• Carbonatite complex in SGT offers high REE exploration potential. • EO-1 Hyperion data used for REE-fertile lithology mapping. • Applied SDM with PCA, ICA, MNF, BRC, and SVM methods. • ICA gave highest separability (JM > 1.9); SVM accuracy 85.56%. • Combines spectroscopy and ML for complex terrain mineral targeting. The Proterozoic alkaline carbonatite complex, lies along the Samalpatti shear zone, is linked to a post-collisional rift setting in the Southern Granulitic Terrain (SGT), provides a geologically intriguing and economically prospective terrain for rare earth element (REE) exploration. This study involves a multiproxy approach by integrating the hyperspectral remote sensing, machine learning, and field validation techniques to delineate the REE fertile lithology units using the EO-1 Hyperion imagery. The pre-processed dataset was subjected to noise reduction and dimensionality reduction using spectral dispersion matching (SDM) methods. SDM was performed in 3 steps; initially, noise reduction algorithms such as principal component analysis (PCA), independent component analysis (ICA), minimum noise fraction (MNF), and band ratio combinations (BRC) were applied to enhance data quality. This was followed by correlation-based feature selection using support vector machines (SVM), focusing on spectral behaviour. Subsequently, mineralogical characteristics were integrated and validated to emphasize their spectrochemical properties. Among the reduction algorithms, ICA achieved the highest spectral class separability, as confirmed by Jeffries–Matusita distance analysis, with values >1.9 for key lithological pairs. The correlation-based feature selection was performed with a radial basis function (RBF) kernel, yielding an overall accuracy of 85.56% and a Kappa coefficient of 0.80. The multiproxy approach using SDM highlights the efficacy of imaging spectroscopy combined with advanced classification techniques in complex lithological terrains and offers a scalable framework for mineral exploration targeting REE-fertile zones.

Overweight-related alterations in brain structural covariance networks and their potential impact on cognitive function in subjective cognitive decline.

A method for sparse and robust independent component analysis

Beyond CRBN and VHL: Parkin driven lysine-based ternary complexes for selective β3-Tubulin degradation.

We investigated the pain connectome in patients with diabetic peripheral neuropathy (DPN), particularly those with neuropathic pain, by characterizing functional interactions within the traditional pain matrix and across large-scale brain networks. Sixty-two patients with type 2 diabetes mellitus (25 without DPN, 20 with painless DPN, and 17 with painful DPN) and 33 healthy controls were enrolled in this study. All participants underwent resting-state functional MRI, and region of interest (ROI)-to-ROI functional connectivity (FC) analysis was performed within predefined core regions of the pain matrix, supplemented by group independent component analysis to evaluate interactions between pain related networks. Partial correlation analyses were conducted to assess associations between connectivity strength and pain-related measures, peripheral nerve function, and neuropsychological outcomes. Compared with HCs, patients with painful DPN showed presented decreased FC between the amygdala and the anterior cingulate cortex (ACC), medial prefrontal cortex (mPFC), thalamus, and periaqueductal gray (PAG), as well as between the thalamus and PAG. Higher peak pain scores and lower Douleur Neuropathique 4 questionnaire (DN4) questionnaire scores were associated with lower FC in these pathways (all P < 0.05). Lower FC in these pathways (except FC between the amygdala and the PAG) was also associated with poorer peripheral nerve functions (all P < 0.05). Higher anxiety levels were associated with increased FC between the right amygdala and the right ACC (r = 0.600, P = 0.023). Patients with painful DPN exhibit distinct FC alterations, and the amygdala may play a key role in pain perception and emotional regulation in the neuropathic pathophysiology of DPN.

ABSTRACT Objectives Schizophrenia is a neuropsychiatric disorder that affects emotional, behavioral, and brain functions that can be tracked using electroencephalography (EEG). This research conducts a comparative evaluation of deep learning models utilizing EEG time-frequency and spectral analysis methods to automate schizophrenia detection. Methods Two compatible EEG datasets were merged, yielding a total of 934 EEG samples from 237 subjects (121 schizophrenia patients and 116 controls). Independent Component Analysis (ICA) was applied for signal decomposition. By deriving time-frequency representations using Continuous Wavelet Transform (CWT) and Fast Fourier Transform (FFT), scalogram and spectral inputs for deep learning models were obtained. Six architectures, including CNN variants, CNN-FFT, CNN-ELM, CNN-LSTM, ResNet Transfer, and a Transformer-based model, were evaluated with data augmentation and class balancing to improve robustness. Results While variations in numerical performance were observed across models, statistical analysis indicated that these differences were not significant. Discussion The study presents results that underscore the benefits of combining time-frequency analysis with deep learning for EEG-based schizophrenia diagnosis, especially via spectral feature extraction in CNN architectures. Furthermore, it provides insights consistent with known neurophysiological patterns in schizophrenia, emphasizing the significance of model interpretability for clinical translation. Future research will focus on the integration of multimodal neuroimaging and the enhancement of explainability frameworks to augment diagnostic reliability.

