Biophys J. 2024 Dec 24:S0006-3495(24)04105-5. doi: 10.1016/j.bpj.2024.12.024. Online ahead of print.

ABSTRACT

Advanced deep learning and statistical methods can predict structural models for RNA molecules. However, RNAs are flexible, and it remains difficult to describe their macromolecular conformations in solutions where varying conditions can induce conformational changes. Small-angle X-ray scattering (SAXS) in solution is an efficient technique to validate structural predictions by comparing the experimental SAXS profile with those calculated from predicted structures. There are two main challenges in comparing SAXS profiles to RNA structures: the absence of cations essential for stability and charge neutralization in predicted structures and the inadequacy of a single structure to represent RNA's conformational plasticity. We introduce Solution Conformation Predictor for RNA (SCOPER) to address these challenges. This pipeline integrates kinematics-based conformational sampling with the innovative deep-learning model, IonNet, designed for predicting Mg2+ ion binding sites. Validated through benchmarking against fourteen experimental datasets, SCOPER significantly improved the quality of SAXS profile fits by including Mg2+ ions and sampling of conformational plasticity. We observe that an increased content of monovalent and bivalent ions leads to decreased RNA plasticity. Therefore, carefully adjusting the plasticity and ion density is crucial to avoid overfitting experimental SAXS data. SCOPER is an efficient tool for accurately validating the solution state of RNAs given an initial, sufficiently accurate structure and provides the corrected atomistic model, including ions.

PMID:39722452 | DOI:10.1016/j.bpj.2024.12.024

Biol Psychol. 2024 Dec 23:108976. doi: 10.1016/j.biopsycho.2024.108976. Online ahead of print.

ABSTRACT

Within the domain of neurodevelopmental disorders, autism spectrum disorder (ASD) emerges as a distinctive neurological condition characterized by multifaceted challenges. The delayed identification of ASD poses a considerable hurdle in effectively managing its impact and mitigating its severity. Addressing these complexities requires a nuanced understanding of data modalities and the underlying patterns. Existing studies have focused on a single data modality for ASD diagnosis. Recently, there has been a significant shift towards multimodal architectures with deep learning strategies due to their ability to handle and incorporate complex data modalities. In this paper, we developed a novel multimodal ASD diagnosis architecture, referred to as Multi-Head CNN with BERT (MCBERT), which integrates bidirectional encoder representations from transformers (BERT) for meta-features and a multi-head convolutional neural network (MCNN) for the brain image modality. The MCNN incorporates two attention mechanisms to capture spatial (SAC) and channel (CAC) features. The outputs of BERT and MCNN are then fused and processed through a classification module to generate the final diagnosis. We employed the ABIDE-I dataset, a multimodal dataset, and conducted a leave-one-site-out classification to assess the model's effectiveness comprehensively. Experimental simulations demonstrate that the proposed architecture achieves a high accuracy of 93.4%. Furthermore, the exploration of functional MRI data may provide a deeper understanding of the underlying characteristics of ASD.

PMID:39722324 | DOI:10.1016/j.biopsycho.2024.108976

J Agric Food Chem. 2024 Dec 25. doi: 10.1021/acs.jafc.4c10830. Online ahead of print.

ABSTRACT

This study utilized deep learning to optimize antihypertensive peptides from whey protein hydrolysate. Using the Large Language Models (LLMs), we identified an optimal multienzyme combination (MC5) with an ACE inhibition rate of 89.08% at a concentration of 1 mg/mL, significantly higher than single-enzyme hydrolysis. MC5 (1 mg/mL) exhibited excellent biological stability, with the ACE inhibition decreasing by only 6.87% after simulated digestion. In in vivo experiments, MC5 reduced the systolic and diastolic blood pressure of hypertensive rats to 125.00 and 89.00 mmHg, respectively. MC5 significantly lowered inflammatory markers (TNF-α and IL-6) and increased antioxidant enzyme activity (SOD, GSH-Px, GR, and CAT). Compared to the MC group, the MC5 group showed significantly reduced serum renin and ET-1 levels by 1.25-fold and 1.04-fold, respectively, while serum NO content increased by 3.15-fold. Furthermore, molecular docking revealed four potent peptides (LPEW, LKPTPEGDL, LNYW, and LLL) with high ACE binding affinity. This approach demonstrated the potential of combining computational methods with traditional hydrolysis processes to develop effective dietary interventions for hypertension.

