Changes for page Methodology

Last modified by manuelmenendez on 2025/03/14 08:31
From version 3.1
edited by manuelmenendez
on 2025/01/27 23:28
Change comment: There is no comment for this version
To version 7.1
edited by manuelmenendez
on 2025/02/01 14:11
Change comment: There is no comment for this version

Summary

Details

Page properties
Content
... ... @@ -1,7 +1,23 @@
1 -=== **Overview** ===
1 +==== **Overview** ====
2 2  
3 -This section describes the step-by-step process used in the **Neurodiagnoses** project to develop a novel diagnostic framework for neurological diseases. The methodology integrates artificial intelligence (AI), biomedical ontologies, and computational neuroscience to create a structured, interpretable, and scalable diagnostic system.
3 +This project develops a **tridimensional diagnostic framework** for **CNS diseases**, incorporating **AI-powered annotation tools** to improve **interpretability, standardization, and clinical utility**. The methodology integrates **multi-modal data**, including **genetic, neuroimaging, neurophysiological, and biomarker datasets**, and applies **machine learning models** to generate **structured, explainable diagnostic outputs**.
4 4  
5 +=== **Workflow** ===
6 +
7 +1. (((
8 +**We Use GitHub for AI Development**
9 +
10 +* Create a **GitHub repository** for AI scripts and models.
11 +* Use **GitHub Projects** to manage research milestones.
12 +)))
13 +1. (((
14 +**We Use EBRAINS for Data & Collaboration**
15 +
16 +* Store **biomarker and neuroimaging data** in **EBRAINS Buckets**.
17 +* Run **Jupyter Notebooks** in **EBRAINS Lab** to test AI models.
18 +* Use **EBRAINS Wiki** for structured documentation and research discussion.
19 +)))
20 +
5 5  ----
6 6  
7 7  === **1. Data Integration** ===
... ... @@ -8,99 +8,166 @@
8 8  
9 9  ==== **Data Sources** ====
10 10  
11 -* **Biomedical Ontologies**:
12 -** Human Phenotype Ontology (HPO) for phenotypic abnormalities.
13 -** Gene Ontology (GO) for molecular and cellular processes.
14 -* **Neuroimaging Datasets**:
15 -** Example: Alzheimer’s Disease Neuroimaging Initiative (ADNI), OpenNeuro.
16 -* **Clinical and Biomarker Data**:
17 -** Anonymized clinical reports, molecular biomarkers, and test results.
27 +**Biomedical Ontologies & Databases:**
18 18  
19 -==== **Data Preprocessing** ====
29 +* **Human Phenotype Ontology (HPO)** for symptom annotation.
30 +* **Gene Ontology (GO)** for molecular and cellular processes.
20 20  
21 -1. **Standardization**: Ensure all data sources are normalized to a common format.
22 -1. **Feature Selection**: Identify relevant features for diagnosis (e.g., biomarkers, imaging scores).
23 -1. **Data Cleaning**: Handle missing values and remove duplicates.
32 +**Dimensionality Reduction and Interpretability:**
24 24  
34 +* **Evaluate interpretability** using metrics like the **Area Under the Interpretability Curve (AUIC)**.
35 +* **Leverage DEIBO (Data-driven Embedding Interpretation Based on Ontologies)** to connect model dimensions to ontology concepts.
36 +
37 +**Neuroimaging & EEG/MEG Data:**
38 +
39 +* **MRI volumetric measures** for brain atrophy tracking.
40 +* **EEG functional connectivity patterns** (AI-Mind).
41 +
42 +**Clinical & Biomarker Data:**
43 +
44 +* **CSF biomarkers** (Amyloid-beta, Tau, Neurofilament Light).
45 +* **Sleep monitoring and actigraphy data** (ADIS).
46 +
47 +**Federated Learning Integration:**
48 +
49 +* **Secure multi-center data harmonization** (PROMINENT).
50 +
25 25  ----
26 26  
53 +==== **Annotation System for Multi-Modal Data** ====
54 +
55 +To ensure **structured integration of diverse datasets**, **Neurodiagnoses** will implement an **AI-driven annotation system**, which will:
56 +
57 +* **Assign standardized metadata tags** to diagnostic features.
