Nat Commun. 2025 Mar 8;16(1):2342. doi: 10.1038/s41467-025-57721-w.

ABSTRACT

High-quality nuclear magnetic resonance (NMR) spectra can be rapidly acquired by combining non-uniform sampling techniques (NUS) with reconstruction algorithms. However, current deep learning (DL) based reconstruction methods focus only on single-domain reconstruction (time or frequency domain), leading to drawbacks like peak loss and artifact peaks and ultimately failing to achieve optimal performance. Moreover, the lack of fully sampled spectra makes it difficult, even impossible, to determine the quality of reconstructed spectra, presenting challenges in the practical applications of NUS. In this study, a joint time-frequency domain deep learning network, referred to as JTF-Net, is proposed. It effectively combines time domain and frequency domain features, exhibiting better reconstruction performance on protein spectra across various dimensions compared to traditional algorithms and single-domain DL methods. In addition, the reference-free quality assessment metric, denoted as REconstruction QUalIty assuRancE Ratio (REQUIRER), is proposed base on an established quality space in the field of NMR spectral reconstruction. The metric is capable of evaluating the quality of reconstructed NMR spectra without the fully sampled spectra, making it more suitable for practical applications.

PMID:40057512 | DOI:10.1038/s41467-025-57721-w

Respir Res. 2025 Mar 8;26(1):91. doi: 10.1186/s12931-025-03173-1.

ABSTRACT

BACKGROUND: High altitude pulmonary edema (HAPE) poses a significant medical challenge to individuals ascending rapidly to high altitudes. Hypoxia-induced cellular morphological changes in the alveolar-capillary barrier such as mitochondrial structural alterations and cytoskeletal reorganization, play a crucial role in the pathogenesis of HAPE. These morphological changes are critical in understanding the cellular response to hypoxia and represent potential therapeutic targets. However, there is still a lack of effective and valid drug discovery strategies for anti-HAPE treatments based on these cellular morphological features. This study aims to develop a pipeline that focuses on morphological alterations in Cell Painting images to identify potential therapeutic agents for HAPE interventions.

METHODS: We generated over 100,000 full-field Cell Painting images of human alveolar adenocarcinoma basal epithelial cells (A549s) and human pulmonary microvascular endothelial cells (HPMECs) under different hypoxic conditions (1%~5% of oxygen content). These images were then submitted to our newly developed segmentation network (SegNet), which exhibited superior performance than traditional methods, to proceed to subcellular structure detection and segmentation. Subsequently, we created a hypoxia scoring network (HypoNet) using over 200,000 images of subcellular structures from A549s and HPMECs, demonstrating outstanding capacity in identifying cellular hypoxia status.

RESULTS: We proposed a deep neural network-based drug screening pipeline (CPHNet), which facilitated the identification of two promising natural products, ferulic acid (FA) and resveratrol (RES). Both compounds demonstrated satisfactory anti-HAPE effects in a 3D-alveolus chip model (ex vivo) and a mouse model (in vivo).

CONCLUSION: This work provides a brand-new and effective pipeline for screening anti-HAPE agents by integrating artificial intelligence (AI) tools and Cell Painting, offering a novel perspective for AI-driven phenotypic drug discovery.

PMID:40057746 | DOI:10.1186/s12931-025-03173-1

Respir Res. 2025 Mar 8;26(1):94. doi: 10.1186/s12931-025-03174-0.

ABSTRACT

BACKGROUND: To evaluate the clinical applicability of deep learning (DL) models based on automatic segmentation in preoperatively predicting tumor spread through air spaces (STAS) in peripheral stage I lung adenocarcinoma (LUAD).

METHODS: This retrospective study analyzed data from patients who underwent surgical treatment for lung tumors from January 2022 to December 2023. An external validation set was introduced to assess the model's generalizability. The study utilized conventional radiomic features and DL models for comparison. ROI segmentation was performed using the VNet architecture, and DL models were developed with transfer learning and optimization techniques. We assessed the diagnostic accuracy of our models via calibration curves, decision curve analysis, and ROC curves.

