Neurophotonics最新文献

Significance: Functional near-infrared spectroscopy (fNIRS) enables portable and noninvasive monitoring of cerebral hemodynamics, but hemodynamic changes originating from extracerebral tissues may influence the signals. To avoid this, short-channel regression (SCR) is widely used, yet physical short-separation detectors are not always available or optimally positioned due to hardware limitations or the experimental setup. In such cases, a virtual, data-driven alternative to physical short-channel detectors may be a viable solution.

Aim: We aimed to (i) develop a transformer-based deep learning model to predict short-separation optical density (OD) signals from long-separation channels and (ii) evaluate whether these virtual signals enable effective SCR.

Approach: We trained the model on a resting-state fNIRS dataset (69 subjects) with paired short- and long-separation recordings. Dual-wavelength OD signals in segmented time windows were used as input for a transformer encoder trained to reconstruct the extracerebral hemodynamic component measured by short channels. Model performance was evaluated using 3 independent datasets: a holdout subset of the same resting-state dataset (23 subjects), a second dataset acquired using a different system (40 subjects), and a task-based finger-tapping dataset (4 subjects). A wavelet coherence-based channel rejection step was optionally applied during preprocessing. Predictions were evaluated using signal similarity metrics (mean squared error [MSE], normalized MSE [NMSE], and Pearson correlation [ $r$ ]) and denoising efficacy (residual variance after regression).

Results: Predicted short-channel signals showed high correspondence with ground-truth measurements in OD (median $r=0.70$ and $\mathrm{NMSE}=0.047$ ) and concentration data (up to $r=0.67$ ). When used for SCR, virtual regressors effectively denoise long-channel data. Performance was robust across all datasets, with greater accuracy when low-coherence channels were excluded. In motor task blocks, predicted regressors preserved task-evoked activations and reduced residual variance.

Conclusion: Transformer-based models accurately reconstruct extracerebral hemodynamic signals from long-separation fNIRS data, providing a virtual alternative to physical short channels and supporting standardized, hardware-independent preprocessing.

Significance: Myelin breakdown is prevalent in a range of neurodegenerative diseases, aging, and following various forms of trauma. Yet, current imaging techniques have limited capacity for large-scale study of myelin structural damage. A high-throughput, quantitative imaging method would greatly enhance our understanding of myelin degradation in different contexts.

Aim: We aim to establish birefringence microscopy (BRM) as a high-throughput, label-free imaging technique for large-scale, quantitative assessment of myelin pathology in post-mortem brain tissue. BRM has the capacity to provide rapid myelin imaging, which will provide information complementary to other myelin imaging techniques.

Approach: BRM enables label-free structural imaging of myelin with high spatial resolution. We leverage the high-throughput imaging capability of BRM to characterize the distribution of myelin pathology in a rhesus monkey model of cortical injury across the corpus callosum. This framework is applied at two different post-injury survival times (6 and 12 weeks).

Results: We validate BRM for label-free structural imaging of myelin pathology across large regions of tissue (within the corpus callosum) using a fluorescent myelin stain and several immunohistochemical labels. Next, we train and validate a deep learning-based object detection network for automated identification of myelin pathology, using BRM, in the corpus callosum of monkeys with an induced cortical lesion. BRM, paired with deep learning, revealed significantly higher myelin damage through the corpus callosum, resulting from the lesion, in 6-week recovery monkeys compared with 12-week recovery and age-matched controls ( $P<0.01$ ). There was no significant difference between 12-week recovery monkeys and age-matched controls.

Conclusions: BRM enables large-scale assessment of myelin structural alterations in post-mortem brain tissue. When combined with deep-learning object detection, BRM enables rapid quantification of myelin damage in the corpus callosum after cortical injury. With proper training, this can be extended to study structural changes in other diseases and regions such as Alzheimer's disease and chronic traumatic encephalopathy as well as normal aging.

Significance: Rapid acquisition of high-resolution volumetric images has been critical to effectively monitor dynamic biological processes in vivo, yet it faces tradeoffs between image resolution, penetration depth, and imaging speed. These limitations hinder the ability to study rapid neurophysiological events such as cerebrovascular dynamics and cellular activity, highlighting the need for advanced high-speed 3D imaging system.

Aim: To address these challenges in volumetric imaging performances, we aimed to develop a high-speed volumetric imaging system capable of resolving fast biological dynamics with minimal compromise in spatial resolution or imaging depth.

