Functional Near-Infrared Spectroscopy is an optical brain monitoring technique which uses NIRS to perform functional neuroimaging. It uses near-infrared light for measuring brain activity and to estimate the cortical hemodynamic activity in the brain due to motor activity. Functional NIRS measures the changes in oxygen levels in oxygenated and deoxygenated hemoglobin by optical absorption. One of the main challenges in the analysis of fNIRS signals is the signal degradation due to the interference from noise and artifacts from multiple sources. In this context, this research aims to analyze the connectivity between different regions of the brain using graph theory and hence the geometrical association of brain networks in terms of functional parameters. In this study, the impact of two noise removal processes (CBSI and TDDR), along with two types of correlation fNIRS such as Pearson's Correlation (PC), and Cross Correlation (CC) and various whole-brain network architectures on the reproducibility of graph measurements for individual participants has been carefully examined for different densities ranging from 5% to 50%.The graph measures' repeatability at the individual level was studied using the test-retest variability (TRT) The test-retest variability for global measurements in binary networks was substantially large at low densities, regardless of the noise removal method or the kind of correlation. Very low test –reset values are observed for weighted networks and great reproducibility for measures of the entire graph. When comparing the test-retest values for various methods, the kind of correlation, the absolute value of the correlation, and the weight calculation method on the raw correlation value all had significant major effects. Based on a weighted network with the absolute cross correlation functioning as the weight, this study revealed that normalized global graph measurements were reliable. The node definition techniques that were utilized to remove noise were not essential for the normalized graph measures to be reproducible.

Recent research in the field of cognitive motor action decoding focuses on data acquired from Functional Near-Infrared Spectroscopy (fNIRS) and its analysis. This research aims to classify two different motor activities, namely, mental drawing (MD) and spatial navigation (SN), using fNIRS data from non-motor baseline data and other motor activities. Accurate activity detection in non-stationary signals like fNIRS is challenging and requires complex feature descriptors. As a novel framework, a new feature generation by fusion of wavelet feature, Hilbert, symlet, and Hjorth parameters is proposed for improving the accuracy of the classification. This new fused feature has statistical descriptor elements, time-localization in the frequency domain, edge feature, texture features, and phase information to detect and locate the activity accurately. Three types of independent component analysis, including FastICA, Picard, and Infomax were implemented for preprocessing which removes noises and motion artifacts. Two independent binary classifiers are designed to handle the complexity of classification in which one is responsible for mental drawing (MD) detection and the other one is spatial navigation (SN). Four different types of algorithms including nearest neighbors (KNN), Linear Discriminant Analysis (LDA), light gradient-boosting machine (LGBM), and Extreme Gradient Boosting (XGBOOST) were implemented. It has been identified that the LGBM classifier gives high accuracies—98% for mental drawing and 97% for spatial navigation. Comparison with existing research proves that the proposed method gives the highest classification accuracies. Statistical validation of the proposed new feature generation by the Kruskal–Wallis H-test and Mann–Whitney U non-parametric test proves the reliability of the proposed mechanism.

Alzheimer’s disease (AD) is a neuro-degenerative disorder categorized by the progressive decline in cognitive function. Prompt and accurate detection is essential for both intervention and management of the disease. We present a Hybrid CNN-BiLSTM Framework with Adaptive Learning for early detection of AD in this work. Firstly, the model utilizes Convolutional Neural Networks (CNNs) for extraction of features from medical imaging data, capturing the disease-related spatial patterns. Then, a Bidirectional Long Short-Term Memory (Bi-LSTM) network is used to learn the sequential dependencies from the extracted features. To optimize the learning process and improve the performance of the classification, an adaptive learning mechanism is implemented that enables dynamic adjustment of various model parameters throughout the training phase. The evaluation of the proposed framework on several benchmark datasets shows that it outperforms traditional deep learning models in terms of accuracy and robustness. Early stages of AD can be diagnosed in this way, which may help in clinical decision-making and improve patients’ outcomes.