RESULTS: The DL model based on automatic segmentation achieved an AUC of 0.880 (95% CI 0.780-0.979), outperforming the conventional radiomics model with an AUC of 0.833 (95% CI 0.707-0.960). The DL model demonstrated superior performance in both internal validation and external testing cohorts. Calibration curves, decision curve analysis, and ROC curves confirmed the enhanced diagnostic accuracy and clinical utility of the DL approach.

CONCLUSION: The DL model based on automatic segmentation technology shows significant promise in preoperatively predicting STAS in peripheral stage I LUAD, surpassing traditional radiomics models in diagnostic accuracy and clinical applicability. Clinical trial number The clinical trial was registered on April 22, 2024, with the registration number researchregistry10213 ( www.researchregistry.com ).

PMID:40057743 | DOI:10.1186/s12931-025-03174-0

BMC Pediatr. 2025 Mar 8;25(1):177. doi: 10.1186/s12887-025-05532-9.

ABSTRACT

BACKGROUND: Turner syndrome (TS) is one of the important causes of short stature in girls, but there are cases of misdiagnosis and missed diagnosis in clinical practice. Our aim is to analyze the hand skeletal characteristics of TS patients and establish a disease screening model using deep learning.

METHODS: A total of 101 pediatric patients with TS were included in this retrospective case-control study. Their radiation parameters from hand X-rays were summarized and compared. Receiver operating characteristic (ROC) curves for parameters with differences between the groups were plotted. Additionally, we used deep learning networks to establish a predictive model.

RESULTS: Four parameters were identified as having diagnostic value for TS: the length ratio of metacarpal IV and metacarpal III, the distance between ulnoradial tangents, the carpal angle, and the ulnar-radial angle. When the cutoff value of the distance between the ulnoradial tangents was 0.40 cm, the specificity reached 92.57%. And for the ulnar- radius angle, according to the ROC analysis, the maximum value of Youden's index was obtained when the cut-off value was 170°, with a sensitivity of 66.34% and specificity of 61.38%. The ResNet50 deep neural network architecture was utilized, resulting in an accuracy of 78.89%, specificity of 76.67%, and sensitivity of 83.33% on a test dataset.

CONCLUSIONS: We propose that certain hand radiograph parameters have the potential to serve as diagnostic indicators for TS. The utilization of deep learning models has significantly enhanced the precision of disease diagnosis.

PMID:40057693 | DOI:10.1186/s12887-025-05532-9

Sci Rep. 2025 Mar 8;15(1):8114. doi: 10.1038/s41598-025-92776-1.

ABSTRACT

In the diagnosis and treatment of brain tumors, the automatic classification and segmentation of medical images play a pivotal role. Early detection facilitates timely intervention, significantly improving patient survival rates. This study introduces a novel method for the automated classification and segmentation of brain tumors, aiming to enhance both diagnostic accuracy and efficiency. Magnetic Resonance (MR) imaging remains the gold standard in clinical brain tumor diagnostics; however, it is a time-intensive and labor-intensive process. Consequently, the integration of automated detection, localization, and classification methods is not only desirable but essential. In this research, we present a novel framework that enables both tumor classification and post-classification diagnostic feature extraction, allowing for the first-time classification of multiple tumor types. To improve tumor characterization, we applied data augmentation techniques to MR images and developed a hierarchical multiscale deformable attention module (MS-DAM). This model effectively captures irregular and complex tumor patterns, enhancing classification performance. Following classification, a comprehensive segmentation process was conducted across a large dataset, reinforcing the model's role as a decision support system. Utilizing a Kaggle dataset containing 14 different tumor types with highly similar morphologic structures, we validated the proposed model's efficacy. Compared to existing multi-scale channel attention modules, MS-DAM achieved superior accuracy, exceeding 96.5%. This study presents a highly promising approach for the automated classification and segmentation of brain tumors in medical imaging, offering significant advancements for diagnostic imaging clinics and paving the way for more efficient, accurate, and scalable tumor detection methodologies.

