One of the leading causes of death for women worldwide is still breast cancer. Effective diagnosis and treatment planning depend on the accurate classification of the benign and the malignant tissues. Histopathological examination of a biopsy sample under a microscope is the standard method used for diagnosing breast cancer. Manual reading by pathologists is time-consuming and susceptible to human error and inter-observer variation. Thus, there is a need to automate this task for the timely diagnosis of the disease. The use of a Vision Transformer (ViT) for binary classification of the histopathology images from the BreakHis dataset at $\mathbf{4 0 0 X}$ magnification publicly available, is investigated in this study. Rotation, resizing, and horizontal flipping were used for the data augmentation techniques. Holdout cross-validation was implemented for dividing the dataset into training & test data. Moreover, a comparison with classifiers like Convolutional Neural Network (CNN) and Support Vector Machine (SVM) was also carried out. With an evaluation loss of 0.0848, accuracy of 97.53%, sensitivity of 97.67%, specificity of 97.20%, and AUC of 0.9950, the ViT model demonstrated encouraging outcomes. A comparison with the recent research highlights the ViT model's superior performance. These findings suggest that the ViT model performs better than many recently reported methods and thus can be used for the application in automated medical image diagnosis. Future research in multi-class classification and model explainability is suggested by the study, which also emphasizes the potential of transformer-based models in medical image classification.

Background Heartfulness meditation (HM) is a heart-based meditation practice that features a unique yogic transmission that involves transfer of subtle energy from a certified, experienced meditator to an aspirant. Purpose This study was conducted on HM practice to assess the state (momentary changes during meditation result in temporary physiological and neurological shifts) and trait (long-term personality changes developed with experience) effects on the brain using electroencephalogram (EEG) signal analysis. Methods Thirty-four participants were recruited into three groups (long-term meditators [LTM], short-term meditators [STM] and non-meditators). EEG signals from the prefrontal cortex were recorded across four states: baseline, meditation, transmission and post-transmission. Functional connectivity and graph theory measures were calculated for each frequency band (delta, theta, alpha and beta). Within-group and between-group analyses were performed to quantify HM-induced state and trait effects. Significance level was tested through statistical analysis. Results The results show improved coordination among brain networks, increased global and local efficiency, representing a positive state effect. While performing between-group analysis, significant trait changes were observed in the LTM group over frequency bands (delta, theta and alpha) of the EEG signal. These findings suggest that the altered EEG resulting from meditation reflects neurophysiological changes and functional integration of the brain network. Conclusion The positive state and trait effects were observed for HM practice in the prefrontal cortex of the brain. Furthermore, the enhanced connectivity in EEG frequency bands may lead to significant neuroplastic changes. This finding provides insights into employing functional connectivity parameters as measures for quantifying meditation quality.

Dyslexia is a learning disability characterised by a lack of reading abilities and difficulties naming words quickly and correctly. Dyslexic people struggle with the reading and comprehending of words or letters. Hence, this research presents a methodology for classifying dyslexic and normal children using rest electroencephalogram (EEG) signals. The EEG signals are collected from 19 different channels located in the occipital, temporal, frontal, and parietal regions of the brain of twenty children with dyslexia and sixteen normal youngsters (ages 8 to 16 years). Data segmentation is used in the captured signal which is at rest for 2 min, 10-s-long sample is extracted, resulting in a total of 12 samples for one participant. Data balancing is applied to improve the accuracy of the machine learning models. A total of fourteen features from each channel are collected and fed into the classifier. The prediction model employs support vector machine (SVM) learnings such as linear, quadratic, cubic, and median Gaussian. The preprocessed raw EEG signal extracts band separated features using the discrete wavelet transform (DWT). In total, 266 features were extracted. The ensemble-based feature selection method is used to determine the most significant features, reducing the computational pressure on classifiers. Experimental results show that the most significant features derived by the gradient boost technique combined with the SVM classifier produced the maximum classification accuracy of 97.2