PMID:39721995 | DOI:10.1021/acs.jafc.4c10830

J Chem Inf Model. 2024 Dec 25. doi: 10.1021/acs.jcim.4c01801. Online ahead of print.

ABSTRACT

Enzymes are ubiquitous catalysts with enormous application potential in biomedicine, green chemistry, and biotechnology. However, accurately predicting whether a molecule serves as a substrate for a specific enzyme, especially for novel entities, remains a significant challenge. Compared with traditional experimental methods, computational approaches are much more resource-efficient and time-saving, but they often compromise on accuracy. To address this, we introduce the molecule-enzyme interaction (MEI) model, a novel machine learning framework designed to predict the probability that a given molecule is a substrate for a specified enzyme with high accuracy. Utilizing a comprehensive data set that encapsulates extensive information on enzymatic reactions and enzyme sequences, the MEI model seamlessly combines atomic environmental data with amino acid sequence features through an advanced attention mechanism within a hierarchical neural network. Empirical evaluations have confirmed that the MEI model outperforms the current state-of-the-art model by at least 6.7% in prediction accuracy and 8.5% in AUROC, underscoring its enhanced predictive capabilities. Additionally, the MEI model demonstrates remarkable generalization across data sets of varying qualities and sizes. This adaptability is further evidenced by its successful application in diverse areas, such as predicting interactions within the CYP450 enzyme family and achieving an outstanding accuracy of 90.5% in predicting the enzymatic breakdown of complex plastics within environmental applications. These examples illustrate the model's ability to effectively transfer knowledge from coarsely annotated enzyme databases to smaller, high-precision data sets, robustly modeling both sparse and high-quality databases. We believe that this versatility firmly establishes the MEI model as a foundational tool in enzyme research with immense potential to extend beyond its original scope.

PMID:39721977 | DOI:10.1021/acs.jcim.4c01801

J Integr Med. 2024 Dec 6:S2095-4964(24)00412-6. doi: 10.1016/j.joim.2024.12.001. Online ahead of print.

ABSTRACT

OBJECTIVE: Rheumatoid arthritis (RA) is a systemic autoimmune disease that affects the small joints of the whole body and degrades the patients' quality of life. Zhengqing Fengtongning (ZF) is a traditional Chinese medicine preparation used to treat RA. ZF may cause liver injury. In this study, we aimed to develop a prediction model for abnormal liver function caused by ZF.

METHODS: This retrospective study collected data from multiple centers from January 2018 to April 2023. Abnormal liver function was set as the target variable according to the alanine transaminase (ALT) level. Features were screened through univariate analysis and sequential forward selection for modeling. Ten machine learning and deep learning models were compared to find the model that most effectively predicted liver function from the available data.

RESULTS: This study included 1,913 eligible patients. The LightGBM model exhibited the best performance (accuracy = 0.96) out of the 10 learning models. The predictive metrics of the LightGBM model were as follows: precision = 0.99, recall rate = 0.97, F1_score = 0.98, area under the curve (AUC) = 0.98, sensitivity = 0.97 and specificity = 0.85 for predicting ALT < 40 U/L; precision = 0.60, recall rate = 0.83, F1_score = 0.70, AUC = 0.98, sensitivity = 0.83 and specificity = 0.97 for predicting 40 ≤ ALT < 80 U/L; and precision = 0.83, recall rate = 0.63, F1_score = 0.71, AUC = 0.97, sensitivity = 0.63 and specificity = 1.00 for predicting ALT ≥ 80 U/L. ZF-induced abnormal liver function was found to be associated with high total cholesterol and triglyceride levels, the combination of TNF-α inhibitors, JAK inhibitors, methotrexate + nonsteroidal anti-inflammatory drugs, leflunomide, smoking, older age, and females in middle-age (45-65 years old).