PMID:40057634 | DOI:10.1038/s41598-025-92776-1

Sci Rep. 2025 Mar 8;15(1):8086. doi: 10.1038/s41598-025-85678-9.

ABSTRACT

Lung cancer remains a major global health challenge, and accurate pathological examination is crucial for early detection. This study aims to enhance hyperspectral pathological image analysis by refining annotations at the cell level and creating a high-quality hyperspectral dataset of lung tumors. We address the challenge of coarse manual annotations in hyperspectral lung cancer datasets, which limit the effectiveness of deep learning models requiring precise labels for training. We propose a semi-automated annotation refinement method that leverages hyperspectral data to enhance pathological diagnosis. Specifically, we employ K-means unsupervised clustering combined with human-guided selection to refine coarse annotations into cell-level masks based on spectral features. Our method is validated using a hyperspectral lung squamous cell carcinoma dataset containing 65 image samples. Experimental results demonstrate that our approach improves pixel-level segmentation accuracy from 77.33% to 92.52% with a lower level of prediction noise. The time required to accurately label each pathological slide is significantly reduced. While pixel-level labeling methods for an entire slide can take over 30 mins, our semi-automated method requires only about 5 mins. To enhance visualization for pathologists, we apply a conservative post-processing strategy for instance segmentation. These results highlight the effectiveness of our method in addressing annotation challenges and improving the accuracy of hyperspectral pathological analysis.

PMID:40057531 | DOI:10.1038/s41598-025-85678-9

Nucleic Acids Res. 2025 Feb 27;53(5):gkaf158. doi: 10.1093/nar/gkaf158.

ABSTRACT

Spatial transcriptomics technology has revolutionized our understanding of cellular systems by capturing RNA transcript levels in their original spatial context. Single-cell spatial transcriptomics (scST) offers single-cell resolution expression level and precise spatial information of RNA transcripts, while it has a limited capacity for simultaneously detecting a wide range of RNA transcripts, hindering its broader applications. Characterizing the whole transcriptome level and comprehensively annotating cell types represent two significant challenges in scST applications. Despite several proposed methods for one or both tasks, their performance remains inadequate. In this work, we introduce stAI, a deep learning-based model designed to address both missing gene imputation and cell-type annotation for scST data. stAI leverages a joint embedding for the scST and the reference scRNA-seq data with two separate encoder-decoder modules. Both the imputation and annotation are performed within the latent space in a supervised manner, utilizing scRNA-seq data to guide the processes. Experiments for datasets generated from diverse platforms with varying numbers of measured genes were conducted and compared with the updated methods. The results demonstrate that stAI can predict the unmeasured genes, especially the marker genes, with much higher accuracy, and annotate the cell types, including those of small size, with high precision.

PMID:40057378 | DOI:10.1093/nar/gkaf158

Med Eng Phys. 2025 Mar;137:104302. doi: 10.1016/j.medengphy.2025.104302. Epub 2025 Feb 6.

ABSTRACT

Cardiovascular diseases (CVDs) are the leading global cause of death, which requires the early and accurate detection of cardiac abnormalities. Abnormal heart sounds, indicative of potential cardiac problems, pose a challenge due to their low-frequency nature. Utilizing digital signal processing and Phonocardiogram (PCG) analysis, this study employs advanced deep learning techniques for automated heart sound classification. Time-frequency representations capture multiple heart sound features, including gammatonegram, Mel-spectrogram, and Constant-Q Transform (CQT). A Convolutional Neural Network with Directed Acyclic Graph (DAG-CNN) architecture is designed and rigorously evaluated, achieving high classification accuracies of 100%, 99.7%, and 99.5% for gammatonegram, Mel-spectrogram, and CQT, respectively. Comparative analysis with pre-trained CNN models demonstrates the superior performance of the proposed model. This advancement in automated heart sound classification offers a promising and cost-effective tool for early diagnosis, particularly in resource-limited settings, helping to address the diagnostic gap and enhance cardiac care accessibility.