Approach: We devised a rapid axially scanned and de-scanned (RASAD) scheme by integrating a TAG lens (tunable acoustic gradient index of refraction lens) into a line-scan confocal microscope. The TAG lens enabled axial (depth) scanning frequency at 70 kHz, allowing 3D projection imaging at rates up to 200 Hz with a detection depth of $135\mu m$ while minimally sacrificing the image quality (i.e., a lateral resolution of $\sim 2.6\mu m$ ).

Results: We validated its performance through in vitro imaging of spontaneously contracting cardiomyocyte aggregates, capturing real-time calcium transients and synchronized contractions, and through in vivo imaging of the mouse cortical tissue, where volumetric acquisition over a $450\times 450\times 100{\mu m}^3$ region enabled quantification of blood flow velocities up to $3.64\mathrm{mm}/s$ across various vessel types.

Conclusions: The RASAD system enables high-speed, high-resolution 3D imaging of dynamic biological processes, providing a valuable tool for advancing studies of neurophysiological mechanisms and biomedical applications.

Significance: Myelin degradation is a critical yet understudied pathological feature in neurodegenerative disorders. Manual detection of myelin defects in volumetric microscopy images is prohibitively time-consuming, limiting large-scale studies. There is a need for rapid, accurate, and scalable defect-detection methods to accelerate advances in the field.

Aim: We aim to develop and evaluate a human-in-the-loop deep learning approach to accelerate myelin defect detection.

Approach: We imaged brain tissue samples from the dorsolateral prefrontal cortex from 15 subjects (i.e., five controls, five Alzheimer's disease, and five chronic traumatic encephalopathy) using RGB circular crossed-polarized birefringence microscopy. We created a dataset of 5600 manually annotated myelin defects and trained a YOLOv8-based defect detection model with iterative expert verification.

Results: Our approach achieved 0.85 mAP@50 and reduced analysis time from 8 h to 33 min per $1{\mathrm{mm}}^2$ of tissue while maintaining high accuracy for disease comparison studies. The method can process complete 3D volumetric images up to 300 GB, enabling comprehensive assessment across large tissue volumes.

Conclusions: This approach effectively streamlines myelin defect detection and can enable the scale up of myelin degradation studies in neurodegenerative disorders.

Significance: Understanding the regional functionality during early development has significance across various domains such as developmental disorders.

Aim: We aim to investigate the impact of early bilingual exposure on infant brain activity at the age of 4 months and to determine whether differences exist in the activity of specific brain regions between monolingual and bilingual infants in rest.

Approach: To reach that aim, we utilize a combination of functional near-infrared spectroscopy and effective connectivity analysis (EC), integrated with our previously proposed graph construction method, importance of channel based on EC (ICEC), to assess neural mechanisms underlying bilingualism in infancy. Importantly, we represent a secondary analysis of a publicly available dataset.

Results: Employing group-level analysis techniques, our findings reveal that bilingual experience is associated with anatomically specific rather than widespread alterations in EC. Differences were most pronounced in the superior frontal gyrus, superior temporal gyrus, and opercular regions, with the frontal and temporal cortices primarily acting as sources and the operculum functioning as both sources and sinks. Notably, bilingual infants exhibited a gradual increase in connectivity within the rolandic operculum during rest. Temporal analyses further indicated that early rest was marked by stronger inflow into right frontal-opercular hubs, whereas later rest showed a redistribution toward temporal and opercular regions with increased outflow. Together, these results provide evidence that bilingual exposure reorganizes infant brain connectivity in anatomically specific and temporally dynamic ways.

Conclusion: These findings provide novel insights into the neurobiological foundations of early bilingual exposure, highlighting distinct patterns of EC in bilingual infants.

Dr. Anna Roe, Director of Translational Neuroscience at the Nathan Kline Institute for Psychiatric Research and Professor of Psychiatry and Neuroscience at New York University, discusses her journey as a neuroscientist, highlighting breakthroughs in brain plasticity, imaging, and the value of curiosity and mentorship.

Significance: Accurate and consistent probe placement is crucial in functional near-infrared spectroscopy (fNIRS) and electroencephalogram (EEG) experiments, especially in longitudinal and group-based studies. Both operator experience and subject head shape variability can affect placement accuracy.

