Deep learning-based motion artifact removal in functional near-infrared spectroscopy

Disentangling the impact of motion artifact correction algorithms on functional near-infrared spectroscopy-based brain network analysis

Significance

Functional near-infrared spectroscopy (fNIRS) has been widely used to assess brain functional networks due to its superior ecological validity. Generally, fNIRS signals are sensitive to motion artifacts (MA), which can be removed by various MA correction algorithms. Yet, fNIRS signals may also undergo varying degrees of distortion due to MA correction, leading to notable alternation in functional connectivity (FC) analysis results.

Aim

We aimed to investigate the effect of different MA correction algorithms on the performance of brain FC and topology analyses.

Approach

We evaluated various MA correction algorithms on simulated and experimental datasets, including principal component analysis, spline interpolation, correlation-based signal improvement, Kalman filtering, wavelet filtering, and temporal derivative distribution repair (TDDR). The mean FC of each pre-defined network, receiver operating characteristic (ROC), and graph theory metrics were investigated to assess the performance of different algorithms.

Results

Although most algorithms did not differ significantly from each other, the TDDR and wavelet filtering turned out to be the most effective methods for FC and topological analysis, as evidenced by their superior denoising ability, the best ROC, and an enhanced ability to recover the original FC pattern.

Conclusions

The findings of our study elucidate the varying impact of MA correction algorithms on brain FC analysis, which could serve as a reference for choosing the most appropriate method for future FC research. As guidance, we recommend using TDDR or wavelet filtering to minimize the impact of MA correction in brain network analysis.

An Ultralow-Power Real-Time Machine Learning Based fNIRS Motion Artifacts Detection

Due to iterative matrix multiplications or gradient computations, machine learning modules often require a large amount of processing power and memory. As a result, they are often not feasible for use in wearable devices, which have limited processing power and memory. In this study, we propose an ultralow-power and real-time machine learning-based motion artifact detection module for functional near-infrared spectroscopy (fNIRS) systems. We achieved a high classification accuracy of 97.42%, low field-programmable gate array (FPGA) resource utilization of 38354 lookup tables and 6024 flip-flops, as well as low power consumption of 0.021 W in dynamic power. These results outperform conventional CPU support vector machine (SVM) methods and other state-of-the-art SVM implementations. This study has demonstrated that an FPGA-based fNIRS motion artifact classifier can be exploited while meeting low power and resource constraints, which are crucial in embedded hardware systems while keeping high classification accuracy.

Parkinson's disease patients show delayed hemodynamic changes in primary motor cortex in fine motor tasks and decreased resting-state interhemispheric functional connectivity: a functional near-infrared spectroscopy study

Significance

People with Parkinson's disease (PD) experience changes in fine motor skills, which is viewed as one of the hallmark signs of this disease. Due to its non-invasive nature and portability, functional near-infrared spectroscopy (fNIRS) is a promising tool for assessing changes related to fine motor skills.

Aim

We aim to compare activation patterns in the primary motor cortex using fNIRS, comparing volunteers with PD and sex- and age-matched control participants during a fine motor task and walking. Moreover, inter and intrahemispheric functional connectivity (FC) was investigated during the resting state.

Approach

We used fNIRS to measure the hemodynamic changes in the primary motor cortex elicited by a finger-tapping task in 20 PD patients and 20 controls matched for age, sex, education, and body mass index. In addition, a two-minute walking task was carried out. Resting-state FC was also assessed.

Results

Patients with PD showed delayed hypoactivation in the motor cortex during the fine motor task with the dominant hand and delayed hyperactivation with the non-dominant hand. The findings also revealed significant correlations among various measures of hemodynamic activity in the motor cortex using fNIRS and different cognitive and clinical variables. There were no significant differences between patients with PD and controls during the walking task. However, there were significant differences in interhemispheric connectivity between PD patients and control participants, with a statistically significant decrease in PD patients compared with control participants.