CONCLUSION: This study developed a model for predicting ZF-induced abnormal liver function, which may help improve the safety of integrated administration of ZF and Western medicine. Please cite this article as: Yu Z, Kou F, Gao Y, Lyu CM, Gao F, Wei H. A machine learning model for predicting abnormal liver function induced by a Chinese herbal medicine preparation (Zhengqing Fengtongning) in patients with rheumatoid arthritis based on real-world study. J Integr Med. 2024; Epub ahead of print.

PMID:39721810 | DOI:10.1016/j.joim.2024.12.001

Water Res. 2024 Dec 19;273:123011. doi: 10.1016/j.watres.2024.123011. Online ahead of print.

ABSTRACT

Machine learning has been increasingly used to solve management problems of water distribution networks (WDNs). A critical research gap, however, remains in the effective incorporation of WDN hydraulic characteristics in machine learning. Here we present a new water distribution network embedding (WDNE) method that transforms the hydraulic relationships of WDN topology into a vector form to be best suited for machine learning algorithms. The nodal relationships are characterized by local structure, global structure and attribute information. A conjoint use of two deep auto-encoder embedding models ensures that the hydraulic relationships and attribute information are simultaneously preserved and are effectively utilized by machine learning models. WDNE provides a new way to bridge WDN hydraulics with machine learning. It is first applied to a pipe burst localization problem. The results show that it can increase the performance of machine learning algorithms, and enable a lightweight machine learning algorithm to achieve better accuracy with less training data compared with a deep learning method reported in the literature. Then, applications in node grouping problems show that WDNE enables machine learning algorithms to make use of WDN hydraulic information, and integrates WDN structural relationships to achieve better grouping results. The results highlight the potential of WDNE to enhance WDN management by improving the efficiency of machine learning models and broadening the range of solvable problems. Codes are available at https://github.com/ZhouGroupHFUT/WDNE.

PMID:39721501 | DOI:10.1016/j.watres.2024.123011

Artif Intell Med. 2024 Dec 16;160:103055. doi: 10.1016/j.artmed.2024.103055. Online ahead of print.

ABSTRACT

In this paper, we introduce a novel concordance-based predictive uncertainty (CPU)-Index, which integrates insights from subgroup analysis and personalized AI time-to-event models. Through its application in refining lung cancer screening (LCS) predictions generated by an individualized AI time-to-event model trained with fused data of low dose CT (LDCT) radiomics with patient demographics, we demonstrate its effectiveness, resulting in improved risk assessment compared to the Lung CT Screening Reporting & Data System (Lung-RADS). Subgroup-based Lung-RADS faces challenges in representing individual variations and relies on a limited set of predefined characteristics, resulting in variable predictions. Conversely, personalized AI time-to-event models are hindered by transparency issues and biases from censored data. By measuring the prediction consistency between subgroup analysis and AI time-to-event models, the CPU-Index framework offers a nuanced evaluation of the bias-variance trade-off and improves the transparency and reliability of predictions. Consistency was estimated by the concordance index of subgroup analysis-based similarity rank and model prediction similarity rank. Subgroup analysis-based similarity loss was defined as the sum-of-the-difference between Lung-RADS and feature-level 0-1 loss. Model prediction similarity loss was defined as squared loss. To test our approach, we identified 3,326 patients who underwent LDCT for LCS from 1/1/2015 to 6/30/2020 with confirmation of lung cancer on pathology within one year. For each LDCT image, the lesion associated with a Lung-RADS score was detected using a pretrained deep learning model from Medical Open Network for AI (MONAI), from which radiomic features were extracted. Radiomics were optimally fused with patient demographics via a positional encoding scheme and used to train a neural multi-task logistic regression time-to-event model that predicts malignancy. Performance was maximized when radiomics features were fused with positionally encoded demographic features. In this configuration, our algorithm raised the AUC from 0.81 ± 0.04 to 0.89 ± 0.02. Compared to standard Lung-RADS, our approach reduced the False-Positive-Rate from 0.41 ± 0.02 to 0.30 ± 0.12 while maintaining the same False-Negative-Rate. Our methodology enhances lung cancer risk assessment by estimating prediction uncertainty and adjusting accordingly. Furthermore, the optimal integration of radiomics and patient demographics improved overall diagnostic performance, indicating their complementary nature.