EEG is a non-invasive technique for recording brain activity, valued for its high temporal resolution and low cost. Independent Component Analysis (ICA), a blind source separation method, effectively decomposes EEG signals into independent components. Widely used in both medical research and brain-computer interfaces (BCIs), ICA has become essential for improving EEG data quality. As demand grows for real-time and robust neural signal processing, advanced methods like ICA remain crucial for advancing EEG applications. This paper investigated the application of ICA in EEG signal processing. The research first reviewed the core principles of ICA, including its mathematical model and algorithms such as Infomax and FastICA. The results clearly demonstrated that ICA can effectively decompose raw EEG signals into statistically independent components. This enables the identification and removal of artifacts while preserving neural information. It also showed that while ICA is a mature method for EEG preprocessing, challenges such as source number determination and real-time processing bottlenecks still exist. At the end of the research, future developing trends of ICA and its integration with other methods are discussed.

Parkinson’s disease (PD) may benefit from non-pharmacological motor–cognitive rehabilitation, but sensitive neuroimaging markers of training-related brain changes remain limited. This study investigated whether 4 weeks of daily Quadrato Motor Training (QMT) modulate resting-state functional connectivity (FC) in PD and secondarily explored whether whole-brain radiomic features derived from T1-weighted and fractional anisotropy (FA) images could detect pre–post differences over this short intervention interval. Fifty patients with idiopathic PD were randomized to QMT or a SHAM repetitive stepping condition, and 48 completed the protocol (25 SHAM, 23 QMT). MRI was acquired at baseline and after 4 weeks and included resting-state fMRI, 3D T1-weighted imaging, and diffusion-derived FA maps. Resting-state fMRI was analyzed using independent component analysis and dual regression, whereas an IBSI-compliant radiomics workflow and machine-learning models were used for exploratory scan-level classification. Compared with baseline, the SHAM group showed reduced synchronization across several resting-state networks, whereas the QMT group showed increased synchronization in the right sensorimotor and frontoparietal networks and no significant reductions. Between-group analyses showed lower delta-FC in SHAM than QMT in the cerebellar and sensorimotor networks. In contrast, radiomics showed limited discrimination between pre- and post-QMT scans; the best model achieved a ROC-AUC of 0.65 with near-chance accuracy, and no selected predictor remained significant after multiple-comparison correction. These findings suggest that QMT may support short-term functional network stability or task-relevant reorganization in PD relative to the SHAM condition, whereas whole-brain structural radiomics appears less sensitive for detecting early training-related effects in this setting.

Independent component analysis of brain network alterations associated with cognitive impairment in coal workers' pneumoconiosis.

The fetal electrocardiogram (FECG) signal provides valuable insights into the fetal cardiac status during pregnancy. As the maternal abdominal ECG (AECG) signal is influenced by various sources, including the maternal ECG (MECG) signal, it becomes challenging to separate the FECG signal from the AECG effectively. In this paper, we introduce a pipeline centered on Independent Component Analysis (ICA) for the extraction of FECG from a single-channel abdominal recording, termed "Single Channel Time Delay ICA (SCTD-ICA)". A notable limitation of ICA is its requirement for multi-channel data. To overcome this constraint, the proposed pipeline uses the time delay method to map single-channel data into multidimensional data. Subsequently, the multidimensional data is employed as input for the ICA algorithm. Ultimately, the analysis of the power spectrum leads to the automated identification of the fetal component. The effectiveness of the proposed pipeline is assessed on two real datasets: Abdominal and Direct Fetal Electrocardiogram Database (ADFECGDB) and Set-A of 2013 PhysioNet/Computing in Cardiology Challenge Database (PCDB). F1 metrics for fetal QRS detection in ADFECGDB and PCDB are obtained at 96.14% and 95.76%, respectively. In comparison to state-of-the-art methods, the proposed pipeline demonstrates comparable performance with approaches based on deep learning. As a result, the proposed single-channel pipeline is appropriate for continuous maternal and fetal health monitoring.

Decreased functional connectivity in post-COVID syndrome patients with high neuroinflammatory activity.

Explainable machine learning algorithm for classifying resting-state functional MRI in amyotrophic lateral sclerosis.

Disordered connectivity configuration of triple-network model and visual-network in tobacco use disorder.