Conclusions

Decreased interhemispheric FC and delayed activity in the primary motor cortex elicited by a fine motor task may one day serve as one of the many potential neuroimaging biomarkers for diagnosing PD.

Learning based motion artifacts processing in fNIRS: a mini review

This paper provides a concise review of learning-based motion artifacts (MA) processing methods in functional near-infrared spectroscopy (fNIRS), highlighting the challenges of maintaining optimal contact during subject movement, which can lead to MA and compromise data integrity. Traditional strategies often result in reduced reliability of the hemodynamic response and statistical power. Recognizing the limited number of studies focusing on learning-based MA removal, we examine 315 studies, identifying seven pertinent to our focus area. We discuss the current landscape of learning-based MA correction methods and highlight research gaps. Noting the absence of standard evaluation metrics for quality assessment of MA correction, we suggest a novel framework, integrating signal and model quality considerations and employing metrics like ΔSignal-to-Noise Ratio (ΔSNR), confusion matrix, and Mean Squared Error. This work aims to facilitate the application of learning-based methodologies to fNIRS and improve the accuracy and reliability of neurovascular studies.

CGAN-rIRN: a data-augmented deep learning approach to accurate classification of mental tasks for a fNIRS-based brain-computer interface

Functional near-infrared spectroscopy (fNIRS) is increasingly used to investigate different mental tasks for brain-computer interface (BCI) control due to its excellent environmental and motion robustness. Feature extraction and classification strategy for fNIRS signal are essential to enhance the classification accuracy of voluntarily controlled BCI systems. The limitation of traditional machine learning classifiers (MLCs) lies in manual feature engineering, which is considered as one of the drawbacks that reduce accuracy. Since the fNIRS signal is a typical multivariate time series with multi-dimensionality and complexity, it makes the deep learning classifier (DLC) ideal for classifying neural activation patterns. However, the inherent bottleneck of DLCs is the requirement of substantial-scale, high-quality labeled training data and expensive computational resources to train deep networks. The existing DLCs for classifying mental tasks do not fully consider the temporal and spatial properties of fNIRS signals. Therefore, a specifically-designed DLC is desired to classify multi-tasks with high accuracy in fNIRS-BCI. To this end, we herein propose a novel data-augmented DLC to accurately classify mental tasks, which employs a convolution-based conditional generative adversarial network (CGAN) for data augmentation and a revised Inception-ResNet (rIRN) based DLC. The CGAN is utilized to generate class-specific synthetic fNIRS signals to augment the training dataset. The network architecture of rIRN is elaborately designed in accordance with the characteristics of the fNIRS signal, with serial multiple spatial and temporal feature extraction modules (FEMs), where each FEM performs deep and multi-scale feature extraction and fusion. The results of the paradigm experiments show that the proposed CGAN-rIRN approach improves the single-trial accuracy for mental arithmetic and mental singing tasks in both the data augmentation and classifier, as compared to the traditional MLCs and the commonly used DLCs. The proposed fully data-driven hybrid deep learning approach paves a promising way to improve the classification performance of volitional control fNIRS-BCI.

A Machine Learning Framework for Diagnosing and Predicting the Severity of Coronary Artery Disease

Background

Although machine learning (ML)-based prediction of coronary artery disease (CAD) has gained increasing attention, assessment of the severity of suspected CAD in symptomatic patients remains challenging.

Methods

The training set for this study consisted of 284 retrospective participants, while the test set included 116 prospectively enrolled participants from whom we collected 53 baseline variables and coronary angiography results. The data was pre-processed with outlier processing and One-Hot coding. In the first stage, we constructed a ML model that used baseline information to predict the presence of CAD with a dichotomous model. In the second stage, baseline information was used to construct ML regression models for predicting the severity of CAD. The non-CAD population was included, and two different scores were used as output variables. Finally, statistical analysis and SHAP plot visualization methods were employed to explore the relationship between baseline information and CAD.