58 +* **Provide contextual explanations** for AI-based classifications.
59 +* **Track temporal disease progression annotations** to identify long-term trends.
60 +
61 +----
62 +
27 27  === **2. AI-Based Analysis** ===
28 28  
29 -==== **Model Development** ====
65 +==== **Machine Learning & Deep Learning Models** ====
30 30  
31 -* **Embedding Models**: Use pre-trained models like BioBERT or BioLORD for text data.
32 -* **Classification Models**:
33 -** Algorithms: Random Forest, Support Vector Machines (SVM), or neural networks.
34 -** Purpose: Predict the likelihood of specific neurological conditions based on input data.
67 +**Risk Prediction Models:**
35 35  
36 -==== **Dimensionality Reduction and Interpretability** ====
69 +* **LETHE’s cognitive risk prediction model** integrated into the annotation framework.
37 37  
38 -* Leverage [[DEIBO>>https://drive.ebrains.eu/f/8d7157708cde4b258db0/]] (Data-driven Embedding Interpretation Based on Ontologies) to connect model dimensions to ontology concepts.
39 -* Evaluate interpretability using metrics like the Area Under the Interpretability Curve (AUIC).
71 +**Biomarker Classification & Probabilistic Imputation:**
40 40  
73 +* **KNN Imputer** and **Bayesian models** used for handling **missing biomarker data**.
74 +
75 +**Neuroimaging Feature Extraction:**
76 +
77 +* **MRI & EEG data** annotated with **neuroanatomical feature labels**.
78 +
79 +==== **AI-Powered Annotation System** ====
80 +
81 +* Uses **SHAP-based interpretability tools** to explain model decisions.
82 +* Generates **automated clinical annotations** in structured reports.
83 +* Links findings to **standardized medical ontologies** (e.g., **SNOMED, HPO**).
84 +
41 41  ----
42 42  
43 -=== **3. Diagnostic Framework** ===
87 +=== **3. Diagnostic Framework & Clinical Decision Support** ===
44 44  
45 -==== **Axes of Diagnosis** ====
89 +==== **Tridimensional Diagnostic Axes** ====
46 46  
47 -The framework organizes diagnostic data into three axes:
91 +**Axis 1: Etiology (Pathogenic Mechanisms)**
48 48  
49 -1. **Etiology**: Genetic and environmental risk factors.
50 -1. **Molecular Markers**: Biomarkers such as amyloid-beta, tau, and alpha-synuclein.
51 -1. **Neuroanatomical Correlations**: Results from neuroimaging (e.g., MRI, PET).
93 +* Classification based on **genetic markers, cellular pathways, and environmental risk factors**.
94 +* **AI-assisted annotation** provides **causal interpretations** for clinical use.
52 52  
53 -==== **Recommendation System** ====
96 +**Axis 2: Molecular Markers & Biomarkers**
54 54  
55 -* Suggests additional tests or biomarkers if gaps are detected in the data.
56 -* Prioritizes tests based on clinical impact and cost-effectiveness.
98 +* **Integration of CSF, blood, and neuroimaging biomarkers**.
99 +* **Structured annotation** highlights **biological pathways linked to diagnosis**.
57 57  
101 +**Axis 3: Neuroanatomoclinical Correlations**
102 +
103 +* **MRI and EEG data** provide anatomical and functional insights.
104 +* **AI-generated progression maps** annotate **brain structure-function relationships**.
105 +
58 58  ----
59 59  
60 -=== **4. Computational Workflow** ===
108 +=== **4. Computational Workflow & Annotation Pipelines** ===
61 61  
62 -1. **Data Loading**: Import data from storage (Drive or Bucket).
63 -1. **Feature Engineering**: Generate derived features from the raw data.
64 -1. **Model Training**:
65 -1*. Split data into training, validation, and test sets.
66 -1*. Train models with cross-validation to ensure robustness.
67 -1. **Evaluation**:
68 -1*. Metrics: Accuracy, F1-Score, AUIC for interpretability.
