Advancements in Neural Fingerprinting for Brain Research

Nikolas Kampel, Christian M. Kiefer, N. Jon Shah, Irene Neuner, Jürgen Dammers

20th September 2023

Historically, neuroscience has primarily relied on population-level data and commonalities among individuals to understand the brain. However, contemporary insights emphasise the importance of individual variations, leading to the emergence of neural fingerprinting, a field that focuses on identifying individuals within a group using various neuroimaging techniques, such as MEG.

Neuroimaging techniques like magneto- and electroencephalography (M/EEG) have relied on correlation or connectivity matrices for neural fingerprinting, which often ignored individual temporal dynamics and required channel coupling. However, a recent study conducted by scientists from INM-4 has brought significant progress.

In this study, researchers have leveraged recent developments in multivariate time series classification, including the RandOm Convolutional KErnel Transformation (ROCKET) classifier, to perform direct classification on short time segments from MEG resting-state recordings. This novel approach achieved a 99% accuracy rate in assigning 1-second time series windows to the correct individuals in a group of 124 subjects, outperforming previous methods while also reducing the time required for analysis.

These findings hold promise for gaining deeper insights into brain function, particularly in understanding neurological conditions, and may lead to more targeted interventions and treatments.

Advancements in Neural Fingerprinting for Brain Research Nikolas Kampel, Christian M. Kiefer, N. Jon Shah, Irene Neuner, Jürgen Dammers 20th September 2023 Historically, neuroscience has primarily relied on population-level data and commonalities among individuals to understand the brain. However, contemporary insights emphasise the importance of individual variations, leading to the emergence of neural fingerprinting, a field that focuses on identifying individuals within a group using various neuroimaging techniques, such as MEG. Neuroimaging techniques like magneto- and electroencephalography (M/EEG) have relied on correlation or connectivity matrices for neural fingerprinting, which often ignored individual temporal dynamics and required channel coupling. However, a recent study conducted by scientists from INM-4 has brought significant progress. In this study, researchers have leveraged recent developments in multivariate time series classification, including the RandOm Convolutional KErnel Transformation (ROCKET) classifier, to perform direct classification on short time segments from MEG resting-state recordings. This novel approach achieved a 99% accuracy rate in assigning 1-second time series windows to the correct individuals in a group of 124 subjects, outperforming previous methods while also reducing the time required for analysis. These findings hold promise for gaining deeper insights into brain function, particularly in understanding neurological conditions, and may lead to more targeted interventions and treatments. Neural fingerprinting on MEG time series using MiniRockethttps://www.frontiersin.org/articles/10.3389/fnins.2023.1229371/full

Personalising Brain Research: An artistic portrayal of a brain marked with a distinctive fingerprint, symbolizing the advent of neural fingerprinting in the field of neuroscience

Origional publication: Neural fingerprinting on MEG time series using MiniRocket

Last Modified: 29.09.2023