MCH-PAN: gastrointestinal polyp detection model integrating multi-scale feature information.

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Clinical decision support systems benefit from object detection models.
• Gastrointestinal polyp detection faces challenges in identification accuracy and scale variation.

Purpose of the Study:

• To propose a novel object detection model for gastrointestinal polyp identification.
• To enhance accuracy and robustness in detecting polyps of varying scales and complexities.

Main Methods:

• Integration of multi-channel information for robust feature expression.
• Implementation of a hierarchical structure for multi-scale target adaptability.
• Inclusion of a channel attention mechanism to improve positioning accuracy.

Main Results:

• The proposed model demonstrates superior performance in gastrointestinal polyp detection.
• Experimental results confirm enhanced accuracy and reduced diagnostic uncertainty.
• The model effectively addresses challenges in polyp scale variations.

Conclusions:

• The novel gastrointestinal polyp object detection model provides reliable references for clinicians.
• The model contributes to improving the diagnostic level of digestive system diseases.
• This research offers valuable insights for related fields in medical AI.