Aim: We aim to develop an easy-to-use software, NeuroNavigatAR (NNAR), utilizing augmented reality (AR) and machine learning to estimate and display in real-time the subject's cranial and head landmarks to guide consistent headgear placement.

Approach: By applying a facial recognition toolbox to the image frames extracted from a video camera, we can obtain and continuously track subject-specific three-dimensional facial landmarks. Separately, we have precomputed a robust linear transformation between facial landmarks and key cranial landmarks, including nasion and preauricular points, using a large public head-model library consisting of over 1000 subjects. These allow us to rapidly estimate subject-specific cranial landmarks and subsequently render atlas-derived head landmarks to the subject's camera stream.

Results: An open-source graphical user interface implementing this AR system has achieved a speed of 15 frames per second using a laptop. A median 10-20 position error of 1.52 cm was found when using a general adult atlas and is further reduced to 1.33 cm and 0.75 cm when using age-matched atlas models and subject-specific head surfaces, respectively. NNAR demonstrated consistent head-landmark prediction errors across repeated measurement sessions; there is also no statistically significant difference in accuracy across age groups.

Conclusions: NNAR is an easy-to-use AR headgear placement monitoring tool that is expected to significantly enhance consistency and reduce setup time for fNIRS and EEG probe donning across a wide range of studies.

Significance: Accurate modeling of light diffusion in the human brain is crucial for applications in optogenetics and spectroscopic diagnostic techniques. White matter tissue is composed of myelinated axon bundles, suggesting the occurrence of enhanced light diffusion along their local orientation direction, which however has never been characterized experimentally. Existing diffuse optics models assume isotropic properties, limiting their accuracy.

Aim: We aim to characterize the anisotropic scattering properties of human white matter tissue by directly measuring its tensor scattering components along different directions and correlating them with the local axon fiber orientation.

Approach: Using a time- and space-resolved setup, we image the transverse propagation of diffusely reflected light across two perpendicular directions in a post-mortem human brain sample. Local fiber orientation is independently determined using light sheet fluorescence microscopy and two-photon fluorescence microscopy.

Results: The directional dependence of light propagation in organized myelinated axon bundles is characterized via Monte Carlo simulations accounting for a tensor scattering coefficient, revealing a weaker scattering rate parallel to the fiber orientation. The effects of white matter anisotropy are further assessed by simulating a typical time-domain near-infrared spectroscopy measurement in a four-layer human head model.

Conclusions: We provide a first characterization of the anisotropic scattering properties in post-mortem human white matter, highlighting its direct correlation with axon fiber orientation, and opening the way to the realization of quantitatively accurate anisotropy-aware human head 3D meshes for diffuse optics applications.

Background: Functional near-infrared spectroscopy (fNIRS) has recently gained momentum as a reliable and accurate tool for assessing brain states based on the vascular response to neural activity. This increase in popularity is due to its robustness to movement, non-invasive nature, portability, and user-friendly application. However, compared with other hemodynamic functional brain-imaging methods such as functional magnetic resonance imaging (fMRI), fNIRS is constrained by its limited spatial resolution and coverage with a particularly limited penetration depth. In addition, due to comparatively fewer methodological advancements, the performance of fNIRS-based brain-state classification still lags behind more prevalent methods such as fMRI.

Methods: We introduce a classification approach grounded in Riemannian geometry for the classification of kernel matrices, leveraging the temporal and spatial relationships between channels and the inherent duality of fNIRS signals, specifically oxygenated and deoxygenated hemoglobin. For the Riemannian-geometry-based models, we compared different kernel matrix estimators and two classifiers: Riemannian Support Vector Classifier and Tangent Space Logistic Regression. These were benchmarked against four models employing traditional feature extraction methods. Our approach was tested on seven participants in two brain-state classification scenarios based on the same fNIRS dataset: an eight-choice classification, which includes seven established plus an individually selected imagery task, and a two-choice classification of all possible 28 two-task combinations.

Results: This approach achieved a mean eight-choice classification accuracy of 65%, significantly surpassing the mean accuracy of 42% obtained with traditional methods. In addition, the best-performing model achieved an average accuracy of 96% for two-choice classification across all task combinations, compared with 78% with traditional models.

Conclusion: To our knowledge, we are the first to demonstrate that the proposed Riemannian-geometry-based classification approach is both powerful and viable for fNIRS data, substantially increasing the accuracy in binary and multi-class classification of brain activation patterns.