EEG-Based Brain Age Modelling in African Populations for MIRASOL @ MICCAI 2026
# EEG-Based Brain Age Modelling in African Populations
### Quantifying Cross-Population Generalisation Gap and First Brain Age Gap Reference for Healthy Nigerian Adults Using Low-Cost Portable EEG
---
## Table of Contents
1. Project Overview & Roadmap
2. Multi-National Dataset Registry
3. Reproducible Pipeline Architecture
4. Computation Relay Model & Division of Labor
5. Interactive Command-Line Run Guide
6. Repository Structure
7. Results Artifacts
8. Citation & Ethical Compliance
9. License
---
## 1. Project Overview & Roadmap
### 1.1 Synopsis
Brain age modelling estimates an individual's biological brain age from neurophysiological signals, with the deviation from chronological age — the **Brain Age Gap (BAG)** — serving as a quantitative biomarker of accelerated or decelerated neural ageing. Despite over a decade of methodological advancement in this field, **no published study has trained or validated an EEG-based brain age model on an indigenous African population as its primary target cohort.**
This repository contains the complete, reproducible implementation for the first such study. We construct an internationally pre-trained EEG brain age regression model using a multinational, resource-constrained-setting (RCS) inclusive training corpus, and evaluate **zero-shot** and **few-shot fine-tuned** generalisation performance on the **ASZED Nigerian healthy control cohort (N = 77)** — the first open-access African EEG dataset with exact, individually reported chronological ages.
> **Core Research Question:** Does a brain age regression model trained on a multinational, predominantly non-African EEG corpus generalise to a Nigerian population, what is the magnitude of the resulting cross-population generalisation gap (ΔMAE), and what is the normative Brain Age Gap distribution in healthy Nigerian adults?
### 1.2 Project Roadmap
```mermaid
flowchart LR
A[Multi-National
Raw EEG Corpus] --> B[Harmonised
Preprocessing Pipeline]
B --> C[247-D Feature …