Home » Posts tagged 'AI' (Page 3)
Tag Archives: AI
Fr-019-Comparison of Photogrammetry and 3D Scanning Methods to Traditional Digitizer for Localization of fNIRS Channels
Functional near-infrared spectroscopy (fNIRS) requires an accurate spatial localization of fNIRS channel. We investigated the agreement between the fNIRS channel localization methods of photogrammetry and 3D scanning compared to the conventional digitizer. MNI coordinates of simulated fNIRS channels covering the whole head were determined using all three methods. The Euclidian distance between digitizer and photogrammetry, as well as 3D scan, were negligible (median error = 6.96 and 7.07 mm, respectively), demonstrating the utility of photogrammetry as a low-cost and easy-to-use head digitization method, as well as 3D scan method. Continue reading
Sa-015-Uncontrolled False Positives in Filtered fNIRS: A Critical Examination
Filtering is a common preprocessing method in functional near infrared spectroscopy (fNIRS) but can lead to assumption violations in the statistical methods used in fNIRS research. Simulated data with a variety of noise conditions tested whether these assumption violations lead to uncontrolled false positive rate (FPR). Preprocessing with a filter led to a loss of false positive control for all tested statistical methods and measured an FPR much higher than the α-level. We conclude that filtering should be avoided in combination with many common statistical approaches in fNIRS. Continue reading
Sa-053-Reconstructing fNIRS signal with Generative Deep Learning Model
Functional near infrared spectroscopy (fNIRS) is an emerging non-invasive neuroimaging technique with increasing applications across various fields. However, factors like head motion and physiological noise often decrease the quality of fNIRS signals. To solve this issue, we proposed a generative deep learning model for reconstructing fNIRS time series with poor quality. Our results showed accurate reconstruction of poor time series in fNIRS signal while preserving functional connectivity consistency between channels. These findings underscore the potential of generative deep learning in restoring poor fNIRS signals, thereby providing high-quality data for clinical diagnosis and brain research. Continue reading
Sa-092-Classifying the Prefrontal Cortex Reasoning Process Using CLEVR Cognitive Tasks
The prefrontal cortex (PFC) is widely recognized for its pivotal role in relational reasoning. Existing PFC reasoning diagnostic tests show limitations due to learning effect and failure to control for quantitative difficulty. Addressing this challenge, this study proposes a novel cognitive reasoning task built for fNIRS based on AI datasets for computer vision. Tasks were categorized based on their relational or non-relational nature, as well as their level of quantitative difficulty. NIRSIT-Lite was employed to analyze the PFC activation patterns. Significant and consistent activation was observed during relational tasks irrespective of the task difficulty. Continue reading
Sa-040-Assessing Visual Cognitive Motivation through Machine Learning applied to fNIRS data
Motivation plays a fundamental role in the field of psychology, as it influences the cognitive and behavioral dynamics of individuals. This study aimed to classify subjects' motivation for remembering a visual stimulus. Relying on fNIRS features, a Random Forest classifier was developed to distinguish participants who did not want to remember (NWR) a visual stimulus from those who actively wanted to remember (WR) it. On the test data, 71.80% accuracy was reached, whereas during cross-validation, a mean accuracy of 67.32% (± 3.67%) was obtained. This study offers new perspectives for applications where motivation detection is crucial, such as brain-computer interfaces. Continue reading
Su-043-Feasibility of Simultaneous Near Whole-Head fNIRS and Physiological Measurements
Non-cerebral and physiological systemic activity contaminates fNIRS data and could account for the considerable across-subject variability reported in the literature. Aiming to evaluate global systemic physiology's influence, we have designed an fNIRS task battery combined with simultaneous peripheral physiological measurements. Preliminary data quality results validate the feasibility of this multimodal setup toward a larger dataset to be published and openly available for the fNIRS community. Continue reading
Fr-007-The effect of diverse hair and skin properties on fNIRS signal quality
fNIRS signal quality is sensitive to hair and skin properties. In this work, we aim to quantify this by running an fNIRS study in a large cohort of diverse subjects with a wide range of skin and hair properties. Our multiple linear regression analysis revealed that hair characteristics, specifically cumulative hair thickness and hair color, as well as skin pigmentation explain the variability in several signal quality metrics for fNIRS such as Scalp Coupling Index and Peak Spectral Power. Continue reading
Fr-031-Investigating the effect of channel pruning on fNIRS data collected from Gambian children aged 5-60 months
Channel pruning to combat poor scalp-optode coupling is a crucial fNIRS preprocessing step. QT-NIRS allows users to perform pruning efficiently, however there is no consensus on threshold values for the tool parameters in data collected in infants and toddlers. We propose a metric to assess the impact of QT-NIRS parameters Scalp Coupling Index (SCI) and Peak Spectral Power (PSP) on data retention and quality. We demonstrate its application in data collected from Gambian participants tested at 6 time points between 5- and 60 months of age, and its potential utility in guiding fNIRS users working with infant and toddler participants. Continue reading
Fr-030-Characterising infant neurodevelopment in diverse settings using functional change point analysis
Habituation and novelty detection (HaND) of learning. The BRIGHT Study has already investigated infant HaND brain responses via measured changes in the amplitude of evoked hemodynamic responses. However, characterisation of response trajectories would provide further insight into early neurodevelopment. We identify significant changes to the mean structure of the haemodynamic responses using functional change point analysis. We then compare volumetric calculations based on these change points to investigate variation in both the timing and magnitude of infants’ hemodynamic responses during the HaND task. Continue reading
Fr-021-Investigating the NIRS-derived cytochrome-c-oxidase signal using different system wavelength combinations and processing techniques with a diffusion-based model
The effect of using different numbers and combinations of wavelengths, and varying parameter errors on cytochrome-c-oxidase (CCO) and haemoglobin measurement errors were investigated using a diffusion-based simulation. The number of measurement wavelengths was varied from 5-200, and uncertainties in the differential pathlength factor (DPF) and extinction coefficients of up to 10% were introduced to determine their effect on NIRS-derived concentration changes. Deviations from the true concentration values were found to be below 10% in ideal conditions. However when parameter errors are increased to more realistic values, CCO errors are highly wavelength dependent and orders of magnitude larger. Continue reading