Hyperbolic Dirac Nets for medical decision support. Theory, methods, and comparison with Bayes Nets

Area of Science:

• Medical Inference
• Computational Biology
• Graph Theory

Background:

• Traditional Bayes Nets (BNs) are widely used in medicine but struggle with complex interdependencies common in biological systems.
• BNs are defined as directed acyclic graphs, which inherently limit their ability to model cyclic relationships.
• The need for more flexible models in medical inference is apparent due to the intricate nature of biological networks.

Purpose of the Study:

• Introduce the Hyperbolic Dirac Net (HDN) as a more suitable model for medical inference.
• Address the limitations of traditional Bayes Nets in handling cyclic dependencies and self-conditional relationships.
• Provide methods for converting existing BNs to HDNs and introduce a new data-driven approach.

Main Methods:

• The HDN is conceptualized as a bidirectional general graph, utilizing imaginary numbers from quantum mechanics to encode bidirectionality and cycles.
• The P-method is presented for converting traditional BNs into HDNs, enabling the incorporation of cycles.
• The K-method, a simpler and more general approach often derived from data mining, is also introduced as an alternative for constructing HDN-like models.

Main Results:

• The HDN framework accommodates cyclic dependencies, offering a more robust approach to medical inference compared to acyclic BNs.
• The P-method provides a pathway for transitioning from BNs to HDNs, though it may require probability adjustments for coherence.
• The K-method offers a flexible, data-driven alternative for building HDN models, potentially simplifying the process.

Conclusions:

• The Hyperbolic Dirac Net (HDN) represents a significant advancement for medical inference, capable of modeling complex biological interdependencies.
• HDNs can be viewed as an extension of Bayes Nets, offering enhanced capabilities for representing cyclic relationships.
• Both conversion methods (P-method) and data-driven approaches (K-method) contribute to the practical application of HDNs in medical research.