69 -1*. Compare against baseline models and domain benchmarks.
110 +==== **Data Processing Steps** ====
70 70  
112 +**Data Ingestion:**
113 +
114 +* **Harmonized datasets** stored in **EBRAINS Bucket**.
115 +* **Preprocessing pipelines** clean and standardize data.
116 +
117 +**Feature Engineering:**
118 +
119 +* **AI models** extract **clinically relevant patterns** from **EEG, MRI, and biomarkers**.
120 +
121 +**AI-Generated Annotations:**
122 +
123 +* **Automated tagging** of diagnostic features in **structured reports**.
124 +* **Explainability modules (SHAP, LIME)** ensure transparency in predictions.
125 +
126 +**Clinical Decision Support Integration:**
127 +
128 +* **AI-annotated findings** fed into **interactive dashboards**.
129 +* **Clinicians can adjust, validate, and modify annotations**.
130 +
71 71  ----
72 72  
73 -=== **5. Validation** ===
133 +=== **5. Validation & Real-World Testing** ===
74 74  
75 -==== **Internal Validation** ====
135 +==== **Prospective Clinical Study** ====
76 76  
77 -* Test the system using simulated datasets and known clinical cases.
78 -* Fine-tune models based on validation results.
137 +* **Multi-center validation** of AI-based **annotations & risk stratifications**.
138 +* **Benchmarking against clinician-based diagnoses**.
139 +* **Real-world testing** of AI-powered **structured reporting**.
79 79  
80 -==== **External Validation** ====
141 +==== **Quality Assurance & Explainability** ====
81 81  
82 -* Collaborate with research institutions and hospitals to test the system in real-world settings.
83 -* Use anonymized patient data to ensure privacy compliance.
143 +* **Annotations linked to structured knowledge graphs** for improved transparency.
144 +* **Interactive annotation editor** allows clinicians to validate AI outputs.
84 84  
85 85  ----
86 86  
87 87  === **6. Collaborative Development** ===
88 88  
89 -The project is open to contributions from researchers, clinicians, and developers. Key tools include:
150 +The project is **open to contributions** from **researchers, clinicians, and developers**.
90 90  
152 +**Key tools include:**
153 +
91 91  * **Jupyter Notebooks**: For data analysis and pipeline development.
155 +** Example: **probabilistic imputation**
92 92  * **Wiki Pages**: For documenting methods and results.
93 93  * **Drive and Bucket**: For sharing code, data, and outputs.
158 +* **Collaboration with related projects**:
159 +** Example: **Beyond the hype: AI in dementia – from early risk detection to disease treatment**
94 94  
95 95  ----
96 96  
97 97  === **7. Tools and Technologies** ===
98 98  
99 -* **Programming Languages**: Python for AI and data processing.
100 -* **Frameworks**:
101 -** TensorFlow and PyTorch for machine learning.
102 -** Flask or FastAPI for backend services.
103 -* **Visualization**: Plotly and Matplotlib for interactive and static visualizations.
104 -* **EBRAINS Services**:
105 -** Collaboratory Lab for running Notebooks.
106 -** Buckets for storing large datasets.
165 +==== **Programming Languages:** ====
166 +
167 +* **Python** for AI and data processing.
168 +
169 +==== **Frameworks:** ====
170 +
171 +* **TensorFlow** and **PyTorch** for machine learning.
172 +* **Flask** or **FastAPI** for backend services.
173 +
174 +==== **Visualization:** ====
175 +
176 +* **Plotly** and **Matplotlib** for interactive and static visualizations.
177 +
178 +==== **EBRAINS Services:** ====
179 +
180 +* **Collaboratory Lab** for running Notebooks.
181 +* **Buckets** for storing large datasets.
182 +
183 +----
184 +
185 +=== **Why This Matters** ===
186 +
187 +* **The annotation system ensures that AI-generated insights are structured, interpretable, and clinically meaningful.**
188 +* **It enables real-time tracking of disease progression across the three diagnostic axes.**
189 +* **It facilitates integration with electronic health records and decision-support tools, improving AI adoption in clinical workflows.**