Front Neurosci. 2024 Sep 2;18:1441285. doi: 10.3389/fnins.2024.1441285. eCollection 2024.

ABSTRACT

Accuracy-optimized convolutional neural networks (CNNs) have emerged as highly effective models at predicting neural responses in brain areas along the primate ventral stream, but it is largely unknown whether they effectively model neurons in the complementary primate dorsal stream. We explored how well CNNs model the optic flow tuning properties of neurons in dorsal area MSTd and we compared our results with the Non-Negative Matrix Factorization (NNMF) model, which successfully models many tuning properties of MSTd neurons. To better understand the role of computational properties in the NNMF model that give rise to optic flow tuning that resembles that of MSTd neurons, we created additional CNN model variants that implement key NNMF constraints - non-negative weights and sparse coding of optic flow. While the CNNs and NNMF models both accurately estimate the observer's self-motion from purely translational or rotational optic flow, NNMF and the CNNs with nonnegative weights yield substantially less accurate estimates than the other CNNs when tested on more complex optic flow that combines observer translation and rotation. Despite its poor accuracy, NNMF gives rise to tuning properties that align more closely with those observed in primate MSTd than any of the accuracy-optimized CNNs. This work offers a step toward a deeper understanding of the computational properties and constraints that describe the optic flow tuning of primate area MSTd.

PMID:39286477 | PMC:PMC11403719 | DOI:10.3389/fnins.2024.1441285

Data Brief. 2024 Aug 22;56:110855. doi: 10.1016/j.dib.2024.110855. eCollection 2024 Oct.

ABSTRACT

With the soaring demand for healthcare systems, chatbots are gaining tremendous popularity and research attention. Numerous language-centric research on healthcare is conducted day by day. Despite significant advances in Arabic Natural Language Processing (NLP), challenges remain in natural language classification and generation due to the lack of suitable datasets. The primary shortcoming of these models is the lack of suitable Arabic datasets for training. To address this, authors introduce a large Arabic Healthcare Dataset (AHD) of textual data. The dataset consists of over 808k questions and answers across 90 categories, offered to the research community for Arabic computational linguistics. Authors anticipate that this rich dataset would make a great aid for a variety of NLP tasks on Arabic textual data, especially for text classification and generation purposes. Authors present the data in raw form. AHD is composed of main dataset scraped from medical website, which is Altibbi website. AHD is made public and freely available at http://data.mendeley.com/datasets/mgj29ndgrk/5.

PMID:39286413 | PMC:PMC11403399 | DOI:10.1016/j.dib.2024.110855

Heliyon. 2024 Aug 28;10(17):e36352. doi: 10.1016/j.heliyon.2024.e36352. eCollection 2024 Sep 15.

ABSTRACT

BACKGROUND: Alcohol consumption and withdrawal alter social behaviour in humans in a sex-dependent manner. The three-chamber test is a widely used paradigm to assess rodents' social behaviour, including sociability and social novelty. Automatic tracking systems are commonly used to score time spent with conspecifics, despite failing to score direct interaction time with conspecifics rather than time in the nearby zone. Thereby, the automatically scored results are usually inaccurate and need manual corrections.

NEW METHOD: New advances in artificial intelligence (AI) have been used recently to analyze complex behaviours. DeepLabCat is a pose-estimation toolkit that allows the tracking of animal body parts. Thus, we used DeepLabCut, to introduce a scoring model of the three-chamber test to investigate alcohol withdrawal effects on social behaviour in mice considering sex and withdrawal periods. We have compared the results of two automatic pose estimation methods: automatic tracking (AnyMaze) and DeepLabCut considering the manual scoring method, the current gold standard.

RESULTS: We have found that the automatic tracking method (AnyMaze) has failed to detect the significance of social deficits in female mice during acute withdrawal. However, tracking the animal's nose using DeepLabCut showed a significant social deficit in agreement with manual scoring. Interestingly, this social deficit was shown only in females during acute and recovered by the protracted withdrawal. DLC and manually scored results showed a higher Spearman correlation coefficient and a lower bias in the Bland-Altman analysis.

CONCLUSION: our approach helps improve the accuracy of scoring the three-chamber test while outperforming commercial automatic tracking systems.

PMID:39286202 | PMC:PMC11403434 | DOI:10.1016/j.heliyon.2024.e36352

Heliyon. 2024 Aug 31;10(17):e37272. doi: 10.1016/j.heliyon.2024.e37272. eCollection 2024 Sep 15.

