Basal Ganglia Ensemble Detection Using Hierarchical Drift-Diffusion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional systems struggle to analyze neuronal activities governing decision-making processes in the basal ganglia due to the need for hidden variables that cannot be directly observed or recorded from neuronal firing patterns, making it challenging to identify neuronal ensembles in regions like the Sub-Thalamic Nucleus (STN), Globus Pallidus externa (GPe), and Globus Pallidus interna (GPi) during information processing or beta-band oscillations in Parkinson's disease.

Innovation Solution

A method and system using hierarchical drift-diffusion modeling (HDDM) to process microelectrode recording data, classify neuronal states as active or resting based on response times, and identify neuronal ensembles through clustering techniques, leveraging latent variables like drift rate to group neurons with similar dynamics.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If conventional analysis methods are used on neuronal firing patterns, then directly observable parameters can be obtained, but hidden variables governing decision-making processes cannot be identified

Engineering Contradiction:
Improveloss of hidden variable informationVSAvoidcomplexity of modeling approach
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent introduces hierarchical drift-diffusion modeling as an intermediary framework that bridges observable neuronal firing patterns and hidden decision-making variables. The model acts as a mediator that translates spike train data into latent parameters such as drift rate, boundary separation, and non-decision time, enabling indirect observation of cognitive processes that cannot be directly measured from neuronal activity alone.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms the analysis by changing parameters from directly observable spike times to latent cognitive parameters through drift-diffusion modeling. By fitting the model to neuronal data, the system extracts hidden variables including drift rate (information accumulation speed), boundary separation (decision threshold), and non-decision time, thereby converting raw neuronal signals into meaningful cognitive process parameters.

Inventive Principle:
Principle #35Parameter changes

2Loss of information

If hierarchical drift-diffusion modeling is applied to identify hidden variables, then decision-making processes can be analyzed, but computational complexity increases

Engineering Contradiction:
Improverecovery of hidden variable informationVSAvoidcomplexity of computational modeling
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent segments the complex decision-making process into distinct computational components within the drift-diffusion framework: drift rate (information accumulation), boundary separation (decision threshold), and non-decision time (peripheral processing). This segmentation allows the complex cognitive process to be analyzed through manageable, interpretable parameters that can be independently estimated and manipulated.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent adds a temporal dimension to the analysis by modeling decision-making as a dynamic process over time rather than a static state. The drift-diffusion model incorporates time-to-boundary crossing and trial-by-trial variability, transforming the analysis from simple firing rate measurements to a multi-dimensional space that includes decision trajectory, temporal evolution, and latent cognitive parameters.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Measurement precision

If neuronal ensembles are identified using clustering techniques on latent variables, then underlying patterns become visible, but direct observable parameters become insufficient

Engineering Contradiction:
Improveprecision of neuronal pattern identificationVSAvoiddifficulty of measuring hidden variables
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent creates a computational copy of the neuronal decision-making process through drift-diffusion modeling. By fitting the model to observed spike trains, the system generates latent parameter values (drift rate, boundary separation, non-decision time) that replicate the underlying cognitive dynamics. These parameter copies serve as proxies for hidden variables, enabling precise measurement and analysis without direct observation of the actual cognitive processes.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP4666945A1Method and system to identify neuronal ensembles in basal ganglia using hierarchical drift-diffusion modeling
Publication Date: 2025.12.24 TATA CONSULTANCY SERVICES LTD
  • EP4666945A1 patent drawingFigure 1
  • EP4666945A1 patent drawingFigure 2
  • EP4666945A1 patent drawingFigure 3A~3B

AI summary

The present invention generally relates to the field of computational brain modeling. It is challenging to analyse neuronal activities that governs neuronal dynamics during the decision-making process using conventional techniques. Thus, embodiments of present disclosure provide a method and system to identify neuronal ensembles in basal ganglia using hierarchical drift-diffusion modeling. Microelectrode recording data of neuronal activity of neurons in the nuclei within BG of a subject are obtained. Then, spikes and associated spike times are extracted from the obtained data using which Inter-Spike Intervals (ISIs) for each of the neurons are calculated. Response times of each neuron is determined based on the ISIs and they are classified as one of an active state and a resting state which inherently reflected the broader network states responsible for behavioral responses by the neurons. Finally, the neurons are grouped into neuronal ensembles based on HDDM latent variables like drift rate.