This study aimed to explore dynamic functional network connectivity (dFNC) differences between narcolepsy type 1 (NT1), idiopathic hypersomnia (IH), and healthy controls (HCs), and evaluate the potential of dFNC as a neurobiological marker for differentiating these hypersomnolent disorders. We recruited 50 drug-naive NT1 patients, 31 IH patients, and 50 HCs. Resting-state fMRI data were acquired, and intrinsic connectivity networks (ICNs) were identified using group independent component analysis (ICA), yielding 10 networks (e.g., visual network [VIN], auditory network [AUN], sensorimotor network [SMN], default mode network [DMN]). dFNC was analysed via sliding-window and k-means clustering to identify recurring functional connectivity states, and temporal properties (fractional windows, mean dwell time [MDT]) were compared across groups. Machine learning models (support vector machine, random forest [RF], logistic regression) were constructed using state-specific functional connectivity (FC) features to distinguish NT1 and IH. Five distinct FNC states were identified. State II (39% of windows, sparse connectivity with strengthened DMN/SMN/VIN coupling) was more prevalent in NT1 (47.68% ± 34.5%) than in IH (37.07% ± 28.73%) or HCs (31.32% ± 23.67%). Conversely, State I (33% of windows, sparse ICN connectivity) was less frequent in NT1 (13.24% ± 22.04%) versus IH (39.14% ± 35.92%) and HCs (49.28% ± 30.42%). NT1 also showed longer MDT in State II and shorter MDT in State I compared to IH and HCs (p < 0.05, ANOVA with post hoc tests FDR corrected). FC features in State I and II (notably AUN-VIN and SMN-VIN) effectively distinguished NT1 and IH, with the RF model achieving an AUC of 0.9 in State II. These findings reveal distinct dFNC patterns in NT1 and IH, reflecting divergent perturbations in sleep-wake regulatory circuits, particularly involving VIN, which may underpin their neurobiological heterogeneity. dFNC holds promise as a biomarker for differentiating these disorders, with VIN-centered connectivity emerging as a key discriminative feature.

EEG-AI: An agentic system for AI-assisted semi-automated EEG preprocessing and artifact removal.

Facial recognition technology plays a crucial role in modern biometric authentication, surveillance systems, and secure access control applications. However, real-world deployment remains challenging due to variations in illumination, background noise, facial expressions, and other environmental distortions that often reduce recognition accuracy. Most of the prevailing methods perceive face recognition as a pixel-based issue and make extensive use of traditional convolutional neural networks or handcrafted descriptors, which often are not able to realize the intricate interactions between various regions of the faces. This paper will propose a deep learning architecture to overcome these constraints by combining the method of multivariate features extraction with a stacked attention-based stacked Long Short-Term Memory (LSTM) architecture. The implementations of the proposed approach would involve a normalization of Z scores to equalize the pixel intensity of the facial images and then Independent Component Analysis (ICA) to identify statistically independent and discriminating facial features, which include edges, contours and textures. The latter are then handled with a stacked attention LSTM model, which is learned to operate at the sequential level and consequently selectively attends and concentrates on important parts of the face such as the eyes, nose and mouth and attenuates background noise. The framework was tested on two evaluation datasets of CelebA and CASIA-WebFace that consist of large-scale facial images with various variations. Experimental results demonstrate that the proposed system achieves recognition rates of 96% on CelebA and 92% on CASIA-WebFace, with an RMSE of 34.82, indicating improved robustness and generalisation compared with conventional deep learning models. These findings confirm the effectiveness of the proposed approach for reliable and scalable facial recognition applications.

In industrial robots, harmonic drive flexible bearings are prone to faults, and fault diagnosis is essential for preventing unexpected downtime. However, vibration signals acquired from robot joints are often non-stationary and contaminated by strong multi-source interference, including motion-induced interference and vibrations induced by the deformation of flexible components. Such interference severely masks the subtle signatures of faults. To address this issue, this paper proposes a fault diagnosis framework that leverages multi-channel vibration signals to enhance fault-related features. First, angular resampling is applied to eliminate speed-induced non-stationarity. Second, envelope extraction is utilized to obtain demodulated signals suitable for independent component analysis (ICA). Subsequently, ICA is employed to extract fault-related components from the multi-channel signals. Finally, the fault-related independent component is identified and analyzed via envelope order spectrum analysis. Experimental validation on an industrial robot under both single-joint and multi-joint operating conditions demonstrates the effectiveness of the proposed framework. The method suppresses multi-source interference and achieves accurate fault diagnosis for flexible bearings under complex operating conditions, with quantitative validation confirming the diagnostic performance of the proposed framework.