ABSTRACT

RATIONALE AND OBJECTIVES: To analyze morphological changes in patients with COVID-19-associated pneumonia over time, a nonrigid registration technique is required that reduces differences in respiratory phase and imaging position and does not excessively deform the lesion region. A nonrigid registration method using deep learning was applied for lung field alignment, and its practicality was verified through quantitative evaluation, such as image similarity of whole lung region and image similarity of lesion region, as well as visual evaluation by a physician.

MATERIALS AND METHODS: First, the lung field positions and sizes of the first and second CT images were roughly matched using a classical registration method based on iterative calculations as a preprocessing step. Then, voxel-by-voxel transformation was performed using VoxelMorph, a nonrigid deep learning registration method. As an objective evaluation, the similarity of the images was calculated. To evaluate the invariance of image features in the lesion site, primary statistics and 3D shape features were calculated and statistically analyzed. Furthermore, as a subjective evaluation, the similarity of images and whether nonrigid transformation caused unnatural changes in the shape and size of the lesion region were visually evaluated by a pulmonologist.

RESULTS: The proposed method was applied to 509 patient data points with high image similarity. The variances in histogram characteristics before and after image deformation were confirmed. Visual evaluation confirmed the agreement between the shape and internal structure of the lung field and the natural deformation of the lesion region.

CONCLUSION: The developed nonrigid registration method was shown to be effective for quantitative time series analysis of the lungs.

PMID:39286087 | PMC:PMC11403531 | DOI:10.1016/j.heliyon.2024.e37272

Heliyon. 2024 Aug 13;10(17):e36021. doi: 10.1016/j.heliyon.2024.e36021. eCollection 2024 Sep 15.

ABSTRACT

PURPOSE: To use artificial intelligence to identify relationships between morphological characteristics of the Meibomian glands (MGs), subject factors, clinical outcomes, and subjective symptoms of dry eye.

METHODS: A total of 562 infrared meibography images were collected from 363 subjects (170 contact lens wearers, 193 non-wearers). Subjects were 67.2 % female and were 54.8 % Caucasian. Subjects were 18 years of age or older. A deep learning model was trained to take meibography as input, segment the individual MG in the images, and learn their detailed morphological features. Morphological characteristics were then combined with clinical and symptom data in prediction models of MG function, tear film stability, ocular surface health, and subjective discomfort and dryness. The models were analyzed to identify the most heavily weighted features used by the algorithm for predictions.

RESULTS: MG morphological characteristics were heavily weighted predictors for eyelid notching and vascularization, MG expressate quality and quantity, tear film stability, corneal staining, and comfort and dryness ratings, with accuracies ranging from 65 % to 99 %. Number of visible MG, along with other clinical parameters, were able to predict MG dysfunction, aqueous deficiency and blepharitis with accuracies ranging from 74 % to 85 %.

CONCLUSIONS: Machine learning-derived MG morphological characteristics were found to be important in predicting multiple signs, symptoms, and diagnoses related to MG dysfunction and dry eye. This deep learning method illustrates the rich clinical information that detailed morphological analysis of the MGs can provide, and shows promise in advancing our understanding of the role of MG morphology in ocular surface health.

PMID:39286076 | PMC:PMC11403426 | DOI:10.1016/j.heliyon.2024.e36021

Research (Wash D C). 2022 Dec 15;2022:0004. doi: 10.34133/research.0004. eCollection 2022.

ABSTRACT

Accurate prediction of pharmacological properties of small molecules is becoming increasingly important in drug discovery. Traditional feature-engineering approaches heavily rely on handcrafted descriptors and/or fingerprints, which need extensive human expert knowledge. With the rapid progress of artificial intelligence technology, data-driven deep learning methods have shown unparalleled advantages over feature-engineering-based methods. However, existing deep learning methods usually suffer from the scarcity of labeled data and the inability to share information between different tasks when applied to predicting molecular properties, thus resulting in poor generalization capability. Here, we proposed a novel multitask learning BERT (Bidirectional Encoder Representations from Transformer) framework, named MTL-BERT, which leverages large-scale pre-training, multitask learning, and SMILES (simplified molecular input line entry specification) enumeration to alleviate the data scarcity problem. MTL-BERT first exploits a large amount of unlabeled data through self-supervised pretraining to mine the rich contextual information in SMILES strings and then fine-tunes the pretrained model for multiple downstream tasks simultaneously by leveraging their shared information. Meanwhile, SMILES enumeration is used as a data enhancement strategy during the pretraining, fine-tuning, and test phases to substantially increase data diversity and help to learn the key relevant patterns from complex SMILES strings. The experimental results showed that the pretrained MTL-BERT model with few additional fine-tuning can achieve much better performance than the state-of-the-art methods on most of the 60 practical molecular datasets. Additionally, the MTL-BERT model leverages attention mechanisms to focus on SMILES character features essential to target properties for model interpretability.

