A Dual-System Architecture

Neurodiagnoses integrates explainable AI for deep diagnosis with machine learning for robust prognosis.

The "Glass-Box" Bayesian Engine

For Diagnosis: Understanding the "Now"

The reasoning core of the system, designed for transparency. It processes multi-modal patient data through our tridimensional framework to generate a comprehensive "Neurodegenerative Signature."

Input: Evidence from Etiology, Molecular Pathology, and Phenotype Axes.
Process: Probabilistic inference against a dynamic, machine-readable Knowledge Base.
Output: A dual report with a classical differential and a rich, tridimensional annotation.

The "Black-Box" ML Pipelines

For Prognosis: Predicting the "Next"

Leverages proven Machine Learning models (Cox Proportional Hazards, Polygenic Hazard Scores) trained on large-scale datasets to predict future outcomes, rescuing powerful legacy components of the project.

Input: Longitudinal data and genetic profiles.
Process: Survival analysis and risk scoring models.
Output: Actionable predictions, such as progression risk and estimated decline rates.

A detailed technical breakdown of all components is available in `ARCHITECTURE.md` within the private engine repository.