Breast cancer (BC) is the most often diagnosed malignancy among women. It accounts for one-third of oncological patients making it the second most common cause of cancer in women worldwide and fifth most common cause of death due to cancer in women [1]. Mammography is the most preferred imaging modality for BC; however, it has limitation of exposure to radiation, high cost and unavailability in low-resource setting. Thus, ultrasound (US) has gained popularity for BC screening due to its low cost, ease of availability and radiation less procedure. However, US is noisy as accurate diagnosis remains challenging due to inter-observer variability and lack of availability of experienced radiologists in low resource areas. This motivated researchers to use computational approaches like machine learning (ML) and deep learning (DL) for automated classification of BC. The study aims to enhance diagnostic precision by developing an ensemble ML-based approach for the classification of BC using US images. The methodology involves extracting the Region of Interest (ROI), followed by extraction of traditional features as well as deep features. Experiment where performed on dataset consisting of 1875 US images having 722 benign and 342 malignant cases. Because of inherent class imbalance in the dataset further processing has been done prior to model training on various ML algorithms. Further a voting-based ML ensemble is proposed for classification using combined traditional and deep features. The proposed model achieved an overall accuracy of 91.64%, showcasing the effectiveness of proposed method in distinguishing benign and malignant cases, providing reliable decision support for clinicians.

Respiratory disorders, a leading global health issue, are often inadequately diagnosed in both developing and developed countries despite the availability of effective treatments. While lung sounds have been widely studied for early diagnosis, speech parameter-based analysis. remains underexplored. This review examines the potential of speech analysis for detecting respiratory disorders, focusing on the challenges and existing methods. We analyze studies that employ speech parameters in respiratory disease detection, comparing them with traditional methods such as lung sound analysis and medical imaging. Our review identifies a significant gap in the use of speech-based diagnostics for respiratory conditions, with most studies relying on lung sounds and imaging. The few speech-based studies highlight the potential for speech to serve as an early indicator, though the lack of open-source speech databases for various disorders limits progress. We analyzed peer-reviewed studies published between 2010 and 2024, identifying key trends in feature extraction methods, speech datasets, and machine learning techniques. Our review reveals that only 17

This article proposes a machine learning framework for risk assessment of breast cancer using factors such as those related to the patient’s history, anthropometric measures, clinical symptoms, etc. Breast cancer risk factor dataset including 14,382 cases from Breast Cancer Surveillance Consortium dataset was used in the experiments. The dataset consisted of a healthy group and a control group consisting of patients with breast cancer detected within a year of the screening mammography. Statistical analysis using Mann–Whitney U-test and classification using various classifiers such as support vector machine (SVM) different kernels, Gaussian (fine, medium, and course), Naïve Bayes, k-nearest neighbor (k = 5), and linear discriminant analysis (LDA) were applied on this data set for risk estimation of breast cancer under fivefold data division protocol. Experimental results show that the SVM classifier with cubic kernel achieves the superior classification accuracy of 99.43

Schizophrenia (SZ) is a complex psychiatric disorder that affects 1% of the global population. It disrupts thought processes, perceptions, emotional responsiveness, and social interactions. Researchers have used electroencephalogram (EEG) signals as clinical indicators for a variety of neurological and psychiatric conditions, including SZ. These signals record the electrical activity of the brain. This study uses a novel time-frequency analysis and convolutional neural networks (CNNs) to detect SZ using EEG recorded from 24 SZ patients and 24 healthy controls (HCs). Initially, the Hilbert transform was used to convert EEG signals into analytic representations, which improves time-frequency resolution. After that, the proposed adaptive window-based smoothed pseudo Wigner-Ville distribution (AD-SPWVD) is applied to extract instantaneous frequency and amplitude variations from the modified signals. The AD-SPWVD coefficient values at the outputs provide a detailed depiction of neuronal dynamics, which serve as inputs to the CNN model for detecting SZ. Results show that combining the Hilbert transform and AD-SPWVD significantly improves feature extraction from EEG data, yielding informative and discriminative features. The proposed model achieved a classification accuracy of 97%, demonstrating its effectiveness in identifying neural patterns associated with SZ. This work contributes to the development of real-time, noninvasive, and automated tools that support clinicians in the early diagnosis, detection, and continuous monitoring of SZ in practical healthcare settings. Future research will validate these findings in larger cohorts and explore their generalizability across diverse clinical conditions.