PMID:39285949 | PMC:PMC11404312 | DOI:10.34133/research.0004

Innovation (Camb). 2024 Aug 22;5(5):100691. doi: 10.1016/j.xinn.2024.100691. eCollection 2024 Sep 9.

ABSTRACT

This paper explores the evolution of geoscientific inquiry, tracing the progression from traditional physics-based models to modern data-driven approaches facilitated by significant advancements in artificial intelligence (AI) and data collection techniques. Traditional models, which are grounded in physical and numerical frameworks, provide robust explanations by explicitly reconstructing underlying physical processes. However, their limitations in comprehensively capturing Earth's complexities and uncertainties pose challenges in optimization and real-world applicability. In contrast, contemporary data-driven models, particularly those utilizing machine learning (ML) and deep learning (DL), leverage extensive geoscience data to glean insights without requiring exhaustive theoretical knowledge. ML techniques have shown promise in addressing Earth science-related questions. Nevertheless, challenges such as data scarcity, computational demands, data privacy concerns, and the "black-box" nature of AI models hinder their seamless integration into geoscience. The integration of physics-based and data-driven methodologies into hybrid models presents an alternative paradigm. These models, which incorporate domain knowledge to guide AI methodologies, demonstrate enhanced efficiency and performance with reduced training data requirements. This review provides a comprehensive overview of geoscientific research paradigms, emphasizing untapped opportunities at the intersection of advanced AI techniques and geoscience. It examines major methodologies, showcases advances in large-scale models, and discusses the challenges and prospects that will shape the future landscape of AI in geoscience. The paper outlines a dynamic field ripe with possibilities, poised to unlock new understandings of Earth's complexities and further advance geoscience exploration.

PMID:39285902 | PMC:PMC11404188 | DOI:10.1016/j.xinn.2024.100691

ACS Appl Mater Interfaces. 2024 Sep 16. doi: 10.1021/acsami.4c13223. Online ahead of print.

ABSTRACT

Artificial intelligence and human-computer interaction advances demand bioinspired sensing modalities capable of comprehending human affective states and speech. However, endowing skin-like interfaces with such intricate perception abilities remains challenging. Here, we have developed a flexible piezoresistive artificial ear (AE) sensor based on gold nanoparticles, which can convert sound signals into electrical signals through changes in resistance. By testing the sensor's performance at both frequency and sound pressure level (SPL), the AE has a frequency response range of 20 Hz to 12 kHz and can sense sound signals from up to 5 m away at a frequency of 1 kHz and an SPL of 126 dB. Furthermore, through deep learning, the device achieves up to 96.9% and 95.0% accuracy in classification and recognition applications for seven emotional and eight urban environmental noises, respectively. Hence, on one hand, our device can monitor the patient's emotional state by their speech, such as sudden yelling and screaming, which can help healthcare workers understand patients' condition in time. On the other hand, the device could also be used for real-time monitoring of noise levels in aircraft, ships, factories, and other high-decibel equipment and environments.

PMID:39285705 | DOI:10.1021/acsami.4c13223

Eur J Radiol Open. 2024 Aug 31;13:100599. doi: 10.1016/j.ejro.2024.100599. eCollection 2024 Dec.

ABSTRACT

PURPOSE: To compare image quality and detection accuracy of renal stones between deep learning image reconstruction (DLIR) and Adaptive Statistical Iterative Reconstruction-Veo (ASIR-V) reconstructed virtual non-contrast (VNC) images and true non-contrast (TNC) images in spectral CT Urography (CTU).

METHODS: A retrospective analysis was conducted on images of 70 patients who underwent abdominal-pelvic CTU in TNC phase using non-contrast scan and contrast-enhanced corticomedullary phase (CP) and excretory phase (EP) using spectral scan. The TNC scan was reconstructed using ASIR-V70 % (TNC-AR70), contrast-enhanced scans were reconstructed using AR70, DLIR medium-level (DM), and high-level (DH) to obtain CP-VNC-AR70/DM/DH and EP-VNC-AR70/DM/DH image groups, respectively. CT value, image quality and kidney stones quantification accuracy were measured and compared among groups. The subjective evaluation was independently assessed by two senior radiologists using the 5-point Likert scale for image quality and lesion visibility.