PMID:39721356 | DOI:10.1016/j.artmed.2024.103055

Int J Med Inform. 2024 Dec 19;195:105769. doi: 10.1016/j.ijmedinf.2024.105769. Online ahead of print.

ABSTRACT

BACKGROUND: The automated segmentation of individual teeth from 3D models of the human dental arch is challenging due to variations in tooth alignment, arch form and overall maxillofacial anatomy. Domain adaptation is a specialised technique in deep learning which allows models to adapt to data from different domains, such as varying tooth and dental arch forms, without requiring human annotations.

PURPOSE: This study aimed to segment individual teeth from various dental arch morphologies in 3D intraoral scans using domain adaptation.

MATERIALS AND METHODS: Twenty scanned dental arches from various age groups and developmental stages were used to generate 20 simplified synthetic variants of the scans. These synthetic variants, along with 16 natural scanned dental arches, were used to train the deep learning models. Domain adaptation was employed using Gradient Reversal Layer and Siamese Network techniques. The PointNet and PointNet++ model backbones were trained to align the latent space distribution of real and synthetic domains. Validations were performed on four unseen natural scanned arches, with and without domain adaptation enabled, to evaluate whether a 3D deep neural network can be trained without any human-annotated 3D models.

RESULTS: PointNet and PointNet++ models demonstrated a mean intersection-over-union between 0.34 and 0.36 mIoU without domain adaptation enabled and 0.80 and 0.95 mIoU, respectively with domain adaptation enabled when assessing natural scanned dental arches.

CONCLUSION: Domain adaptation techniques can enable training a segmentation deep learning model using synthetically generated 3D jaw scans without requiring human operators annotating the training data.

PMID:39721113 | DOI:10.1016/j.ijmedinf.2024.105769

Neural Netw. 2024 Dec 19;184:107074. doi: 10.1016/j.neunet.2024.107074. Online ahead of print.

ABSTRACT

The field of traffic forecasting has been the subject of considerable attention as a critical component in alleviating traffic congestion and improving urban services. Given the regular patterns of human activities, it is evident that traffic flow is inherently periodic. However, most of existing studies restrict themselves to recent historical observations and typically yield structurally and computationally complex models, which greatly limits the forecasting accuracy and hinders the application of models in realistic situations. To this end, this paper proposes a concise framework named Spatial-Temporal Gated Multi-Layer Perceptron (ST-GMLP), aiming to enhance the forecasting performance by leveraging the temporal patterns of different scales with a simple and effective structure. Nevertheless, due to the incorporation of more historical features, the presence of distribution shifts between periods further restricts the forecasting accuracy. To address the above issue, ST-GMLP employs a parallel structure of learning the interdependencies of traffic flow in both spatial node and temporal directions, and then establishes the interactions between time and space to effectively mitigate the adverse effects due to temporal distribution shifts. Owing to the utilization of MLP with gated mechanisms (GMLP) for modeling the spatial-temporal interdependencies, ST-GMLP has significant advantages in terms of training efficiency and resources occupation. Extensive experimental findings indicate that ST-GMLP exhibits superior performance in comparison to state-of-the-art methods.

PMID:39721105 | DOI:10.1016/j.neunet.2024.107074

Cell Physiol Biochem. 2024 Dec 19;58(6):739-755. doi: 10.33594/000000746.

ABSTRACT

BACKGROUND/AIMS: Alzheimer's Disease (AD) is a progressive neurodegenerative disorder that severely affects cognitive functions and memory. Early detection is crucial for timely intervention and improved patient outcomes. However, traditional diagnostic tools, such as MRI and PET scans, are costly and less accessible. This study aims to develop an automated, cost-effective digital diagnostic approach using deep learning (DL) and computer-aided detection (CAD) methods for early AD identification and classification.

METHODS: The proposed framework utilizes pretrained convolutional neural networks (CNNs) for feature extraction, integrated with two classifiers: multi-class support vector machine (MSVM) and artificial neural network (ANN). A dataset categorized into four groups-non-demented, very mild demented, mild demented, and moderate demented-was employed for evaluation. To optimize the classification process, a texture-based algorithm was applied for feature reduction, enhancing computational efficiency and reducing processing time.

