Automatic Extraction of Appendix from Ultrasonography with Self-Organizing Map and Shape-Brightness Pattern Learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Diagnostic Ultrasound

Background:

• Accurate diagnosis of acute appendicitis is challenging, particularly in pediatric and pregnant patients.
• Current diagnostic methods for appendicitis can be limited by patient factors and imaging interpretation.

Purpose of the Study:

• To develop a fully automatic appendix extractor for ultrasonography.
• To improve the accuracy and efficiency of appendicitis diagnosis using AI-driven image analysis.

Main Methods:

• Application of image processing algorithms and an unsupervised neural learning algorithm (self-organizing map).
• Definition and learning of four distinct appendix shape patterns based on clinical input.
• Pixel clustering for pattern recognition by the self-organizing map.

Main Results:

• The automated appendix extractor demonstrated high success rates across tested appendicitis shape patterns.
• The system achieved high accuracy, with only 1 failure in 45 cases across three patterns.
• One specific appendix shape pattern showed an 80% success rate, indicating an area for further refinement.

Conclusions:

• The proposed fully automatic appendix extractor shows significant potential for aiding in the diagnosis of acute appendicitis.
• This AI-based approach can enhance diagnostic accuracy and efficiency in challenging patient populations.
• Further development is warranted to address performance variations in specific appendix morphologies.