RESULTS: DH images were superior to AR70 and DM images in objective image quality evaluation. There was no statistical difference in the liver and spleen (both P > 0.05), or within 6HU in renal and fat in CT value between VNC and TNC images. EP-VNC-DH had the lowest image noise, highest SNR, and CNR, and VNC-AR70 images had better noise and SNR performance than TNC-AR70 images (all p < 0.05). EP-VNC-DH had the highest subjective image quality, and CP-VNC-DH performed the best in lesion visibility. In stone CT value and volume measurements, there was no statistical difference between VNC and TNC (P > 0.05).

CONCLUSION: The DLIR-reconstructed VNC images in CTU provide better image quality than the ASIR-V reconstructed TNC images and similar quantification accuracy for kidney stones for potential dose savings.The study highlights that deep learning image reconstruction (DLIR)-reconstructed virtual non-contrast (VNC) images in spectral CT Urography (CTU) offer improved image quality compared to traditional true non-contrast (TNC) images, while maintaining similar accuracy in kidney stone detection, suggesting potential dose savings in clinical practice.

PMID:39280122 | PMC:PMC11402413 | DOI:10.1016/j.ejro.2024.100599

Eur J Radiol Open. 2024 Aug 31;13:100594. doi: 10.1016/j.ejro.2024.100594. eCollection 2024 Dec.

ABSTRACT

PURPOSE: To assess radiomics and deep learning (DL) methods in identifying symptomatic Carotid Artery Disease (CAD) from carotid CT angiography (CTA) images. We further compare the performance of these novel methods to the conventional calcium score.

METHODS: Carotid CT angiography (CTA) images from symptomatic patients (ischaemic stroke/transient ischaemic attack within the last 3 months) and asymptomatic patients were analysed. Carotid arteries were classified into culprit, non-culprit and asymptomatic. The calcium score was assessed using the Agatston method. 93 radiomic features were extracted from regions-of-interest drawn on 14 consecutive CTA slices. For DL, convolutional neural networks (CNNs) with and without transfer learning were trained directly on CTA slices. Predictive performance was assessed over 5-fold cross validated AUC scores. SHAP and GRAD-CAM algorithms were used for explainability.

RESULTS: 132 carotid arteries were analysed (41 culprit, 41 non-culprit, and 50 asymptomatic). For asymptomatic vs symptomatic arteries, radiomics attained a mean AUC of 0.96(± 0.02), followed by DL 0.86(± 0.06) and then calcium 0.79(± 0.08). For culprit vs non-culprit arteries, radiomics achieved a mean AUC of 0.75(± 0.09), followed by DL 0.67(± 0.10) and then calcium 0.60(± 0.02). For multi-class classification, the mean AUCs were 0.95(± 0.07), 0.79(± 0.05), and 0.71(± 0.07) for radiomics, DL and calcium, respectively. Explainability revealed consistent patterns in the most important radiomic features.

CONCLUSIONS: Our study highlights the potential of novel image analysis techniques in extracting quantitative information beyond calcification in the identification of CAD. Though further work is required, the transition of these novel techniques into clinical practice may eventually facilitate better stroke risk stratification.

PMID:39280120 | PMC:PMC11402422 | DOI:10.1016/j.ejro.2024.100594

Comput Intell Neurosci. 2023 Feb 20;2023:9041355. doi: 10.1155/2023/9041355. eCollection 2023.

ABSTRACT

As the network is closely related to people's daily life, network security has become an important factor affecting the physical and mental health of human beings. Network flow classification is the foundation of network security. It is the basis for providing various network services such as network security maintenance, network monitoring, and network quality of service (QoS). Therefore, this field has always been a hot spot of academic and industrial research. Existing studies have shown that through appropriate data preprocessing techniques, machine learning methods can be used to classify network flows, most of which, however, are based on manually and expert-originated feature sets; it is a time-consuming and laborious work. Moreover, only features extracted by a single model can be used in classification tasks, which can easily make the model inefficient and prone to overfitting. In order to solve the abovementioned problems, this study proposes a multimodal automatic analysis framework based on spatial and sequential features. The framework is completely based on the deep learning method and realizes automatic extraction of two types of features, which is very suitable for processing large-flow information; this improves the efficiency of network flow classification. There are two types of frameworks based on pretraining and joint-training, respectively, with analyzing the advantages and disadvantages of them in practice. In terms of evaluation, compared with the previous methods, the experimental results show that the framework has good performance in both accuracy and stability.