RESULTS: The system demonstrated high statistical performance, achieving an accuracy of 91%, precision of 95%, and recall of 90%. Among the initial set of twenty-two texture features, seven were identified as particularly effective in differentiating normal cases from mild AD stages, significantly streamlining the classification process. These results validate the robustness and efficacy of the proposed DL-based CAD system.

CONCLUSION: This study presents a reliable and affordable solution for early AD detection and diagnosis. The proposed system outperforms existing state-of-the-art models and offers a valuable tool for timely treatment planning. Future research should explore its application to larger, more diverse datasets and investigate integration with other imaging modalities, such as MRI, to further enhance diagnostic precision.

PMID:39720940 | DOI:10.33594/000000746

Intern Med J. 2024 Dec 25. doi: 10.1111/imj.16607. Online ahead of print.

ABSTRACT

Enhancing patient comprehension of their health is crucial in improving health outcomes. The integration of artificial intelligence (AI) in distilling medical information into a conversational, legible format can potentially enhance health literacy. This review aims to examine the accuracy, reliability, comprehensiveness and readability of medical patient education materials (PEMs) simplified by AI models. A systematic review was conducted searching for articles assessing outcomes of use of AI in simplifying PEMs. Inclusion criteria are as follows: publication between January 2019 and June 2023, various modalities of AI, English language, AI use in PEMs and including physicians and/or patients. An inductive thematic approach was utilised to code for unifying topics which were qualitatively analysed. Twenty studies were included, and seven themes were identified (reproducibility, accessibility and ease of use, emotional support and user satisfaction, readability, data security, accuracy and reliability and comprehensiveness). AI effectively simplified PEMs, with reproducibility rates up to 90.7% in specific domains. User satisfaction exceeded 85% in AI-generated materials. AI models showed promising readability improvements, with ChatGPT achieving 100% post-simplification readability scores. AI's performance in accuracy and reliability was mixed, with occasional lack of comprehensiveness and inaccuracies, particularly when addressing complex medical topics. AI models accurately simplified basic tasks but lacked soft skills and personalisation. These limitations can be addressed with higher-calibre models combined with prompt engineering. In conclusion, the literature reveals a scope for AI to enhance patient health literacy through medical PEMs. Further refinement is needed to improve AI's accuracy and reliability, especially when simplifying complex medical information.

PMID:39720869 | DOI:10.1111/imj.16607

Phys Imaging Radiat Oncol. 2024 Dec 4;32:100684. doi: 10.1016/j.phro.2024.100684. eCollection 2024 Oct.

ABSTRACT

BACKGROUND AND PURPOSE: Segmentation imperfections (noise) in radiotherapy organ-at-risk segmentation naturally arise from specialist experience and image quality. Using clinical contours can result in sub-optimal convolutional neural network (CNN) training and performance, but manual curation is costly. We address the impact of simulated and clinical segmentation noise on CNN parotid gland (PG) segmentation performance and provide proof-of-concept for an easily implemented auto-curation countermeasure.

METHODS AND MATERIALS: The impact of segmentation imperfections was investigated by simulating noise in clean, high-quality segmentations. Curation efficacy was tested by removing lowest-scoring Dice similarity coefficient (DSC) cases early during CNN training, both in simulated (5-fold) and clinical (10-fold) settings, using our full radiotherapy clinical cohort (RTCC; N = 1750 individual PGs). Statistical significance was assessed using Bonferroni-corrected Wilcoxon signed-rank tests. Curation efficacies were evaluated using DSC and mean surface distance (MSD) on in-distribution and out-of-distribution data and visual inspection.

RESULTS: The curation step correctly removed median(range) 98(90-100)% of corrupted segmentations and restored the majority (1.2 %/1.3 %) of DSC lost from training with 30 % corrupted segmentations. This effect was masked when using typical (non-curated) validation data. In RTCC, 20 % curation showed improved model generalizability which significantly improved out-of-distribution DSC and MSD (p < 1.0e-12, p < 1.0e-6). Improved consistency was observed in particularly the medial and anterior lobes.

