Search Results (170)

Introduction: Congenital hypertrophy of the retinal pigment epithelium (CHRPE) is a benign proliferation of the melanin-producing retinal pigment epithelium (RPE). Although often a benign and incidental finding, multifocal CHRPE may mimic lesions associated with familial adenomatous polyposis (FAP). Case Description: We describe an 8-year-old girl presenting with optic disc pallor and widespread multifocal bear track CHRPE observed bilaterally on dilated fundoscopy. Fundus autofluorescence (FAF) imaging showed uniform areas of hypoautofluorescence corresponding to the bear track lesions. Spectral domain optical coherence tomography (SD-OCT) demonstrated normal lamination without atrophy. The full-field electroretinogram (ffERG) was within normal limits. Whole-genome sequencing (WGS) revealed a likely pathogenic heterozygous variant in the STAG2 gene (c.3222dup, p.Ser1075IlefsTer12). Conclusions: We present a rare case of bilateral, panretinal bear track CHRPE in a child with a likely pathogenic variant in STAG2. Using multimodal imaging, we contrast bear track lesions of the retina with FAP-associated CHRPE. We also present possible ophthalmic manifestations in carriers of pathogenic STAG2 variants. Full article

Emotion recognition in speech is essential for enhancing human–computer interaction (HCI) systems. Despite progress in Bangla speech emotion recognition, challenges remain, including low accuracy, speaker dependency, and poor generalization across emotional expressions. Previous approaches often rely on traditional machine learning or basic deep learning models, struggling with robustness and accuracy in noisy or varied data. In this study, we propose a novel multi-stream deep learning feature fusion approach for Bangla speech emotion recognition, addressing the limitations of existing methods. Our approach begins with various data augmentation techniques applied to the training dataset, enhancing the model’s robustness and generalization. We then extract a comprehensive set of handcrafted features, including Zero-Crossing Rate (ZCR), chromagram, spectral centroid, spectral roll-off, spectral contrast, spectral flatness, Mel-Frequency Cepstral Coefficients (MFCCs), Root Mean Square (RMS) energy, and Mel-spectrogram. Although these features are used as 1D numerical vectors, some of them are computed from time–frequency representations (e.g., chromagram, Mel-spectrogram) that can themselves be depicted as images, which is conceptually close to imaging-based analysis. These features capture key characteristics of the speech signal, providing valuable insights into the emotional content. Sequentially, we utilize a multi-stream deep learning architecture to automatically learn complex, hierarchical representations of the speech signal. This architecture consists of three distinct streams: the first stream uses 1D convolutional neural networks (1D CNNs), the second integrates 1D CNN with Long Short-Term Memory (LSTM), and the third combines 1D CNNs with bidirectional LSTM (Bi-LSTM). These models capture intricate emotional nuances that handcrafted features alone may not fully represent. For each of these models, we generate predicted scores and then employ ensemble learning with a soft voting technique to produce the final prediction. This fusion of handcrafted features, deep learning-derived features, and ensemble voting enhances the accuracy and robustness of emotion identification across multiple datasets. Our method demonstrates the effectiveness of combining various learning models to improve emotion recognition in Bangla speech, providing a more comprehensive solution compared with existing methods. We utilize three primary datasets—SUBESCO, BanglaSER, and a merged version of both—as well as two external datasets, RAVDESS and EMODB, to assess the performance of our models. Our method achieves impressive results with accuracies of 92.90%, 85.20%, 90.63%, 67.71%, and 69.25% for the SUBESCO, BanglaSER, merged SUBESCO and BanglaSER, RAVDESS, and EMODB datasets, respectively. These results demonstrate the effectiveness of combining handcrafted features with deep learning-based features through ensemble learning for robust emotion recognition in Bangla speech. Full article