Schizophrenia (SCZ) is a severe psychotic disorder characterized by cognitive and social deficits. Its diagnosis traditionally relies on subjective clinical assessments, which can be time-consuming and inconsistent. This study introduces a newly curated resting-state electroencephalogram (EEG) dataset comprising SCZ patients and healthy controls. A deep learning-driven analysis is performed to identify significant EEG frequency bands and brain regions that can serve as reliable biomarkers for accurate SCZ detection. We propose a novel two-stage investigation framework utilizing spectrograms generated via short-term Fourier transform (STFT). In the first stage, the most discriminative EEG frequency bands are identified across the entire brain. In the second stage, these bands are mapped to specific brain regions to localize the most diagnostically relevant signals. For classification, a deep learning-based ResNet-18 model is employed to analyze the EEG spectrograms. The findings reveal that delta band activity in the posterior brain region serves as the most effective biomarker, achieving a classification accuracy of 91.3

Mental arithmetic tasks (MAT) are useful for evaluating cognitive load and detecting neurological disorders since they assess working memory, attention, and executive function. Previous studies have shown changes in connectivity that are unique to frequency, increases in frontal midline theta, and decreases in alpha that are connected to tasks. However, most of these studies have only looked at power, connectivity, or source localization independently. This research examined electroencephalography (EEG) data from 36 participants engaged in serial subtraction, integrating power, connectivity, and source location to offer a more holistic perspective on brain processes. Preprocessing involved filtering, ICA artifact removal, and later by applying manual channel inspection and segmentation. We used fast Fourier transform (FFT) to find the power spectral density (PSD), Coherence to check connectivity, and Low-resolution electromagnetic tomography (LORETA) to find the sources. The findings indicated alpha suppression in the parieto-occipital regions, an increase in delta activity in the frontal and occipital areas, mild reductions in theta activity, steady beta power, and robust theta/beta connection within the frontal-parietal-occipital networks. LORETA confirmed the activations in the frontal, parietal, and occipital cortices. These findings underscore the significance of EEG analysis in elucidating cognitive function and identifying possible markers of cognitive decline.

Electroencephalography (EEG) has emerged as an effective method for psychiatric disorder classification owing to its non-invasive and real-time properties. In the proposed study, 64channel EEG data were obtained, with one channel, F2, serving as a reference, resulting in a 63-channel dataset from schizophrenia (SZ), bipolar disorder (BPD), and healthy subjects (HS), each with 8 participants, in an eye-closed condition for three minutes. The recorded EEG data are then segmented into 10second samples to increase the data sample size and further processed using adaptive filtering-based Short-Time Fourier Transform (AF-STFT) to obtain the decomposed signal. The resulting signals were transformed into two-dimensional Fast Fourier Transform (2DFFT) image representations and classified using Shifted Window (SWIN) and Vision Transformer (ViT)based deep learning models. The SWIN Transformer achieved an accuracy of 93.06%, while the ViT achieved 90.25%, demonstrating strong performance in multiclass EEG classification. These findings highlight the potential of transformer architectures for identifying neuropsychiatric disorders. Clinically, the proposed framework may enable rapid and automated differentiation between psychiatric conditions, offering support for early diagnosis and treatment monitoring. Integration into portable EEG systems could further provide scalable and accessible real-time decision support in psychiatric practice.