Results

The study included 269 CAD patients and 131 healthy controls. The eXtreme Gradient Boosting (XGBoost) model exhibited the best performance amongst the different models for predicting CAD, with an area under the receiver operating characteristic curve of 0.728 (95% CI 0.623-0.824). The main correlates were left ventricular ejection fraction, homocysteine, and hemoglobin (p < 0.001). The XGBoost model performed best for predicting the SYNTAX score, with the main correlates being brain natriuretic peptide (BNP), left ventricular ejection fraction, and glycated hemoglobin (p < 0.001). The main relevant features in the model predictive for the GENSINI score were BNP, high density lipoprotein, and homocysteine (p < 0.001).

Conclusions

This data-driven approach provides a foundation for the risk stratification and severity assessment of CAD.

Clinical Trial Registration

The study was registered in www.clinicaltrials.gov protocol registration system (number NCT05018715).

Optical neuroimaging and neurostimulation in surgical training and assessment: A state-of-the-art review

Introduction

Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical neuroimaging technique used to assess surgeons' brain function. The aim of this narrative review is to outline the effect of expertise, stress, surgical technology, and neurostimulation on surgeons' neural activation patterns, and highlight key progress areas required in surgical neuroergonomics to modulate training and performance.

Methods

A literature search of PubMed and Embase was conducted to identify neuroimaging studies using fNIRS and neurostimulation in surgeons performing simulated tasks.

Results

Novice surgeons exhibit greater haemodynamic responses across the pre-frontal cortex than experts during simple surgical tasks, whilst expert surgical performance is characterized by relative prefrontal attenuation and upregulation of activation foci across other regions such as the supplementary motor area. The association between PFC activation and mental workload follows an inverted-U shaped curve, activation increasing then attenuating past a critical inflection point at which demands outstrip cognitive capacity Neuroimages are sensitive to the impact of laparoscopic and robotic tools on cognitive workload, helping inform the development of training programs which target neural learning curves. FNIRS differs in comparison to current tools to assess proficiency by depicting a cognitive state during surgery, enabling the development of cognitive benchmarks of expertise. Finally, neurostimulation using transcranial direct-current-stimulation may accelerate skill acquisition and enhance technical performance.

Conclusion

FNIRS can inform the development of surgical training programs which modulate stress responses, cognitive learning curves, and motor skill performance. Improved data processing with machine learning offers the possibility of live feedback regarding surgeons' cognitive states during operative procedures.

System Derived Spatial-Temporal CNN for High-Density fNIRS BCI

An intuitive and generalisable approach to spatial-temporal feature extraction for high-density (HD) functional Near-Infrared Spectroscopy (fNIRS) brain-computer interface (BCI) is proposed, demonstrated here using Frequency-Domain (FD) fNIRS for motor-task classification. Enabled by the HD probe design, layered topographical maps of Oxy/deOxy Haemoglobin changes are used to train a 3D convolutional neural network (CNN), enabling simultaneous extraction of spatial and temporal features. The proposed spatial-temporal CNN is shown to effectively exploit the spatial relationships in HD fNIRS measurements to improve the classification of the functional haemodynamic response, achieving an average F1 score of 0.69 across seven subjects in a mixed subjects training scheme, and improving subject-independent classification as compared to a standard temporal CNN.

Motion artifacts removal and evaluation techniques for functional near-infrared spectroscopy signals: A review

With the emergence of an increasing number of functional near-infrared spectroscopy (fNIRS) devices, the significant deterioration in measurement caused by motion artifacts has become an essential research topic for fNIRS applications. However, a high requirement for mathematics and programming limits the number of related researches. Therefore, here we provide the first comprehensive review for motion artifact removal in fNIRS aiming to (i) summarize the latest achievements, (ii) present the significant solutions and evaluation metrics from the perspective of application and reproduction, and (iii) predict future topics in the field. The present review synthesizes information from fifty-one journal articles (screened according to three criteria). Three hardware-based solutions and nine algorithmic solutions are summarized, and their application requirements (compatible signal types, the availability for online applications, and limitations) and extensions are discussed. Five metrics for noise suppression and two metrics for signal distortion were synthesized to evaluate the motion artifact removal methods. Moreover, we highlight three deficiencies in the existing research: (i) The balance between the use of auxiliary hardware and that of an algorithmic solution is not clarified; (ii) few studies mention the filtering delay of the solutions, and (iii) the robustness and stability of the solution under extreme application conditions are not discussed.