PMID:39280017 | PMC:PMC11401685 | DOI:10.1155/2023/9041355

aBIOTECH. 2024 May 21;5(3):281-297. doi: 10.1007/s42994-024-00169-1. eCollection 2024 Sep.

ABSTRACT

Bakanae disease, caused by Fusarium fujikuroi, poses a significant threat to rice production and has been observed in most rice-growing regions. The disease symptoms caused by different pathogens may vary, including elongated and weak stems, slender and yellow leaves, and dwarfism, as example. Bakanae disease is likely to cause necrosis of diseased seedlings, and it may cause a large area of infection in the field through the transmission of conidia. Therefore, early disease surveillance plays a crucial role in securing rice production. Traditional monitoring methods are both time-consuming and labor-intensive and cannot be broadly applied. In this study, a combination of hyperspectral imaging technology and deep learning algorithms were used to achieve in situ detection of rice seedlings infected with bakanae disease. Phenotypic data were obtained on the 9th, 15th, and 21st day after rice infection to explore the physiological and biochemical performance, which helps to deepen the research on the disease mechanism. Hyperspectral data were obtained over these same periods of infection, and a deep learning model, named Rice Bakanae Disease-Visual Geometry Group (RBD-VGG), was established by leveraging hyperspectral imaging technology and deep learning algorithms. Based on this model, an average accuracy of 92.2% was achieved on the 21st day of infection. It also achieved an accuracy of 79.4% as early as the 9th day. Universal characteristic wavelengths were extracted to increase the feasibility of using portable spectral equipment for field surveillance. Collectively, the model offers an efficient and non-destructive surveillance methodology for monitoring bakanae disease, thereby providing an efficient avenue for disease prevention and control.

SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s42994-024-00169-1.

PMID:39279856 | PMC:PMC11399517 | DOI:10.1007/s42994-024-00169-1

ArXiv [Preprint]. 2024 Sep 3:arXiv:2409.02303v1.

ABSTRACT

We propose a lesion-aware graph neural network (LEGNet) to predict language ability from resting-state fMRI (rs-fMRI) connectivity in patients with post-stroke aphasia. Our model integrates three components: an edge-based learning module that encodes functional connectivity between brain regions, a lesion encoding module, and a subgraph learning module that leverages functional similarities for prediction. We use synthetic data derived from the Human Connectome Project (HCP) for hyperparameter tuning and model pretraining. We then evaluate the performance using repeated 10-fold cross-validation on an in-house neuroimaging dataset of post-stroke aphasia. Our results demonstrate that LEGNet outperforms baseline deep learning methods in predicting language ability. LEGNet also exhibits superior generalization ability when tested on a second in-house dataset that was acquired under a slightly different neuroimaging protocol. Taken together, the results of this study highlight the potential of LEGNet in effectively learning the relationships between rs-fMRI connectivity and language ability in a patient cohort with brain lesions for improved post-stroke aphasia evaluation.

PMID:39279836 | PMC:PMC11398550

ArXiv [Preprint]. 2024 Sep 4:arXiv:2409.03080v1.

ABSTRACT

Deep learning models have shown promise in histopathology image analysis, but their opaque decision-making process poses challenges in high-risk medical scenarios. Here we introduce HIPPO, an explainable AI method that interrogates attention-based multiple instance learning (ABMIL) models in computational pathology by generating counterfactual examples through tissue patch modifications in whole slide images. Applying HIPPO to ABMIL models trained to detect breast cancer metastasis reveals that they may overlook small tumors and can be misled by non-tumor tissue, while attention maps$\unicode{x2014}$widely used for interpretation$\unicode{x2014}$often highlight regions that do not directly influence predictions. By interpreting ABMIL models trained on a prognostic prediction task, HIPPO identified tissue areas with stronger prognostic effects than high-attention regions, which sometimes showed counterintuitive influences on risk scores. These findings demonstrate HIPPO's capacity for comprehensive model evaluation, bias detection, and quantitative hypothesis testing. HIPPO greatly expands the capabilities of explainable AI tools to assess the trustworthy and reliable development, deployment, and regulation of weakly-supervised models in computational pathology.