Breast cancer (BC) is one of the topmost causes of mortality in women all over the world. Early detection and classification of the tumor allow proper treatment of patients and chances of survival. In this article, we propose a hybrid residual neural network (ResNet) and machine learning framework and integrate the features of both mammography (MG) and ultrasound (US) images to perform the multimodal classification of BC images as benign or malignant. The features are extracted automatically from the input images of each modality using the residual neural network from the average pooling layer. Next, the feature level fusion is carried out to obtain a feature vector by combining features of MG & US. Finally, the multimodal classification is performed using the support vector machine (SVM) as a classifier. Experiments are performed on a real-time dataset collected from patients who have undergone both MG and US examinations. The classification accuracy obtained for the multimodal approach with SVM is 99.22%, which is higher than unimodal systems. Results show that the proposed multimodal approach performs better in classifying breast tumors than unimodal mammogram and ultrasound systems.

Electromyography (EMG) signals are essential, as they are used to measure muscular activity in different parts of the human body. The measurement and analysis of EMG signal lead to various applications of muscle disorders such as muscular dystrophy, myopathy, hand movements, etc. In this paper, an improved and effective hand movement classification model is developed for amputee subjects. It includes: (1) EMG feature extraction using Discrete Wavelet Transform (DWT), (2) EMG feature selection using binary Global Best Guided Gaussian Artificial Bee Colony (BGGABC), (3) Hand movements classification using Optimized k-nearest neighbors (OKNN) classifier. The EMG signal is taken from the DB3 of NinaPro dataset comprising 17 different prosthetic hand movements recorded from 11 amputee subjects. Thereafter, DWT is applied to decompose the EMG signal for extracting features. An improved wrapper-based feature selection technique (BGGABC) is used to select the optimal feature subset for effective classification. The two variants of KNN, i.e. Smallest Modified KNN and Largest Modified KNN are taken in which item strength to a class is optimized for efficient classification. The strength of an item to a class depends on distance and weight of an item to a class. Therefore, a multi-objective Non-dominated sorting genetic algorithm-II (NSGA-II) is used for optimizing these two contradictory parameters (distance and weight) simultaneously to have optimized variants, namely: Optimized Smallest KNN (OSKNN) and Optimized Largest KNN (OLKNN). Extensive results show that the proposed method OKNN achieved the highest classification accuracy of 93.07% (OLKNN) and 89.43% (OSKNN) compared with KNN variants and competitors.

Neurodevelopmental disorders (NDs) often hamper multiple functional prints of a child brain. Despite several studies on their neural and speech responses, multimodal researches on NDs are extremely rare. The present work examined the electroencephalography (EEG) and speech signals of the ND and control children, who performed "Hindi language" vocal tasks (V) of seven different categories, viz. 'vowel', 'consonant', 'one syllable', 'multi-syllable', 'compound', 'complex', and 'sentence' (V1-V7). Statistical testing of EEG parameters showed substantially high beta and gamma band energies in frontal, central, and temporal head sites of NDs for tasks V1-V5 and in parietal too for V6. For the 'sentence' task (V7), the NDs yielded significantly high theta and low alpha energies in the parietal area. These findings imply that even performing a general context-based task exerts a heavy cognitive loading in neurodevelopmental subjects. They also exhibited poor auditory comprehension while executing a long phrasing. Further, the speech signal analysis manifested significantly high amplitude (for V1-V7) and frequency (for V3-V7) perturbations in the voices of ND children. Moreover, the classification of subjects as ND or control was done via EEG and speech features. We attained 100% accuracy, precision, and F-measure using EEG features of all tasks, and using speech features of the 'complex' task. Jointly, the 'complex' task transpired as the best vocal stimuli among V1-V7 for characterizing ND brains. Meanwhile, we also inspected inter-relations between EEG energies and speech attributes of the ND group. Our work, thus, represents a unique multimodal layout to explore the distinctiveness of neuro-impaired children.