Block-As-Domain Adaptation for Workload Prediction from fNIRS Data

Overview

• Functional near-infrared spectroscopy (fNIRS) is a non-invasive way to measure brain activity
• Predicting cognitive workload from fNIRS data has used a variety of methods
• To be useful in real-world settings, models need to perform well across different sessions and subjects
• Existing work assumes training and testing data come from the same subjects, limiting generalization
• fNIRS data has high variation between subjects and within the same subject across sessions
• This paper proposes a new method called class-aware-block-aware domain adaptation (CABA-DA) to address these challenges

Plain English Explanation

The paper is about using functional near-infrared spectroscopy (fNIRS) to measure brain activity and predict a person's cognitive workload. fNIRS is a non-invasive technology that can track blood flow changes in the brain, which are linked to mental effort.

Predicting cognitive workload from fNIRS data has been done in various ways, but to be useful in real-world settings, the models need to work well across different people and testing sessions, not just the same person. The problem is that most existing methods can't generalize well to new people who weren't in the original training data.

There are also challenges with fNIRS data itself - it can vary a lot between people, and even change for the same person across different testing sessions. To address these issues, the researchers propose a new method called CABA-DA. This aims to minimize the differences in the fNIRS data within the same person's testing sessions, while also maximizing the differences between people with different cognitive workloads.

Along with the CABA-DA technique, the paper also introduces a new machine learning model based on an MLPMixer architecture for classifying cognitive workload levels. The authors show this combined approach outperforms other baseline methods on several publicly available fNIRS datasets.

Technical Explanation

The paper proposes a new method called class-aware-block-aware domain adaptation (CABA-DA) to address the challenges of predicting cognitive workload from functional near-infrared spectroscopy (fNIRS) data.

The key elements of the CABA-DA approach are:

• Viewing different testing blocks from the same subject as separate "domains" to minimize intra-session variance
• Explicitly minimizing the intra-class domain discrepancy and maximizing the inter-class domain discrepancy
• Using this domain adaptation technique in conjunction with an MLPMixer-based classification model

The paper evaluates the proposed method on three public fNIRS datasets related to cognitive workload, including n-back tasks and finger tapping. Compared to baseline models, the CABA-DA approach demonstrates improved performance in classifying the level of cognitive load.

The authors also show that their contrastive learning-based domain adaptation can enhance the baseline models as well.

Critical Analysis

The paper presents a novel and effective approach to the challenge of predicting cognitive workload from fNIRS data, which is an important application for neuroergonomics and real-time cognitive load assessment.

However, the authors acknowledge some limitations of their work:

• The method has only been evaluated on a few public datasets, and may not generalize as well to other tasks or settings
• The high variability in fNIRS data remains a significant challenge, even with the CABA-DA approach
• The paper does not address the problem of temporal differences across electrodes, which can also impact model performance

Additionally, while the CABA-DA technique is a promising contribution, it would be valuable to see how it compares to other domain adaptation or meta-learning methods that could also help with generalizing to new subjects.

Overall, this paper takes an important step towards making fNIRS-based cognitive workload prediction more practical for real-world applications, but there is still room for further research and refinement.

Conclusion

This paper proposes an effective method called CABA-DA to address the challenges of predicting cognitive workload from functional near-infrared spectroscopy (fNIRS) data. The key innovations are:

• Treating different testing blocks from the same subject as separate "domains" to minimize intra-session variance
• Explicitly optimizing the intra-class and inter-class domain discrepancies
• Using this domain adaptation approach with an MLPMixer-based classification model

Experiments on several public fNIRS datasets show the proposed approach outperforms baseline methods, demonstrating its potential to make fNIRS-based cognitive load assessment more practical and generalizable to real-world settings. While the method has some limitations, it represents an important step forward in this important area of neuroergonomics and cognitive state monitoring.