Special Section Guest Editorial: Computational Approaches for Neuroimaging

This guest editorial provides an introduction to the Special Section on Computational Approaches for Neuroimaging.

Deep learning in fNIRS: a review

Significance: Optical neuroimaging has become a well-established clinical and research tool to monitor cortical activations in the human brain. It is notable that outcomes of functional near-infrared spectroscopy (fNIRS) studies depend heavily on the data processing pipeline and classification model employed. Recently, deep learning (DL) methodologies have demonstrated fast and accurate performances in data processing and classification tasks across many biomedical fields. Aim: We aim to review the emerging DL applications in fNIRS studies. Approach: We first introduce some of the commonly used DL techniques. Then, the review summarizes current DL work in some of the most active areas of this field, including brain-computer interface, neuro-impairment diagnosis, and neuroscience discovery. Results: Of the 63 papers considered in this review, 32 report a comparative study of DL techniques to traditional machine learning techniques where 26 have been shown outperforming the latter in terms of the classification accuracy. In addition, eight studies also utilize DL to reduce the amount of preprocessing typically done with fNIRS data or increase the amount of data via data augmentation. Conclusions: The application of DL techniques to fNIRS studies has shown to mitigate many of the hurdles present in fNIRS studies such as lengthy data preprocessing or small sample sizes while achieving comparable or improved classification accuracy.

Review of Applications of Near-Infrared Spectroscopy in Two Rare Disorders with Executive and Neurological Dysfunction: UCD and PKU

Studying rare diseases, particularly those with neurological dysfunction, is a challenge to researchers and healthcare professionals due to their complexity and small population with geographical dispersion. Universal and standardized biomarkers generated by tools such as functional neuroimaging have been forged to collect baseline data as well as treatment effects. However, the cost and heavily infrastructural requirement of those technologies have substantially limited their availability. Thus, developing non-invasive, portable, and inexpensive modalities has become a major focus for both researchers and clinicians. When considering neurological disorders and diseases with executive dysfunction, EEG is the most convenient tool to obtain biomarkers which can correlate the objective severity and clinical observation of these conditions. However, studies have also shown that EEG biomarkers and clinical observations alone are not sensitive enough since not all the patients present classical phenotypical features or EEG evidence of dysfunction. This article reviews disorders, including two rare disorders with neurological dysfunction and the usefulness of functional near-infrared spectroscopy (fNIRS) as a non-invasive optical modality to obtain hemodynamic biomarkers of diseases and for screening and monitoring the disease.

Deep-learning informed Kalman filtering for priori-free and real-time hemodynamics extraction in functional near-infrared spectroscopy

Separation of the physiological interferences and the neural hemodynamics has been a vitally important task in the realistic implementation of functional near-infrared spectroscopy (fNIRS). Although many efforts have been devoted, the established solutions to this issue additionally rely on priori information on the interferences and activation responses, such as time-frequency characteristics and spatial patterns, etc., also hindering the realization of real-time. To tackle the adversity, we herein propose a novel priori-free scheme for real-time physiological interference suppression. This method combines the robustness of deep-leaning-based interference characterization and adaptivity of Kalman filtering: a long short-term memory (LSTM) network is trained with the time-courses of the absorption perturbation baseline for interferences profiling, and successively, a Kalman filtering process is applied with reference to the noise prediction for real-time activation extraction. The proposed method is validated using both simulated dynamic data and in-vivo experiments, showing the comprehensively improved performance and promisingly appended superiority achieved in the purely data-driven way.