ROSE: decision trees, automatic learning and their applications in cardiac medicine

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Computerized decision support systems are crucial in medicine, but complexity limits their use.
• Mitral valve prolapse (PMV) diagnosis has uncertainties, despite echocardiography's accuracy.
• General practitioners need tools to identify potential PMV patients effectively.

Purpose of the Study:

• To develop a computerized tool, ROSE, for mitral valve prolapse (PMV) determination.
• To enable general practitioners to evaluate PMV using conventional methods and identify at-risk patients.
• To support the definition of new diagnostic criteria for PMV.

Main Methods:

• Developed ROSE, a knowledge-based system using decision trees and automatic learning algorithms.
• Trained ROSE with a dataset of 400 examined volunteers for PMV classification.
• Employed decision trees to classify patients as positive or negative instances of PMV syndrome.

Main Results:

• ROSE successfully aids in the identification of potential PMV patients.
• The decision tree approach provides a visualized model for understanding diagnostic criteria.
• The system facilitates the discovery of PMV-related symptoms and syndromes.

Conclusions:

• ROSE offers a practical solution for general practitioners to screen for mitral valve prolapse.
• The computerized tool enhances diagnostic capabilities by leveraging automatic learning.
• Further research can refine criteria and improve PMV diagnosis accuracy.