Pesticide residues in fruits and vegetables pose a serious threat to food safety. Traditional detection methods have defects such as complex operation, high cost, and long detection time. Therefore, it is of great significance to develop rapid, non-destructive, and efficient detection technologies and equipment. In recent years, the combination of spectroscopic techniques and imaging technologies with machine learning algorithms has developed rapidly, providing a new attempt to solve this problem. This review focuses on the research progress of the combination of spectroscopic techniques (near-infrared spectroscopy (NIRS), hyperspectral imaging technology (HSI), surface-enhanced Raman scattering (SERS), laser-induced breakdown spectroscopy (LIBS), and imaging techniques (visible light (VIS) imaging, NIRS imaging, HSI technology, terahertz imaging) with machine learning algorithms in the detection of pesticide residues in fruits and vegetables. It also explores the huge challenges faced by the application of spectroscopic and imaging technologies combined with machine learning algorithms in the intelligent perception of pesticide residues in fruits and vegetables: the performance of machine learning models requires further enhancement, the fusion of imaging and spectral data presents technical difficulties, and the commercialization of hardware devices remains underdeveloped. This review has proposed an innovative method that integrates spectral and image data, enhancing the accuracy of pesticide residue detection through the construction of interpretable machine learning algorithms, and providing support for the intelligent sensing and analysis of agricultural and food products. Full article

The intranasal delivery of exogenous mitochondria is a potential therapy for Parkinson’s disease (PD). The regulatory mechanisms and effectiveness in genetic models remains uncertain, as well as the impact of modulating the mitochondrial permeability transition pore (mPTP) in grafts. Utilizing UQCRC1 (p.Tyr314Ser) knock-in mice, and a cellular model, this study validated the transplantation of mitochondria with or without cyclosporin A (CsA) preloading as a method to treat mitochondrial dysfunction and improve disease progression through intranasal delivery. Liver-derived mitochondria were labeled with bromodeoxyuridine (BrdU), incubated with CsA to inhibit mPTP opening, and were administered weekly via the nasal route to 6-month-old mice for six months. Both treatment groups showed significant locomotor improvements in open-field tests. PET imaging showed increased striatal tracer uptake, indicating enhanced dopamine synthesis capacity. The immunohistochemical analysis revealed increased neuron survival in the dentate gyrus, a higher number of tyrosine hydroxylase (TH)-positive neurons in the substantia nigra (SN) and striatum (ST), and a thicker granule cell layer. In SN neurons, the function of mitochondrial complex III was reinstated. Additionally, the CsA-accumulated mitochondria reduced more proinflammatory cytokine levels, yet their therapeutic effectiveness was similar to that of unmodified mitochondria. External mitochondria were detected in multiple brain areas through BrdU tracking, showing a 3.6-fold increase in the ST compared to the SN. In the ST, about 47% of TH-positive neurons incorporated exogenous mitochondria compared to 8% in the SN. Notably, GFAP-labeled striatal astrocytes (ASTs) also displayed external mitochondria, while MBP-labeled striatal oligodendrocytes (OLs) did not. On the other hand, fewer ASTs and increased OLs were noted, along with lower S100β levels, indicating reduced reactive gliosis and a more supportive environment for OLs. Intranasally, mitochondrial transplantation showed neuroprotective effects in genetic PD, validating a noninvasive therapeutic approach. This supports mitochondrial recovery and is linked to anti-inflammatory responses and glial modulation. Full article

Magnetotactic bacteria (MTB), a unique group of Gram-negative prokaryotes, have the remarkable ability to biomineralize magnetic nanoparticles (MNPs) intracellularly, making them promising candidates for various biomedical applications such as biosensors, drug delivery, imaging contrast agents, and cancer-targeted therapies. To fully exploit the potential of MTB, a precise understanding of the structural, surface, and functional properties of these biologically produced nanoparticles is required. Given these concerns, this review provides a focused synthesis of the most widely used microscopic and spectroscopic methods applied in the characterization of MTB and their associated MNPs, covering the latest research from January 2022 to May 2025. Specifically, various optical microscopy techniques (e.g., transmission electron microscopy (TEM), scanning electron microscopy (SEM), and atomic force microscopy (AFM)) and spectroscopic approaches (e.g., localized surface plasmon resonance (LSPR), surface-enhanced Raman scattering (SERS), and X-ray photoelectron spectroscopy (XPS)) relevant to ultrasensitive MTB biosensor development are herein discussed and compared in term of their advantages and disadvantages. Overall, the novelty of this work lies in its clarity and structure, aiming to consolidate and simplify access to the most current and effective characterization techniques. Furthermore, several gaps in the characterization methods of MTB were identified, and new directions of methods that can be integrated into the study, analysis, and characterization of these bacteria are suggested in exhaustive manner. Finally, to the authors’ knowledge, this is the first comprehensive overview of characterization techniques that could serve as a practical resource for both younger and more experienced researchers seeking to optimize the use of MTB in the development of advanced biosensing systems and other biomedical tools. Full article