PMID:40057368 | DOI:10.1016/j.medengphy.2025.104302

Med Eng Phys. 2025 Mar;137:104305. doi: 10.1016/j.medengphy.2025.104305. Epub 2025 Feb 7.

ABSTRACT

Lung cancer remains the principal cause of cancer-related deaths. Nodules are the main radiological finding, typically observed from low-dose CT scans. Nonetheless, the nodule characterization diagnosis remains subjective, reporting a moderate agreement among experts' observations, especially in identifying malignancy stratification. The proposed approach presents a deep multi-attention strategy, validated exhaustively to classify nodule masses according to four malignancy degrees. This work introduces a multi-attention architecture dedicated to stratifying nodules among malignancy stages. The architecture receives volumetric nodule regions and learns multi-scale saliency maps, focusing on determinant malignancy patterns of the observed masses. Specialized attention heads capture related patterns associated with lobulated, textural, and spiculated features. Validation includes an extensive analysis regarding multiple attention features, allowing to establish a correlation with other radiological findings. The proposed approach achieves an AUC of 85.35% for a classical multi-classification and a mean AUC of 82.90% in a one-vs-all validation methodology, showing competitive results in the state-of-the-art. The introduced architecture has capabilities to support nodule stratification and to classify nodule features. The exhaustive validation also suggests a proper generalization performance, which is a potential property to transfer this strategy in real scenarios.

PMID:40057364 | DOI:10.1016/j.medengphy.2025.104305

Med Eng Phys. 2025 Mar;137:104300. doi: 10.1016/j.medengphy.2025.104300. Epub 2025 Feb 8.

ABSTRACT

Tuberculosis is a high-mortality infectious disease. Manual sputum smear microscopy is a common and effective method for screening tuberculosis. However, it is time-consuming, labor-intensive, and has low sensitivity. In this study, we propose ResGloTBNet, a framework that integrates convolutional neural network and graph convolutional network for sputum smear image classification with high discriminative power. In this framework, the global reasoning unit is introduced into the residual structure of ResNet to form the ResGloRe module, which not only fully extracts the local features of the image but also models the global relationship between different regions in the image. Furthermore, we applied activation maximization and class activation mapping to generate explanations for the model's predictions on the test sets. ResGloTBNet achieved remarkable results on a publicly available dataset, reaching 97.2 % accuracy and 99.0 % sensitivity. It also maintained a high level of performance on a private dataset, attaining 98.0 % accuracy and 96.6 % sensitivity. In addition, interpretable analysis demonstrated that ResGloTBNet can effectively identify the features and regions in the input images that contribute the most to the model's predictions, providing valuable insights into the decision-making process of the deep learning model.

PMID:40057359 | DOI:10.1016/j.medengphy.2025.104300

SLAS Technol. 2025 Mar 6:100264. doi: 10.1016/j.slast.2025.100264. Online ahead of print.