Abstract—Breast cancer remains one of the most common causes of mortality among women worldwide. Accurate and early diagnosis plays an essential role in improving survival rates. Machine learning techniques have increasingly been applied in medical decision-support systems to assist physicians in identifying malignant tumors more reliably. This research investigates the use of Independent Component Analysis (ICA) as a feature reduction technique for breast cancer classification. The Wisconsin Diagnostic Breast Cancer (WDBC) dataset is utilized, which initially contains thirty diagnostic attributes extracted from digitized biopsy images. ICA is applied to trans- form the original feature space into a reduced representation consisting of a single independent component. To evaluate the effectiveness of this dimensionality reduction approach, several machine learning classifiers are employed, including k-Nearest Neighbor (k-NN), Artificial Neural Networks (ANN), Radial Basis Function Neural Networks (RBFNN), and Support Vector Machines (SVM). The classification performance is examined using both the original 30-feature dataset and the reduced feature representation. Different validation strategies such as 5- fold cross-validation, 10-fold cross-validation, and random data partitioning are used to assess performance. The classifiers are evaluated using multiple metrics including accuracy, sensitivity, specificity, F-score, Youden’s index, discriminant power, and Receiver Operating Characteristic (ROC) analysis. Experi- mental results indicate that reducing the feature dimension through ICA significantly decreases computational cost while maintaining competitive diagnostic accuracy. These findings suggest that ICA-based feature reduction can be beneficial for developing efficient computer-aided breast cancer diagnosis systems. Index Terms—Breast cancer, ICA, Machine learning, Classi- fication

Mild cognitive impairment (MCI) is regarded a potential early stage of Alzheimer's disease (AD) and associated with a significantly increased risk of progression to AD. This study aims to evaluate whole-brain static functional connectivity (SFC) disruptions with the default mode network (DMN) seed points in patients with MCI by resting-state functional magnetic resonance imaging (rs-fMRI), and to explore whether these disruptions could serve as potential markers for MCI progression to AD. Retrospective rs-fMRI data with MCI (n = 36) and corresponding matched healthy controls (HCs) (n = 26) were collected for comparison. Independent component analysis (ICA) was used to extract DMN regions, and SFC was calculated for four seed points within the DMN. Two-sample t-tests were performed to compare group differences in SFC strength between the MCI and HC groups, and Pearson correlation analyses were conducted. Compared to HCs, the MCI group showed both increased and decreased SFC between four subregions and multiple brain regions, decreased SFC was more widely distributed than increased SFC. Abnormal connectivity was more prominent in the first two key nodes compared to the latter two. Affected regions primarily located in the precuneus, frontal gyri, temporal gyri, postcentral gyrus, caudate nucleus, lingual gyrus, and fusiform gyrus. The SFC value between the right angular gyrus and the right insula was significantly negatively correlated with MoCA scores (r = - 0.385, p < 0.05, FDR-corrected). It reveals a decline in the functional integration capacity within the DMN, as well as complex reorganization and abnormal connectivity patterns between the DMN and other brain networks. The altered interactions between DMN subregions and abnormal brain areas are significantly associated with episodic memory disturbance in MCI.

Adolescent depression patients with psychotic symptoms have a worse prognosis and higher risk of suicide compared to those without psychotic symptoms, however, the neurophysiological underlying pathway is still unclear. The fronto-striatal network (FSN) may play an important role in the association between depression and psychotic symptoms. This study aims to investigate the specific alterations of the FSN in adolescent patients with depression, both with and without psychotic symptoms, and determine whether it could serve as an underlying neuropathological underlying pathway contributing to clinical symptoms. Based on the presence or absence of psychotic events such as hallucinations or delusions, adolescent depression patients were divided into a psychotic depression (PD) group (n = 32) and a non-psychotic depression (NPD) group (n = 41). All participants underwent comprehensive clinical assessments, including the Brief Psychiatric Rating Scale (BPRS) for general psychopathology and the Patient Health Questionnaire-9 (PHQ-9) for depressive symptom severity. Resting-state functional magnetic resonance imaging data were acquired using a 3.0T scanner. These data were preprocessed and analyzed using independent component analysis (ICA) to identify the FSN for subsequent between group comparisons and correlation analyses. In contrast to NPD, PD group demonstrated decreased connectivity in the bilateral striatum (bilateral caudate and putamen) (cluster-based family-wise error 0.05 correction). Furthermore, a significantly negative correlation was observed between the left striatum and BPRS scores (r = - 0.385, P = 0.030) in PD group, while no correlation was found in NPD group. The mediation analysis results showed that no mediating effect of PHQ-9 scores was found in the left striatum and the BPRS scores of PD group, nor was any mediating effect of PHQ-9 scores in the right striatum and the BPRS scores of NPD group. Our results indicate a distinct alteration in the function connectivity in the bilateral striatal between PD and NPD, providing insights into the potential pathogenesis of psychotic symptoms in adolescent depression.