Highly infectious and pathogenic viruses seriously threaten global public health, underscoring the need for rapid and accurate diagnostic methods to effectively manage and control outbreaks. In this study, we developed a comprehensive Surface-Enhanced Raman Scattering–Lateral Flow Immunoassay (SERS-LFIA) detection system that integrates SERS scanning imaging with artificial intelligence (AI)-based result discrimination. This system was based on an ultra-sensitive SERS-LFIA strip with SiO₂-Au NSs as the immunoprobe (with a theoretical limit of detection (LOD) of 1.8 pg/mL). On this basis, a negative–positive discrimination method combining SERS scanning imaging with a deep learning model (ResNet-18) was developed to analyze probe distribution patterns near the T line. The proposed machine learning method significantly reduced the interference of abnormal signals and achieved reliable detection at concentrations as low as 2.5 pg/mL, which was close to the theoretical Raman LOD. The accuracy of the proposed ResNet-18 image recognition model was 100% for the training set and 94.52% for the testing set, respectively. In summary, the proposed SERS-LFIA detection system that integrates detection, scanning, imaging, and AI automated result determination can achieve the simplification of detection process, elimination of the need for specialized personnel, reduction in test time, and improvement of diagnostic reliability, which exhibits great clinical potential and offers a robust technical foundation for detecting other highly pathogenic viruses, providing a versatile and highly sensitive detection method adaptable for future pandemic prevention. Full article

Speech emotion recognition (SER) plays a vital role in enhancing human–computer interaction (HCI) and can be applied in affective computing, virtual support, and healthcare. This research presents a high-performance SER framework based on a lightweight 1D Convolutional Neural Network (1D-CNN) and a multi-feature fusion technique. Rather than employing spectrograms as image-based input, frame-level characteristics (Mel-Frequency Cepstral Coefficients, Mel-Spectrograms, and Chroma vectors) are calculated throughout the sequences to preserve temporal information and reduce the computing expense. The model attained classification accuracies of 94.0% on MELD (multi-party talks) and 91.9% on RAVDESS (acted speech). Ablation experiments demonstrate that the integration of complimentary features significantly outperforms the utilisation of a singular feature as a baseline. Data augmentation techniques, including Gaussian noise and time shifting, enhance model generalisation. The proposed method demonstrates significant potential for real-time emotion recognition using audio only in embedded or resource-constrained devices. Full article

Receptor-mediated endocytosis (RME) is a commonly recognized receptor internalization process of receptor degradation or recycling. However, recent studies have supported that RME is closely related to signal propagation and amplification from the plasma membrane to the cytosol. Few studies have elucidated the role of H₂O₂, a mild oxidant among reactive oxygen species (ROS) in RME and second messenger of signal propagation. In the present study, we investigated the regulatory function of H₂O₂ in early endosomes during signaling throughout receptor-mediated endocytosis. In mammalian cells with a physiological amount of H₂O₂ generated during epidermal growth factor (EGF) activation, fluorescence imaging showed that the levels of two activating phosphorylations on Ser⁴⁷³ and Thr³⁰⁸ of Akt were transiently increased in the plasma membrane, but the predominant p-Akt on Ser⁴⁷³ appeared in early endosomes. To examine the role of endosomal H₂O₂ molecules as signaling mediators of Akt activation in endosomes, we modulated endosomal H₂O₂ through the ectopic expression of an endosomal-targeting catalase (Cat-Endo). The forced removal of endosomal H₂O₂ inhibited the Akt phosphorylation on Ser⁴⁷³ but not on Thr³⁰⁸. The levels of mSIN and rictor, two components of mTORC2 that work as a kinase in Akt phosphorylation on Ser⁴⁷³, were also selectively diminished in the early endosomes of Cat-Endo-expressing cells. We also observed a decrease in the endosomal level of the adaptor protein containing the PH domain, the PTB domain, and the Leucine zipper motif 1 (APPL1) protein, which is an effector of Rab5 and key player in the assembly of signaling complexes regulating the Akt pathway in Cat-Endo-expressing cells compared with those in normal cells. Therefore, the H₂O₂-dependent recruitment of the APPL1 adaptor protein into endosomes was required for full Akt activation. We proposed that endosomal H₂O₂ is a promoter of Akt signaling. Full article

