• Incorporating shape blocks for improved model supervision and precise boundary feature identification.

Main Results:

• MCSF-Net demonstrated superior performance compared to state-of-the-art methods across five benchmark datasets.
• The framework achieved real-time segmentation at approximately 45 frames per second (FPS).
• The proposed approach showed significant advantages in scalability and real-time processing capabilities.

Conclusions:

• MCSF-Net offers a promising solution for accurate and efficient polyp segmentation in colonoscopy.
• The real-time performance and high accuracy of MCSF-Net can aid in reducing missed polyp rates and improving early detection of colorectal cancer.
• This deep learning framework has the potential to enhance clinical decision-making and patient outcomes in colorectal cancer screening.