🧠 neuroscience

The Rhythm of Normality: A Comprehensive Normative Database for TMS-EEG Metrics with Reliability Characterization

This study establishes a comprehensive, open-access normative database of TMS-EEG metrics derived from 164 healthy adults, characterizing the reliability and redundancy of 968 features to enable validated individual-level cortical excitability and connectivity assessments.

Original authors: Couto, B. A. N., De Martino, E., Mazhari-Jensen, D., Jakobsen, A., Bach, M. M., Gianotta, A., Ingemann-Molden, S., Graven-Nielsen, T., Casali, A. G., de Andrade, D. C.

Published 2026-04-27

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Couto, B. A. N., De Martino, E., Mazhari-Jensen, D., Jakobsen, A., Bach, M. M., Gianotta, A., Ingemann-Molden, S., Graven-Nielsen, T., Casali, A. G., de Andrade, D. C.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine your brain is a massive, bustling orchestra. For a long time, scientists trying to understand how this orchestra plays have only been able to listen to the entire group playing together. They could tell if the whole orchestra sounded "healthy" or "sick" on average, but they couldn't tell if a single violinist was playing out of tune.

This paper is about building a new kind of sheet music that lets us check the health of individual musicians.

Here is the story of how they did it, broken down into simple steps:

1. The Problem: Too Many Guesses

Scientists use a special tool called TMS-EEG. Think of this as a "brain tap" (TMS) that gently knocks on the skull to wake up a specific part of the brain, while a helmet of sensors (EEG) listens to the brain's electrical reply.

Until now, analyzing these replies was like trying to guess the weather by looking at a single cloud. The data was messy, and scientists mostly compared groups of people. They didn't have a solid "rulebook" to say, "This specific brain signal is normal for you, and this one is weird."

2. The Solution: Building a "Brain Library"

The researchers decided to build a giant, open-access library of "normal" brain signals.

The Collection: They gathered data from 164 healthy adults (mostly young to middle-aged) from nine different studies.
The Cleanup: They used a special "harmonized" cleaning process. Imagine taking 9 different sets of photos taken with different cameras and lighting, and using software to make them all look like they were taken with the same high-quality camera. This made the data consistent and trustworthy.

3. The Quality Check: Is the Ruler Reliable?

Before they could use this library, they had to make sure their measuring tools were accurate. They took a smaller group of 57 people and measured them twice (like measuring your height on Monday and then again on Tuesday).

The Result: They looked at nearly 1,000 different brain signals (features). They found that about 54% of these signals were reliable enough to be used as a trustworthy ruler. The rest were too shaky or inconsistent to rely on for checking a single person.

4. Finding the Clues: Sorting the Noise

With so many signals, there was a lot of repetition. It was like having a dictionary where 50 different words all meant the exact same thing.

The researchers used a "clustering" method to group these signals. They found that the signals naturally fell into three main groups. Within each group, the signals were very similar to each other. This helped them realize they didn't need to measure every single thing; they just needed to measure the key representatives of each group.

5. The Final Product: A Benchmark for Individuals

The end result is a publicly available database that acts as a "Gold Standard" for a healthy brain.

The Test: To prove it works, they took data from one test patient and compared it against their new library. The library successfully flagged that this patient's brain signals were "abnormal" compared to the healthy norm.
The Goal: This isn't just a static report; it's a living platform. It allows doctors and scientists to compare an individual's brain directly against a massive, verified list of healthy brains, rather than just guessing based on averages.

In short: This paper built a reliable, standardized "normal" reference library for brain signals. It filtered out the unreliable measurements, organized the data to remove duplicates, and proved that we can now use this library to spot differences in individual brains, not just groups.

1. Problem Statement

Transcranial Magnetic Stimulation combined with electroencephalography (TMS-EEG) is a powerful technique for probing cortical excitability and connectivity. However, the field faces two critical limitations:

Group-Level Bias: Current analyses predominantly rely on group-level inferences, which obscure individual variations and limit clinical applicability for single-subject diagnosis.
Lack of Validation: There is insufficient validation regarding the reliability of individual TMS-EEG metrics and the redundancy among the high-dimensional feature sets extracted from the data. Without a standardized normative baseline, it is difficult to determine if a specific individual's results are abnormal.

2. Methodology

The authors employed a rigorous, multi-step approach to construct a robust normative database:

Data Aggregation: They pooled TMS-EEG data recorded from the primary motor cortex (M1) across nine distinct studies.
Cohort: The dataset comprises 164 healthy adults (mean age 30.8 ± 9.8 years; 88 female).
Harmonization: A critical methodological step involved applying harmonized acquisition and preprocessing pipelines across all studies to ensure data consistency and comparability.
Reliability Assessment:
- A test-retest subset (N=57) was utilized to evaluate the stability of the metrics.
- 968 distinct features were extracted and analyzed.
- Metrics evaluated included systematic bias, absolute error, and relative reliability using the Intraclass Correlation Coefficient (ICC), specifically focusing on the lower bound of the 95% confidence interval.
Feature Analysis: Cluster analysis was performed to map the high-dimensional feature space, quantifying redundancy and correlation structures among metrics.
Platform Development: The team established normative distributions and launched an online benchmarking platform to facilitate open access.

3. Key Contributions

First Comprehensive Normative Database: Creation of a large-scale, open-access reference dataset specifically designed for individual-level comparison in TMS-EEG.
Reliability Characterization: A systematic evaluation of 968 features, providing the first clear map of which TMS-EEG metrics are stable enough for clinical or individual research use.
Redundancy Mapping: Identification of feature clusters, offering guidance on which metrics provide unique information versus those that are highly correlated (redundant).
Open-Science Infrastructure: Development of a publicly available platform that allows for community contributions and ongoing benchmarking.

4. Key Results

Reliability Standards: Out of 968 features analyzed in the test-retest subset, 525 features (54.3%) met at least moderate reliability standards (defined as an ICC lower bound > 0.5). This confirms that a significant portion of the TMS-EEG feature space is robust enough for individual assessment.
Feature Redundancy: Cluster analysis revealed substantial redundancy within the dataset. Three distinct clusters were identified with moderate-to-high internal correlations ( $|r| = 0.64, 0.48, 0.39$ ), suggesting that many extracted metrics may be measuring overlapping physiological phenomena.
Clinical Feasibility: The study successfully applied the normative database to a test patient, identifying abnormal results. This demonstrated the practical feasibility of using the database for individual-level classification within an open-science framework.

5. Significance

This work represents a paradigm shift in TMS-EEG research from purely group-level statistics to precision medicine applications. By harmonizing data pipelines and establishing a validated normative reference, the authors have provided the field with:

A validated tool for distinguishing pathological from normal cortical activity in single subjects.
A standardized framework that reduces variability caused by methodological differences between studies.
An open platform that fosters community collaboration, ensuring the database remains current and expandable for future neuroscientific and clinical discoveries.