Stroke affects approximately 12 million individuals annually, necessitating swift diagnosis to avert fatal outcomes. Current hospital imaging protocols often delay treatment, underscoring the need for portable diagnostic solutions. We have investigated silver nanostars (AgNS) incubated with human plasma, deposited on a simple aluminum foil substrate, and utilizing Surface-Enhanced Raman Spectroscopy (SERS) combined with machine learning (ML) to provide a proof-of-concept for rapid differentiation of stroke types. These are the seminal steps for the development of low-cost pre-hospital diagnostics at point-of-care, with potential for improving patient outcomes. The proposed SERS assay aims to classify plasma from stroke patients, differentiating hemorrhagic from ischemic stroke. Silver nanostars were incubated with plasma and spiked with glial fibrillary acidic protein (GFAP), a biomarker elevated in hemorrhagic stroke. SERS spectra were analyzed using ML to distinguish between hemorrhagic and ischemic stroke, mimicked by different concentrations of GFAP. Key innovations include optimized AgNS–plasma incubates formation, controlled plasma-to-AgNS ratios, and a low-cost aluminum foil substrate, enabling results within 15 min. Differential analysis revealed stroke-specific protein profiles, while ML improved classification accuracy through ensemble modeling and feature engineering. The integrated ML model achieved rapid and precise stroke predictions within seconds, demonstrating the assay’s potential for immediate clinical decision-making. Full article

Over the past five years, the proliferation of virtual reality platforms and the advancement of metahuman technologies have underscored the importance of natural interaction and emotional expression. As a result, there has been significant research activity focused on developing emotion recognition techniques based on speech data. Despite significant progress in emotion recognition research for the Korean language, a shortage of speech databases applicable to such research has been regarded as the most critical problem in this field, leading to overfitting issues in several models developed by previous studies. To address the issue of overfitting caused by limited data availability in the field of Korean speech emotion recognition (SER), this study focuses on integrating the data augmentation–saturation (DA-S) technique into a traditional ResNet model to enhance SER performance. The DA-S technique enhances data augmentation by adjusting the saturation of an image. We used 11,192 utterance numbers provided by AI-HUB, which were converted into images to extract features such as pitch and intensity of speech. The DA-S technique was then applied to this dataset, using weights of 0 and 2, to augment the utterance numbers to 33,576. This augmented dataset was utilized to classify four emotion categories: happiness, sadness, anger, and neutrality. The results of this study showed that the proposed model using the DA-S technique overcame overfitting issues. Furthermore, its performance for SER increased by 34.19% compared to that of existing ResNet models not using the DA-S technique. This demonstrates that the DA-S technique effectively enhances model performance with limited data and may be applicable to specific areas such as stress monitoring and mental health support. Full article

Surface-enhanced Raman scattering (SERS) stands out as a highly effective molecular identification technique, renowned for its exceptional sensitivity, specificity, and non-destructive nature. It has become a main technology in various sectors, including biological detection and imaging, environmental monitoring, and food safety. With the development of material science and the expansion of application fields, SERS substrate materials have also undergone significant changes: from precious metals to semiconductors, from single crystals to composite particles, from rigid to flexible substrates, and from two-dimensional to three-dimensional structures. This report delves into the advancements of the three latest types of SERS substrates: colloidal, chip-based, and tip-enhanced Raman spectroscopy. It explores the design principles, distinctive functionalities, and factors that influence SERS signal enhancement within various SERS-active nanomaterials. Furthermore, it provides an outlook on the future challenges and trends in the field. The insights presented are expected to aid researchers in the development and fabrication of SERS substrates that are not only more efficient but also more cost-effective. This progress is crucial for the multifunctionalization of SERS substrates and for their successful implementation in real-world applications. Full article