ABSTRACT

Gastrointestinal cancer represents a significant health burden, necessitating innovative approaches for personalized treatment. This study aims to develop an advanced natural language processing (NLP)-driven predictive modeling framework for tailored treatment strategies in gastrointestinal cancer, leveraging the capabilities of deep learning. The Resilient Adam Algorithm-driven Versatile Long-Short Term Memory (RAA-VLSTM) model is proposed to analyze comprehensive clinical data. The dataset comprises extensive electronic health records (EHRs) from multiple healthcare centers, focusing on patient demographics, clinical history, treatment outcomes, and genetic factors. Data preprocessing employs techniques such as tokenization, normalization, and stop-word removal to ensure effective representation of textual data. For feature extraction, state-of-the-art word embeddings are utilized to enhance model performance. The proposed framework outlines a comprehensive process: data collection from EHRs, preprocessing to prepare the data for analysis, and employing NLP techniques to extract meaningful features. The RAA optimization algorithm significantly improves training efficiency by adapting learning rates for each parameter, addressing common issues in gradient descent. This optimization enhances feature learning from sequential clinical data, enabling accurate predictions of treatment responses and outcomes. The overall performance in terms of F1-score (89.4%), accuracy (92.5%), recall (88.7%), and precision (90.1%). Preliminary results demonstrate the model's strong predictive capabilities, achieving high accuracy in predicting treatment outcomes, thereby suggesting its potential to improve individualized care. In conclusion, this study establishes a robust foundation for employing advanced NLP and machine learning techniques in the management of gastrointestinal cancer, paving the way for future research and clinical applications.

PMID:40057234 | DOI:10.1016/j.slast.2025.100264

J Neurosurg. 2025 Mar 7:1-9. doi: 10.3171/2024.10.JNS241060. Online ahead of print.

ABSTRACT

OBJECTIVE: Preoperative workup of trigeminal neuralgia (TN) consists of identification of neurovascular features on MRI. In this study, the authors apply and evaluate the performance of deep learning models for segmentation of the trigeminal nerve and surrounding vasculature to quantify anatomical features of the nerve and vessels.

METHODS: Six U-Net-based neural networks, each with a different encoder backbone, were trained to label constructive interference in steady-state MRI voxels as nerve, vasculature, or background. A retrospective dataset of 50 TN patients at the authors' institution who underwent preoperative high-resolution MRI in 2022 was utilized to train and test the models. Performance was measured by the Dice coefficient and intersection over union (IoU) metrics. Anatomical characteristics, such as surface area of neurovascular contact and distance to the contact point, were computed and compared between the predicted and ground truth segmentations.

RESULTS: Of the evaluated models, the best performing was U-Net with an SE-ResNet50 backbone (Dice score = 0.775 ± 0.015, IoU score = 0.681 ± 0.015). When the SE-ResNet50 backbone was used, the average surface area of neurovascular contact in the testing dataset was 6.90 mm2, which was not significantly different from the surface area calculated from manual segmentation (p = 0.83). The average calculated distance from the brainstem to the contact point was 4.34 mm, which was also not significantly different from manual segmentation (p = 0.29).

CONCLUSIONS: U-Net-based neural networks perform well for segmenting trigeminal nerve and vessels from preoperative MRI volumes. This technology enables the development of quantitative and objective metrics for radiographic evaluation of TN.

PMID:40053933 | DOI:10.3171/2024.10.JNS241060

Mol Pharm. 2025 Mar 7. doi: 10.1021/acs.molpharmaceut.4c01086. Online ahead of print.

ABSTRACT

The optimization of absorption, distribution, metabolism, and excretion (ADME) profiles of compounds is critical to the drug discovery process. As such, machine learning (ML) models for ADME are widely used for prioritizing the design and synthesis of compounds. The effectiveness of ML models for ADME depends on the availability of high-quality experimental data for a diverse set of compounds that is relevant to the emerging chemical space being explored by the drug discovery teams. To that end, ADME data sets from Genentech and Roche were combined to evaluate the impact of expanding the chemical space on the performance of ML models, a first experiment of its kind for large-scale, historical ADME data sets. The combined ADME data set consisted of over 1 million individual measurements distributed across 11 assay end points. We utilized a multitask (MT) neural network architecture that enables the modeling of multiple end points simultaneously and thereby exploits information transfer between interconnected ADME end points. Both single- and cross-site MT models were trained and compared against single-site, single-task baseline models. Given the differences in assay protocols across the two sites, the data for corresponding end points across sites were modeled as separate tasks. Models were evaluated against test sets representing varying degrees of extrapolation difficulty, including cluster-based, temporal, and external test sets. We found that cross-site MT models appeared to provide a greater generalization capacity compared to single-site models. The performance improvement of the cross-site MT models was more pronounced for the relatively "distant" external and temporal test sets, suggesting an expanded applicability domain. The data exchange exercise described here demonstrates the value of expanding the learning from ADME data from multiple sources without the need to aggregate such data when the experimental methods are disparate.