PMID:39279830 | PMC:PMC11398542

Curr Gene Ther. 2024 Sep 13. doi: 10.2174/0115665232312364240902060458. Online ahead of print.

ABSTRACT

INTRODUCTION: Lung cancer stands as one of the most prevalent malignant neoplasms, with microRNAs (miRNAs) playing a pivotal role in the modulation of gene expression, impacting cancer cell proliferation, invasion, metastasis, immune escape, and resistance to therapy.

METHOD: The intricate role of miRNAs in lung cancer underscores their significance as biomarkers for early detection and as novel targets for therapeutic intervention. Traditional approaches for the identification of miRNAs related to lung cancer, however, are impeded by inefficiencies and complexities.

RESULTS: In response to these challenges, this study introduced an innovative deep-learning strategy designed for the efficient and precise identification of lung cancer-associated miRNAs. Through comprehensive benchmark tests, our method exhibited superior performance relative to existing technologies.

CONCLUSION: Further case studies have also confirmed the ability of our model to identify lung cancer-associated miRNAs that have undergone biological validation.

PMID:39279703 | DOI:10.2174/0115665232312364240902060458

ACS Nano. 2024 Sep 15. doi: 10.1021/acsnano.4c06619. Online ahead of print.

ABSTRACT

Two-dimensional (2D) ferroelectric semiconductors present opportunities for integrating ferroelectrics into high-density ultrathin nanoelectronics. Among the few synthesized 2D ferroelectrics, α-In2Se3, known for its electrically addressable vertical polarization, has attracted significant interest. However, the understanding of many fundamental characteristics of this material, such as the existence of spontaneous in-plane polarization and switching mechanisms, remains controversial, marked by conflicting experimental and theoretical results. Here, our combined experimental characterizations with piezoresponse force microscope and symmetry analysis conclusively dismiss previous claims of in-plane ferroelectricity in single-domain α-In2Se3. The processes of vertical polarization switching in monolayer α-In2Se3 are explored with deep-learning-assisted large-scale molecular dynamics simulations, revealing atomistic mechanisms fundamentally different from those of bulk ferroelectrics. Despite lacking in-plane effective polarization, 1D domain walls can be moved by both out-of-plane and in-plane fields, exhibiting avalanche dynamics characterized by abrupt, intermittent moving patterns. The propagating velocity at various temperatures, field orientations, and strengths can be statistically described with a universal creep equation, featuring a dynamical exponent of 2 that is distinct from all known values for elastic interfaces moving in disordered media. This work rectifies a long-held misunderstanding regarding the in-plane ferroelectricity of α-In2Se3, and the quantitative characterizations of domain wall velocity will hold broad implications for both the fundamental understanding and technological applications of 2D ferroelectrics.

PMID:39279156 | DOI:10.1021/acsnano.4c06619

J Pak Med Assoc. 2024 Sep;74(9):1659-1664. doi: 10.47391/JPMA.11215.

ABSTRACT

OBJECTIVE: To evaluate the effectiveness and analyse the influence of YouTube as a learning modality for clinical procedures among medical and dental students in a public-sector setting.

METHODS: The cross-sectional study was conducted at the medical and dental constituent institutes of Jinnah Sindh Medical University and Jinnah Postgraduate Medical Centre between August and October 2023, and comprised undergraduate, graduate, and postgraduate students of either gender aged 18-40 years. Data was collected using a self-administered, structured, closed-ended 16-item questionnaire, which was developed in the English language and explored the usage of YouTube as a source of information about medical and dental clinical procedures. Data was coded and analysed using SPSS 26.

RESULTS: Of the 314 participants, 153(48.7%) were medical students and 161(51.3%) were from the dental stream, 175(55.7%) were females, and 139(44.3%) were males. YouTube was a helpful tool for 143(45.5%) students who used it when needed, 172(54.8%) used it occasionally before attempting procedures, while majority of the dental students 140(44.6%) used it to study for prosthodontics. There were 154(49%) students who supported the idea that faculty should recommend watching relevant videos on YouTube, while 256(81.5%) preferred other websites.

CONCLUSION: YouTube was mostly used for learning clinical procedures by the students.

PMID:39279072 | DOI:10.47391/JPMA.11215

Sci Rep. 2024 Sep 15;14(1):21540. doi: 10.1038/s41598-024-72867-1.

