What is this about and what can I find here?

Overview

The classification and diagnosis of central nervous system (CNS) diseases have long been constrained by traditional phenotype-based approaches, which often fail to capture the complex pathophysiological mechanisms, molecular biomarkers, and neuroanatomical changes that drive disease progression. Neurodegenerative and psychiatric disorders, for example, exhibit significant clinical overlap, co-pathology, and heterogeneity, making current diagnostic models insufficient.

This project proposes a new diagnostic framework—one that shifts from symptom-based classifications to an etiology-driven, tridimensional system. By integrating genetics, proteomics, neuroimaging, computational modeling, and AI-powered annotations, this approach aims to provide a more precise, scalable, and biologically grounded method for diagnosing and managing CNS diseases.

The AI-powered annotation system plays a critical role by structuring, interpreting, and tracking multi-modal data, ensuring real-time disease progression analysis, clinician decision support, and personalized treatment pathways.

Project Aim

Neurodiagnoses is an open-source AI-powered diagnostic system designed for complex central nervous system (CNS) disorders, including neurodegenerative diseases, autoimmune encephalopathies, prion disorders, and genetic syndromes. The project aims to develop a tridimensional diagnostic framework with an AI-powered annotation system, integrating etiology, molecular biomarkers, and neuroanatomoclinical correlations for precise, standardized, and scalable CNS disease diagnostics.

The Tridimensional Diagnostic Framework redefines CNS diseases can be classified and diagnosed by focusing on:

• Axis 1: Etiology (genetic or other causes of diseases).
• Axis 2: Molecular Markers (biomarkers).
• Axis 3: Neuroanatomoclinical correlations (linking clinical symptoms to structural changes in the nervous system).

This methodology enables:

• Greater precision in diagnosis.
• Integration of incomplete datasets using AI-driven probabilistic modeling.
• Stratification of patients for personalized treatment.

The Role of AI-Powered Annotation

To enhance standardization, interpretability, and clinical application, the framework integrates an AI-powered annotation system, which:

• Assigns structured metadata tags to diagnostic features.
• Provides real-time contextual explanations for AI-based classifications.
• Tracks longitudinal disease progression using timestamped AI annotations.
• Improves AI model transparency through interpretability tools (e.g., SHAP analysis).
• Facilitates decision-making for clinicians by linking annotations to standardized biomedical ontologies (SNOMED, HPO).

Neurodiagnoses provides two complementary AI-driven diagnostic approaches:

1. Traditional Probabilistic Diagnosis

• AI provides multiple possible diagnoses, each assigned a probability percentage based on biomarker, imaging, and clinical data.
• Example Output:

75% Alzheimer's Disease  
20% Lewy Body Dementia  
5% Vascular Dementia  

• Useful for differential diagnosis and treatment decision-making.

2. Tridimensional Diagnosis

• Diagnoses are structured based on:
  (1) Etiology (genetic, autoimmune, metabolic, infectious)
  (2) Molecular Biomarkers (amyloid-beta, tau, inflammatory markers, EEG patterns)
  (3) Neuroanatomoclinical Correlations (brain atrophy, connectivity alterations)
• This approach enables precise disease subtyping and biologically meaningful classification.

For every patient case, both systems will be offered, allowing clinicians to compare AI-generated probabilistic diagnosis with a structured tridimensional classification.

The case of neurodegenerative diseases

There have been described these 3 diagnostic axes:

• Axis 1: Etiology

  • Description: Focuses on genetic and sporadic causes, identifying risk factors and potential triggers.
  • Examples: APOE ε4 as a genetic risk factor, or cardiovascular health affecting NDD progression.
  • Tests: Genetic testing, lifestyle, and cardiovascular screening.

   

• Axis 2: Molecular Markers

  • Description: Analyzes primary (amyloid-beta, tau) and secondary biomarkers (NFL, GFAP) for tracking disease progression.
  • Examples: CSF amyloid-beta concentrations to confirm Alzheimer’s pathology.
  • Tests: Blood/CSF biomarkers, PET imaging (Tau-PET, Amyloid-PET).

   

• Axis 3: Neuroanatomoclinical

  • Description: Links clinical symptoms to neuroanatomical changes, such as atrophy or functional impairments.
  • Examples: Hippocampal atrophy correlating with memory deficits.
  • Tests: MRI volumetrics, FDG-PET, neuropsychological evaluations.

Applications

This system enhances:

• Research: By stratifying patients, reduces cohort heterogeneity in clinical trials.
• Clinical Practice: Provides dynamic diagnostic annotations with timestamps for longitudinal tracking.

How to Contribute

• Access the `/docs` folder for guidelines.
• Use `/code` for the latest AI pipelines.
• Share feedback and ideas in the wiki discussion pages.

Key Objectives

• Develop interpretable AI models for diagnosis and progression tracking. 
• Integrate data from Human Phenotype Ontology (HPO), Gene Ontology (GO), and other biomedical resources.
• Foster collaboration among neuroscientists, AI researchers, and clinicians.

Who has access?

We welcome contributions from the global community. Let’s build the future of neurological diagnostics together!

Main contents

• `/docs`: Documentation and contribution guidelines.
• `/code`: Machine learning pipelines and scripts.
• `/data`: Sample datasets for testing.
• `/outputs`: Generated models, visualizations, and reports.
• Methodology
• Notebooks
• Results
• to-do-list