PMID:40053846 | DOI:10.1021/acs.molpharmaceut.4c01086

JMIR Med Inform. 2025 Mar 4;13:e68863. doi: 10.2196/68863.

ABSTRACT

BACKGROUND: Patients' oral expressions serve as valuable sources of clinical information to improve pharmacotherapy. Natural language processing (NLP) is a useful approach for analyzing unstructured text data, such as patient narratives. However, few studies have focused on using NLP for narratives in the Japanese language.

OBJECTIVE: We aimed to develop a high-performance NLP system for extracting clinical information from patient narratives by examining the performance progression with a gradual increase in the amount of training data.

METHODS: We used subjective texts from the pharmaceutical care records of Keio University Hospital from April 1, 2018, to March 31, 2019, comprising 12,004 records from 6559 cases. After preprocessing, we annotated diseases and symptoms within the texts. We then trained and evaluated a deep learning model (bidirectional encoder representations from transformers combined with a conditional random field [BERT-CRF]) through 10-fold cross-validation. The annotated data were divided into 10 subsets, and the amount of training data was progressively increased over 10 steps. We also analyzed the causes of errors. Finally, we applied the developed system to the analysis of case report texts to evaluate its usability for texts from other sources.

RESULTS: The F1-score of the system improved from 0.67 to 0.82 as the amount of training data increased from 1200 to 12,004 records. The F1-score reached 0.78 with 3600 records and was largely similar thereafter. As performance improved, errors from incorrect extractions decreased significantly, which resulted in an increase in precision. For case reports, the F1-score also increased from 0.34 to 0.41 as the training dataset expanded from 1200 to 12,004 records. Performance was lower for extracting symptoms from case report texts compared with pharmaceutical care records, suggesting that this system is more specialized for analyzing subjective data from pharmaceutical care records.

CONCLUSIONS: We successfully developed a high-performance system specialized in analyzing subjective data from pharmaceutical care records by training a large dataset, with near-complete saturation of system performance with about 3600 training records. This system will be useful for monitoring symptoms, offering benefits for both clinical practice and research.

PMID:40053805 | DOI:10.2196/68863

J Med Internet Res. 2025 Mar 7;27:e69068. doi: 10.2196/69068.

ABSTRACT

BACKGROUND: Tuberculosis (TB) remains a significant health concern, contributing to the highest mortality among infectious diseases worldwide. However, none of the various TB diagnostic tools introduced is deemed sufficient on its own for the diagnostic pathway, so various artificial intelligence (AI)-based methods have been developed to address this issue.

OBJECTIVE: We aimed to provide a comprehensive evaluation of AI-based algorithms for TB detection across various data modalities.

METHODS: Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) 2020 guidelines, we conducted a systematic review to synthesize current knowledge on this topic. Our search across 3 major databases (Scopus, PubMed, Association for Computing Machinery [ACM] Digital Library) yielded 1146 records, of which we included 152 (13.3%) studies in our analysis. QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies version 2) was performed for the risk-of-bias assessment of all included studies.