CONCLUSIONS: Up to 30% case removal, the curation benefit outweighed the training variance lost through curation. Considering the notable ease of implementation, high sensitivity in simulations and performance gains already at lower curation fractions, as a conservative middle ground, we recommend 15% curation of training cases when training CNNs using clinical PG contours.

PMID:39720784 | PMC:PMC11667007 | DOI:10.1016/j.phro.2024.100684

Healthc Technol Lett. 2024 Dec 13;11(6):365-373. doi: 10.1049/htl2.12105. eCollection 2024 Dec.

ABSTRACT

Deep learning models are used to minimize the number of polyps that goes unnoticed by the experts and to accurately segment the detected polyps during interventions. Although state-of-the-art models are proposed, it remains a challenge to define representations that are able to generalize well and that mediate between capturing low-level features and higher-level semantic details without being redundant. Another challenge with these models is that they are computation and memory intensive, which can pose a problem with real-time applications. To address these problems, PlutoNet is proposed for polyp segmentation which requires only 9 FLOPs and 2,626,537 parameters, less than 10% of the parameters required by its counterparts. With PlutoNet, a novel decoder consistency training approach is proposed that consists of a shared encoder, the modified partial decoder, which is a combination of the partial decoder and full-scale connections that capture salient features at different scales without redundancy, and the auxiliary decoder which focuses on higher-level semantic features. The modified partial decoder and the auxiliary decoder are trained with a combined loss to enforce consistency, which helps strengthen learned representations. Ablation studies and experiments are performed which show that PlutoNet performs significantly better than the state-of-the-art models, particularly on unseen datasets.

PMID:39720760 | PMC:PMC11665777 | DOI:10.1049/htl2.12105

Healthc Technol Lett. 2024 Oct 23;11(6):471-484. doi: 10.1049/htl2.12093. eCollection 2024 Dec.

ABSTRACT

Breast cancer is one of the most common causes of death in women in the modern world. Cancerous tissue detection in histopathological images relies on complex features related to tissue structure and staining properties. Convolutional neural network (CNN) models like ResNet50, Inception-V1, and VGG-16, while useful in many applications, cannot capture the patterns of cell layers and staining properties. Most previous approaches, such as stain normalization and instance-based vision transformers, either miss important features or do not process the whole image effectively. Therefore, a deep fusion-based vision Transformer model (DFViT) that combines CNNs and transformers for better feature extraction is proposed. DFViT captures local and global patterns more effectively by fusing RGB and stain-normalized images. Trained and tested on several datasets, such as BreakHis, breast cancer histology (BACH), and UCSC cancer genomics (UC), the results demonstrate outstanding accuracy, F1 score, precision, and recall, setting a new milestone in histopathological image analysis for diagnosing breast cancer.

PMID:39720758 | PMC:PMC11665795 | DOI:10.1049/htl2.12093

Healthc Technol Lett. 2024 Sep 15;11(6):452-460. doi: 10.1049/htl2.12091. eCollection 2024 Dec.

ABSTRACT

Missing Alzheimer's disease (AD) data is prevalent and poses significant challenges for AD diagnosis. Previous studies have explored various data imputation approaches on AD data, but the systematic evaluation of deep learning algorithms for imputing heterogeneous and comprehensive AD data is limited. This study investigates the efficacy of denoising autoencoder-based imputation of missing key features of heterogeneous data that comprised tau-PET, MRI, cognitive and functional assessments, genotype, sociodemographic, and medical history. The authors focused on extreme (≥40%) missing at random of key features which depend on AD progression; identified as the history of a mother having AD, APoE ε4 alleles, and clinical dementia rating. Along with features selected using traditional feature selection methods, latent features extracted from the denoising autoencoder are incorporated for subsequent classification. Using random forest classification with 10-fold cross-validation, robust AD predictive performance of imputed datasets (accuracy: 79%-85%; precision: 71%-85%) across missingness levels, and high recall values with 40% missingness are found. Further, the feature-selected dataset using feature selection methods, including autoencoder, demonstrated higher classification score than that of the original complete dataset. These results highlight the effectiveness and robustness of autoencoder in imputing crucial information for reliable AD prediction in AI-based clinical decision support systems.

