Methodology

Version 17.1 by manuelmenendez on 2025/02/09 13:01

Overview

Neurodiagnoses develops a tridimensional diagnostic framework for CNS diseases, incorporating AI-powered annotation tools to improve interpretability, standardization, and clinical utility.

This methodology integrates multi-modal data, including:
Genetic data (whole-genome sequencing, polygenic risk scores).
Neuroimaging (MRI, PET, EEG, MEG).
Neurophysiological data (EEG-based biomarkers, sleep actigraphy).
CSF & Blood Biomarkers (Amyloid-beta, Tau, Neurofilament Light).

By applying machine learning models, Neurodiagnoses generates structured, explainable diagnostic outputs to assist clinical decision-making and biomarker-driven patient stratification.

Data Integration & External Databases

How to Use External Databases in Neurodiagnoses

Neurodiagnoses integrates data from multiple biomedical and neurological research databases. Researchers can follow these steps to access, prepare, and integrate data into the Neurodiagnoses framework.

Potential Data Sources
Reference: List of Potential Databases

Register for Access

Each external database requires individual registration and approval.
✔️ Follow the official data access guidelines of each provider.
✔️ Ensure compliance with ethical approvals and data-sharing agreements (DUAs).

Download & Prepare Data

Once access is granted, download datasets following compliance guidelines and format requirements for integration.

Supported File Formats

Tabular Data: .csv, .tsv
Neuroimaging Data: .nii, .dcm
Genomic Data: .fasta, .vcf
Clinical Metadata: .json, .xml

Mandatory Fields for Integration

Field Name	Description
Subject ID	Unique patient identifier
Diagnosis	Standardized disease classification
Biomarkers	CSF, plasma, or imaging biomarkers
Genetic Data	Whole-genome or exome sequencing
Neuroimaging Metadata	MRI/PET acquisition parameters

Upload Data to Neurodiagnoses

Option 1: Upload to EBRAINS Bucket → Neurodiagnoses Data Storage
Option 2: Contribute via GitHub Repository → GitHub Data Repository

For large datasets, please contact project administrators before uploading.

Integrate Data into AI Models

Use Jupyter Notebooks on EBRAINS for data preprocessing.
Standardize data using harmonization tools.
Train AI models with newly integrated datasets.

Reference: Data Processing Guide

AI-Powered Annotation & Machine Learning Models

Neurodiagnoses applies advanced machine learning models to classify CNS diseases, extract features from biomarkers and neuroimaging, and provide AI-powered annotation.

AI Model Categories

Model Type	Function	Example Algorithms
Probabilistic Diagnosis	Assigns probability scores to multiple CNS disorders.	Random Forest, XGBoost, Bayesian Networks
Tridimensional Diagnosis	Classifies disorders based on Etiology, Biomarkers, and Neuroanatomical Correlations.	CNNs, Transformers, Autoencoders
Biomarker Prediction	Predicts missing biomarker values using regression.	KNN Imputation, Bayesian Estimation
Neuroimaging Feature Extraction	Extracts patterns from MRI, PET, EEG.	CNNs, Graph Neural Networks
Clinical Decision Support	Generates AI-driven diagnostic reports.	SHAP Explainability Tools

Reference: AI Model Documentation

Clinical Decision Support & Tridimensional Diagnostic Framework

Neurodiagnoses generates structured AI reports for clinicians, combining:

Probabilistic Diagnosis: AI-generated ranking of potential diagnoses.
Tridimensional Classification: Standardized diagnostic reports based on:

Axis 1: Etiology → Genetic, Autoimmune, Prion, Toxic, Vascular.
Axis 2: Molecular Markers → CSF, Neuroinflammation, EEG biomarkers.
Axis 3: Neuroanatomoclinical Correlations → MRI atrophy, PET.

Reference: Tridimensional Classification Guide

Data Security, Compliance & Federated Learning

✔ Privacy-Preserving AI: Implements Federated Learning, ensuring that patient data never leaves local institutions.
✔ Secure Data Access: Data remains stored in EBRAINS MIP servers using differential privacy techniques.
✔ Ethical & GDPR Compliance: Data-sharing agreements must be signed before use.

Reference: Data Protection & Federated Learning

Data Processing & Integration with Clinica.Run

Neurodiagnoses now supports Clinica.Run, an open-source neuroimaging platform for multimodal data processing.

How It Works

✔ Neuroimaging Preprocessing: MRI, PET, EEG data is preprocessed using Clinica.Run pipelines.
✔ Automated Biomarker Extraction: Extracts volumetric, metabolic, and functional biomarkers.
✔ Data Security & Compliance: Clinica.Run is GDPR & HIPAA-compliant.