RESULTS: Radiographic biomarkers (n=129, 84.9%) and deep learning (DL; n=122, 80.3%) approaches were predominantly used, with convolutional neural networks (CNNs) using Visual Geometry Group (VGG)-16 (n=37, 24.3%), ResNet-50 (n=33, 21.7%), and DenseNet-121 (n=19, 12.5%) architectures being the most common DL approach. The majority of studies focused on model development (n=143, 94.1%) and used a single modality approach (n=141, 92.8%). AI methods demonstrated good performance in all studies: mean accuracy=91.93% (SD 8.10%, 95% CI 90.52%-93.33%; median 93.59%, IQR 88.33%-98.32%), mean area under the curve (AUC)=93.48% (SD 7.51%, 95% CI 91.90%-95.06%; median 95.28%, IQR 91%-99%), mean sensitivity=92.77% (SD 7.48%, 95% CI 91.38%-94.15%; median 94.05% IQR 89%-98.87%), and mean specificity=92.39% (SD 9.4%, 95% CI 90.30%-94.49%; median 95.38%, IQR 89.42%-99.19%). AI performance across different biomarker types showed mean accuracies of 92.45% (SD 7.83%), 89.03% (SD 8.49%), and 84.21% (SD 0%); mean AUCs of 94.47% (SD 7.32%), 88.45% (SD 8.33%), and 88.61% (SD 5.9%); mean sensitivities of 93.8% (SD 6.27%), 88.41% (SD 10.24%), and 93% (SD 0%); and mean specificities of 94.2% (SD 6.63%), 85.89% (SD 14.66%), and 95% (SD 0%) for radiographic, molecular/biochemical, and physiological types, respectively. AI performance across various reference standards showed mean accuracies of 91.44% (SD 7.3%), 93.16% (SD 6.44%), and 88.98% (SD 9.77%); mean AUCs of 90.95% (SD 7.58%), 94.89% (SD 5.18%), and 92.61% (SD 6.01%); mean sensitivities of 91.76% (SD 7.02%), 93.73% (SD 6.67%), and 91.34% (SD 7.71%); and mean specificities of 86.56% (SD 12.8%), 93.69% (SD 8.45%), and 92.7% (SD 6.54%) for bacteriological, human reader, and combined reference standards, respectively. The transfer learning (TL) approach showed increasing popularity (n=89, 58.6%). Notably, only 1 (0.7%) study conducted domain-shift analysis for TB detection.

CONCLUSIONS: Findings from this review underscore the considerable promise of AI-based methods in the realm of TB detection. Future research endeavors should prioritize conducting domain-shift analyses to better simulate real-world scenarios in TB detection.

TRIAL REGISTRATION: PROSPERO CRD42023453611; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023453611.

PMID:40053773 | DOI:10.2196/69068

J Med Internet Res. 2025 Mar 5;27:e54543. doi: 10.2196/54543.

ABSTRACT

BACKGROUND: Chronic obstructive pulmonary disease (COPD) ranks among the leading causes of global mortality, and COVID-19 has intensified its challenges. Beyond the evident physical effects, the long-term psychological effects of COVID-19 are not fully understood.

OBJECTIVE: This study aims to unveil the long-term psychological trends and patterns in populations with COPD throughout the COVID-19 pandemic and beyond via large-scale Twitter mining.

METHODS: A 2-stage deep learning framework was designed in this study. The first stage involved a data retrieval procedure to identify COPD and non-COPD users and to collect their daily tweets. In the second stage, a data mining procedure leveraged various deep learning algorithms to extract demographic characteristics, hashtags, topics, and sentiments from the collected tweets. Based on these data, multiple analytical methods, namely, odds ratio (OR), difference-in-difference, and emotion pattern methods, were used to examine the psychological effects.