The use of Raman spectroscopy, particularly surface-enhanced Raman spectroscopy (SERS), offers a powerful tool for analyzing biochemical changes in biofluids. This study aims to assess the modifications occurring in saliva collected from patients before and after exposure to cone beam computed tomography (CBCT) and computed tomography (CT) imaging. SERS analysis revealed significantly amplified spectra in post-imaging samples compared to pre-imaging samples, with pronounced intensification of thiocyanate and opiorphin bands, which, together with proteins, dominated the spectra. The changes were more pronounced in the case of CT as compared to CBCT, probably due to the use of a high radiation dose in the case of the first-mentioned technique. These findings underscore the impact of CBCT and CT on salivary composition, highlighting the relevance of SERS as a sensitive method for detecting subtle molecular changes in biofluids post-radiation exposure. This study’s results emphasize the importance of monitoring biochemical markers in patients undergoing diagnostic imaging to better understand the systemic effects of ionizing radiation. Full article

Iron oxide nanoparticles (IONPs) have garnered significant attention in biomedical applications due to their unique magnetic properties, biocompatibility, and versatility. This review comprehensively examines the synthesis methods, surface functionalization techniques, and diverse biomedical applications of IONPs. Various chemical and physical synthesis techniques, including coprecipitation, sol–gel processes, thermal decomposition, hydrothermal synthesis, and sonochemical routes, are discussed in detail, highlighting their advantages and limitations. Surface functionalization strategies, such as ligand exchange, encapsulation, and silanization, are explored to enhance the biocompatibility and functionality of IONPs. Special emphasis is placed on the role of IONPs in biosensing technologies, where their magnetic and optical properties enable significant advancements, including in surface-enhanced Raman scattering (SERS)-based biosensors, fluorescence biosensors, and field-effect transistor (FET) biosensors. The review explores how IONPs enhance sensitivity and selectivity in detecting biomolecules, demonstrating their potential for point-of-care diagnostics. Additionally, biomedical applications such as magnetic resonance imaging (MRI), targeted drug delivery, tissue engineering, and stem cell tracking are discussed. The challenges and future perspectives in the clinical translation of IONPs are also addressed, emphasizing the need for further research to optimize their properties and ensure safety and efficacy in medical applications. This review aims to provide a comprehensive understanding of the current state and future potential of IONPs in both biosensing and broader biomedical fields. Full article

Automatic Speech Emotion Recognition (SER) plays a vital role in making human–computer interactions more natural and effective. A significant challenge in SER development is the limited availability of diverse emotional speech datasets, which hinders the application of advanced deep learning models. Transfer learning is a machine learning technique that helps address this issue by utilizing knowledge from pre-trained models to improve performance on a new task in a target domain, even with limited data. This study investigates the use of transfer learning from various pre-trained networks, including speaker embedding models such as d-vector, x-vector, and r-vector, and image classification models like AlexNet, GoogLeNet, SqueezeNet, ResNet-18, and ResNet-50. We also propose enhanced versions of the x-vector and r-vector models incorporating Multi-Head Attention Pooling and Angular Margin Softmax, alongside other architectural improvements. Additionally, reverberation from the Room Impulse Response datasets was added to the speech utterances to diversify and augment the available data. Notably, the enhanced r-vector model achieved classification accuracies of 74.05% Unweighted Accuracy (UA) and 73.68% Weighted Accuracy (WA) on the IEMOCAP dataset, and 80.25% UA and 79.81% WA on the CREMA-D dataset, outperforming the existing state-of-the-art methods. This study shows that using cross-domain transfer learning is beneficial for low-resource emotion recognition. The enhanced models developed in other domains (for non-emotional tasks) can further improve the accuracy of SER. Full article

The use of surface-enhanced Raman spectroscopy (SERS) in liquid solutions has always been challenging due to signal fluctuations, inconsistent data, and difficulties in obtaining reliable results, especially at very low analyte concentrations. In our study, we introduce a new method using a three-dimensional (3D) SERS substrate made of silica microparticles (SMPs) with attached plasmonic nanoparticles (NPs). These SMPs were placed in low-concentration analyte solutions for SERS analysis. In the first approach to perform SERS in a 3D environment, glycerin was used to immobilize the particles, which enabled high-resolution SERS imaging. Additionally, we conducted time-dependent SERS measurements in an aqueous solution, where freely suspended SMPs passed through the laser focus. In both scenarios, EFs larger than 200 were achieved, which enabled the detection of low-abundance analytes. Our study demonstrates a reliable and reproducible method for performing SERS in liquid environments, offering significant advantages for the real-time analysis of dynamic processes, sensitive detection of low-concentration molecules, and potential applications in biomolecular interaction studies, environmental monitoring, and biomedical diagnostics. Full article