Neurofeedback Threshold Adaptation via Logistic Transition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic thresholding systems in neurofeedback training are influenced by EEG artefacts, fail to allow subjects to learn and progress, and do not mimic the habits of trained specialists, leading to reduced efficacy and limited availability of unsupervised sessions.
Innovation Solution
A system for unsupervised adjustment of the threshold during neurofeedback sessions, where the threshold is computed as a weighted sum of the intermediate threshold and the previous threshold, with a coefficient alpha that adapts over time, allowing for real-time reward reporting and minimizing the impact of artefacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a moving average time window is used for automatic thresholding, then the threshold adapts in real-time, but the system becomes strongly influenced by EEG artefacts
Solution Approach 1:
The patent implements a dynamic threshold adjustment mechanism where the threshold evolves over time based on a logistic function that transitions from an initial value to an intermediate value. This dynamic approach allows the threshold to adapt to changing brain activity patterns while the logistic function's smoothing effect reduces sensitivity to transient artefacts in the EEG signal.
Solution Approach 2:
The system performs preliminary calibration by computing an intermediate threshold from the subject's resting-state brain activity before the actual training session begins. This preliminary action establishes a baseline that accounts for individual subject characteristics while filtering out transient artefacts present during random baseline recording.
2Adaptability or versatility
If the threshold is adjusted too adaptively to maintain constant reward rate, then real-time adaptation is achieved, but the subject cannot learn and progress
Solution Approach 1:
The patent implements a periodic adjustment strategy where the threshold transitions from the initial value to the intermediate value over a specified transition period. During this period, the threshold changes gradually according to a logistic function, providing a structured adaptation schedule that balances reward rate maintenance with learning opportunities. After the transition period, the threshold stabilizes to allow consistent learning conditions.
Solution Approach 2:
The system automatically adjusts the threshold based on pre-computed intermediate values derived from the subject's own baseline activity, eliminating the need for continuous manual intervention. The logistic function parameters are configured to provide appropriate adaptation speed, allowing the system to self-regulate the balance between maintaining engagement through reward rate control and enabling learning through stable threshold periods.
3Reliability
If manual threshold adjustment by specialists is used, then subject engagement and challenge are optimized, but session availability and scalability are limited
Solution Approach 1:
The system enables unsupervised neurofeedback sessions by automatically computing thresholds based on each subject's own baseline brain activity. The automated threshold adjustment mechanism uses the logistic function to transition between initial and intermediate thresholds, replicating the adaptive behavior of specialist-adjusted thresholds without requiring continuous human intervention. This allows subjects to undergo multiple training sessions independently.
Solution Approach 2:
The patent replicates the expert specialist's threshold adjustment strategy by computing an intermediate threshold from baseline data and using a logistic function to transition from an initial to this intermediate value. This copying approach captures the essential adaptive behavior of manual specialist adjustment, maintaining engagement optimization while enabling automated unsupervised operation.
4Extent of automation
If existing automatic thresholding is implemented, then unsupervised sessions are enabled, but the threshold does not mimic specialist habits
Solution Approach 1:
The patent copies the specialist's threshold adjustment methodology by computing an intermediate threshold from baseline brain activity and using a logistic function to transition from an initial threshold to this intermediate value. This approach replicates the gradual adaptation pattern observed in specialist practice, where thresholds are adjusted based on subject performance and baseline characteristics rather than using simple moving average methods.
Solution Approach 2:
The system changes the threshold parameter dynamically using a logistic function that transitions between initial and intermediate values based on a transition period and growth rate parameters. These parameter changes are configured to mimic specialist adjustment patterns, providing a more realistic threshold evolution compared to linear or simple adaptive methods.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The invention relates to a method for biofeedback training of a subject, said method comprising iteratively obtaining a series of biomarkers, each biomarker being representative of a bio-signal of the subject on a first time window; computing an intermediate threshold Thrintermediate(t) based on the series of biomarkers on a second time window, such that said intermediate threshold on said second time window could provide the subject with an expected reward ratio; computing a threshold Thr(t) as the weighted sum of the intermediate threshold Thrintermediate(t) and the threshold of the previous iteration Thr(t-1); and reporting in real-time a reward to the subject based on the difference between the biomarker and the computed threshold Thr(t); wherein at each iteration the time windows are moved forward in time. The invention also relates to a system for implementing the method of the invention.