PMID:39720752 | PMC:PMC11665783 | DOI:10.1049/htl2.12091

Front Neurorobot. 2024 Dec 10;18:1485640. doi: 10.3389/fnbot.2024.1485640. eCollection 2024.

ABSTRACT

Non-invasive brain-computer interfaces (BCI) hold great promise in the field of neurorehabilitation. They are easy to use and do not require surgery, particularly in the area of motor imagery electroencephalography (EEG). However, motor imagery EEG signals often have a low signal-to-noise ratio and limited spatial and temporal resolution. Traditional deep neural networks typically only focus on the spatial and temporal features of EEG, resulting in relatively low decoding and accuracy rates for motor imagery tasks. To address these challenges, this paper proposes a 3D Convolutional Neural Network (P-3DCNN) decoding method that jointly learns spatial-frequency feature maps from the frequency and spatial domains of the EEG signals. First, the Welch method is used to calculate the frequency band power spectrum of the EEG, and a 2D matrix representing the spatial topology distribution of the electrodes is constructed. These spatial-frequency representations are then generated through cubic interpolation of the temporal EEG data. Next, the paper designs a 3DCNN network with 1D and 2D convolutional layers in series to optimize the convolutional kernel parameters and effectively learn the spatial-frequency features of the EEG. Batch normalization and dropout are also applied to improve the training speed and classification performance of the network. Finally, through experiments, the proposed method is compared to various classic machine learning and deep learning techniques. The results show an average decoding accuracy rate of 86.69%, surpassing other advanced networks. This demonstrates the effectiveness of our approach in decoding motor imagery EEG and offers valuable insights for the development of BCI.

PMID:39720668 | PMC:PMC11667157 | DOI:10.3389/fnbot.2024.1485640

Infect Dis Model. 2024 Dec 3;10(1):353-364. doi: 10.1016/j.idm.2024.12.002. eCollection 2025 Mar.

ABSTRACT

Acute flaccid paralysis (AFP) case surveillance is pivotal for the early detection of potential poliovirus, particularly in endemic countries such as Ethiopia. The community-based surveillance system implemented in Ethiopia has significantly improved AFP surveillance. However, challenges like delayed detection and disorganized communication persist. This work proposes a simple deep learning model for AFP surveillance, leveraging transfer learning on images collected from Ethiopia's community key informants through mobile phones. The transfer learning approach is implemented using a vision transformer model pretrained on the ImageNet dataset. The proposed model outperformed convolutional neural network-based deep learning models and vision transformer models trained from scratch, achieving superior accuracy, F1-score, precision, recall, and area under the receiver operating characteristic curve (AUC). It emerged as the optimal model, demonstrating the highest average AUC of 0.870 ± 0.01. Statistical analysis confirmed the significant superiority of the proposed model over alternative approaches (P < 0.001). By bridging community reporting with health system response, this study offers a scalable solution for enhancing AFP surveillance in low-resource settings. The study is limited in terms of the quality of image data collected, necessitating future work on improving data quality. The establishment of a dedicated platform that facilitates data storage, analysis, and future learning can strengthen data quality. Nonetheless, this work represents a significant step toward leveraging artificial intelligence for community-based AFP surveillance from images, with substantial implications for addressing global health challenges and disease eradication strategies.

PMID:39720666 | PMC:PMC11666939 | DOI:10.1016/j.idm.2024.12.002

J Pathol Inform. 2024 Nov 17;16:100410. doi: 10.1016/j.jpi.2024.100410. eCollection 2025 Jan.

ABSTRACT

Classifying breast cancer molecular subtypes is crucial for tailoring treatment strategies. While immunohistochemistry (IHC) and gene expression profiling are standard methods for molecular subtyping, IHC can be subjective, and gene profiling is costly and not widely accessible in many regions. Previous approaches have highlighted the potential application of deep learning models on hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) for molecular subtyping, but these efforts vary in their methods, datasets, and reported performance. In this work, we investigated whether H&E-stained WSIs could be solely leveraged to predict breast cancer molecular subtypes (luminal A, B, HER2-enriched, and Basal). We used 1433 WSIs of breast cancer in a two-step pipeline: first, classifying tumor and non-tumor tiles to use only the tumor regions for molecular subtyping; and second, employing a One-vs-Rest (OvR) strategy to train four binary OvR classifiers and aggregating their results using an eXtreme Gradient Boosting model. The pipeline was tested on 221 hold-out WSIs, achieving an F1 score of 0.95 for tumor vs non-tumor classification and a macro F1 score of 0.73 for molecular subtyping. Our findings suggest that, with further validation, supervised deep learning models could serve as supportive tools for molecular subtyping in breast cancer. Our codes are made available to facilitate ongoing research and development.