ABSTRACT

Unmanned aerial vehicle (UAV)-assisted communication based on automatic modulation classification (AMC) technology is considered an effective solution to improve the transmission efficiency of wireless communication systems, as it can adaptively select the most suitable modulation method according to the current communication environment. However, many existing deep learning (DL)-based AMC methods cannot be directly applied to UAV platform with limited computing power and storage space, because of the contradiction between accuracy and efficiency. This paper mainly studies the lightweight of DL-based AMC networks to improve adaptability in resource-constrained scenarios. To address this challenge, we propose an ultra-lightweight neural network (ULNN). This network incorporates a lightweight convolutional structure, attention mechanism, and cross-channel feature fusion technique. Additionally, we introduce data augmentation (DA) based on signal phase offsets during the model training process, aimed at improving the model's generalization ability and preventing overfitting. Through experimental validation on the public dataset RML2016.10 A, the ULNN we proposed achieves an average precision of 62.83% with only 8815 parameters and reaches a peak classification accuracy of 92.11% at SNR = 10 dB. The experimental results show that ULNN can achieve high recognition accuracy while keeping the model lightweight, and is suitable for UAV platform with limited resources.

PMID:39278962 | DOI:10.1038/s41598-024-72867-1

Sci Rep. 2024 Sep 15;14(1):21529. doi: 10.1038/s41598-024-70527-y.

ABSTRACT

Liver ultrasound is widely used in clinical practice due to its advantages of non-invasiveness, non-radiation, and real-time imaging. Accurate segmentation of the liver region in ultrasound images is essential for accelerating the auxiliary diagnosis of liver-related diseases. This paper proposes BACANet, a deep learning algorithm designed for real-time liver ultrasound segmentation. Our approach utilizes a lightweight network backbone for liver feature extraction and incorporates a convolutional attention mechanism to enhance the network's ability to capture global contextual information. To improve early localization of liver boundaries, we developed a selective large kernel convolution module for boundary feature extraction and introduced explicit liver boundary supervision. Additionally, we designed an enhanced attention gate to efficiently convey liver body and boundary features to the decoder to enhance the feature representation capability. Experimental results across multiple datasets demonstrate that BACANet effectively completes the task of liver ultrasound segmentation, achieving a balance between inference speed and segmentation accuracy. On a public dataset, BACANet achieved a DSC of 0.921 and an IOU of 0.854. On a private test dataset, BACANet achieved a DSC of 0.950 and an IOU of 0.907, with an inference time of approximately 0.32 s per image on a CPU processor.

PMID:39278955 | DOI:10.1038/s41598-024-70527-y

Sci Rep. 2024 Sep 15;14(1):21537. doi: 10.1038/s41598-024-71545-6.

ABSTRACT

Assisted living facilities cater to the demands of the elderly population, providing assistance and support with day-to-day activities. Fall detection is fundamental to ensuring their well-being and safety. Falls are frequent among older persons and might cause severe injuries and complications. Incorporating computer vision techniques into assisted living environments is revolutionary for these issues. By leveraging cameras and complicated approaches, a computer vision (CV) system can monitor residents' movements continuously and identify any potential fall events in real time. CV, driven by deep learning (DL) techniques, allows continuous surveillance of people through cameras, investigating complicated visual information to detect potential fall risks or any instances of falls quickly. This system can learn from many visual data by leveraging DL, improving its capability to identify falls while minimalizing false alarms precisely. Incorporating CV and DL enhances the efficiency and reliability of fall detection and allows proactive intervention, considerably decreasing response times in emergencies. This study introduces a new Deep Feature Fusion with Computer Vision for Fall Detection and Classification (DFFCV-FDC) technique. The primary purpose of the DFFCV-FDC approach is to employ the CV concept for detecting fall events. Accordingly, the DFFCV-FDC approach uses the Gaussian filtering (GF) approach for noise eradication. Besides, a deep feature fusion process comprising MobileNet, DenseNet, and ResNet models is involved. To improve the performance of the DFFCV-FDC technique, improved pelican optimization algorithm (IPOA) based hyperparameter selection is performed. Finally, the detection of falls is identified using the denoising autoencoder (DAE) model. The performance analysis of the DFFCV-FDC methodology was examined on the benchmark fall database. A widespread comparative study reported the supremacy of the DFFCV-FDC approach with existing techniques.

PMID:39278949 | DOI:10.1038/s41598-024-71545-6