PMID:39720418 | PMC:PMC11667687 | DOI:10.1016/j.jpi.2024.100410

Water Res X. 2024 Dec 3;26:100291. doi: 10.1016/j.wroa.2024.100291. eCollection 2025 Jan 1.

ABSTRACT

Sudden shocking load events featuring significant increases in inflow quantities or concentrations of wastewater treatment plants (WWTPs), are a major threat to the attainment of treated effluents to discharge quality standards. To aid in real-time decision-making for stable WWTP operations, this study developed a probabilistic deep learning model that comprises encoder-decoder long short-term memory (LSTM) networks with added capacity of producing probability predictions, to enhance the robustness of real-time WWTP effluent quality prediction under such events. The developed probabilistic encoder-decoder LSTM (P-ED-LSTM) model was tested in an actual WWTP, where bihourly effluent quality prediction of total nitrogen was performed and compared with classical deep learning models, including LSTM, gated recurrent unit (GRU) and Transformer. It was found that under shocking load events, the P-ED-LSTM could achieve a 49.7% improvement in prediction accuracy for bihourly real-time predictions of effluent concentration compared to the LSTM, GRU, and Transformer. A higher quantile of the probability data from the P-ED-LSTM model output, indicated a prediction value more approximate to real effluent quality. The P-ED-LSTM model also exhibited higher predictive power for the next multiple time steps with shocking load scenarios. It captured approximately 90% of the actual over-limit discharges up to 6 hours ahead, significantly outperforming other deep learning models. Therefore, the P-ED-LSTM model, with its robust adaptability to significant fluctuations, has the potential for broader applications across WWTPs with different processes, as well as providing strategies for wastewater system regulation under emergency conditions.

PMID:39720317 | PMC:PMC11667701 | DOI:10.1016/j.wroa.2024.100291

Health Sci Rep. 2024 Dec 24;7(12):e70300. doi: 10.1002/hsr2.70300. eCollection 2024 Dec.

ABSTRACT

BACKGROUND AND AIMS: The rapid expansion of artificial intelligence (AI) within worldwide healthcare systems is occurring at a significant rate. In this context, the Middle East has demonstrated distinctive characteristics in the application of AI within the healthcare sector, particularly shaped by regional policies. This study examined the outcomes resulting from the utilization of AI within healthcare systems in the Middle East.

METHODS: A systematic review was conducted across several databases, including PubMed, Scopus, ProQuest, and the Cochrane Database of Systematic Reviews in 2024. The quality assessment of the included studies was conducted using the Authority, Accuracy, Coverage, Objectivity, Date, Significance checklist. Following this, a thematic analysis was carried out on the acquired data, adhering to the Boyatzis approach.

RESULTS: 100 papers were included. The quality and bias risk of the included studies were delineated to be within an acceptable range. Multiple themes were derived from the thematic analysis including: "Prediction of diseases, their diagnosis, and outcomes," "Prediction of organizational issues and attributes," "Prediction of mental health issues and attributes," "Prediction of polypharmacy and emotional analysis of texts," "Prediction of climate change issues and attributes," and "Prediction and identification of success and satisfaction among healthcare individuals."

CONCLUSION: The findings emphasized AI's significant potential in addressing prevalent healthcare challenges in the Middle East, such as cancer, diabetes, and climate change. AI has the potential to overhaul the healthcare systems. The findings also highlighted the need for policymakers and administrators to develop a concrete plan to effectively integrate AI into healthcare systems.

PMID:39720235 | PMC:PMC11667773 | DOI:10.1002/hsr2.70300