RESULTS: A cohort of 15,347 COPD users was identified from the data that we collected in the Twitter database, comprising over 2.5 billion tweets, spanning from January 2020 to June 2023. The attentiveness toward COPD was significantly affected by gender, age, and occupation; it was lower in females (OR 0.91, 95% CI 0.87-0.94; P<.001) than in males, higher in adults aged 40 years and older (OR 7.23, 95% CI 6.95-7.52; P<.001) than in those younger than 40 years, and higher in individuals with lower socioeconomic status (OR 1.66, 95% CI 1.60-1.72; P<.001) than in those with higher socioeconomic status. Across the study duration, COPD users showed decreasing concerns for COVID-19 and increasing health-related concerns. After the middle phase of COVID-19 (July 2021), a distinct decrease in sentiments among COPD users contrasted sharply with the upward trend among non-COPD users. Notably, in the post-COVID era (June 2023), COPD users showed reduced levels of joy and trust and increased levels of fear compared to their levels of joy and trust in the middle phase of COVID-19. Moreover, males, older adults, and individuals with lower socioeconomic status showed heightened fear compared to their counterparts.

CONCLUSIONS: Our data analysis results suggest that populations with COPD experienced heightened mental stress in the post-COVID era. This underscores the importance of developing tailored interventions and support systems that account for diverse population characteristics.

PMID:40053739 | DOI:10.2196/54543

J Med Internet Res. 2025 Feb 28;27:e67576. doi: 10.2196/67576.

ABSTRACT

BACKGROUND: In-hospital cardiac arrest (IHCA) is a severe and sudden medical emergency that is characterized by the abrupt cessation of circulatory function, leading to death or irreversible organ damage if not addressed immediately. Emergency department (ED)-based IHCA (EDCA) accounts for 10% to 20% of all IHCA cases. Early detection of EDCA is crucial, yet identifying subtle signs of cardiac deterioration is challenging. Traditional EDCA prediction methods primarily rely on structured vital signs or electrocardiogram (ECG) signals, which require additional preprocessing or specialized devices. This study introduces a novel approach using image-based 12-lead ECG data obtained at ED triage, leveraging the inherent richness of visual ECG patterns to enhance prediction and integration into clinical workflows.

OBJECTIVE: This study aims to address the challenge of early detection of EDCA by developing an innovative deep learning model, the ECG-Image-Aware Network (EIANet), which uses 12-lead ECG images for early prediction of EDCA. By focusing on readily available triage ECG images, this research seeks to create a practical and accessible solution that seamlessly integrates into real-world ED workflows.

METHODS: For adult patients with EDCA (cases), 12-lead ECG images at ED triage were obtained from 2 independent data sets: National Taiwan University Hospital (NTUH) and Far Eastern Memorial Hospital (FEMH). Control ECGs were randomly selected from adult ED patients without cardiac arrest during the same study period. In EIANet, ECG images were first converted to binary form, followed by noise reduction, connected component analysis, and morphological opening. A spatial attention module was incorporated into the ResNet50 architecture to enhance feature extraction, and a custom binary recall loss (BRLoss) was used to balance precision and recall, addressing slight data set imbalance. The model was developed and internally validated on the NTUH-ECG data set and was externally validated on an independent FEMH-ECG data set. The model performance was evaluated using the F1-score, area under the receiver operating characteristic curve (AUROC), and area under the precision-recall curve (AUPRC).

RESULTS: There were 571 case ECGs and 826 control ECGs in the NTUH data set and 378 case ECGs and 713 control ECGs in the FEMH data set. The novel EIANet model achieved an F1-score of 0.805, AUROC of 0.896, and AUPRC of 0.842 on the NTUH-ECG data set with a 40% positive sample ratio. It achieved an F1-score of 0.650, AUROC of 0.803, and AUPRC of 0.678 on the FEMH-ECG data set with a 34.6% positive sample ratio. The feature map showed that the region of interest in the ECG was the ST segment.

CONCLUSIONS: EIANet demonstrates promising potential for accurately predicting EDCA using triage ECG images, offering an effective solution for early detection of high-risk cases in emergency settings. This approach may enhance the ability of health care professionals to make timely decisions, with the potential to improve patient outcomes by enabling earlier interventions for EDCA.

PMID:40053733 | DOI:10.2196/67576

J Chem Inf Model. 2025 Mar 7. doi: 10.1021/acs.jcim.4c02216. Online ahead of print.

ABSTRACT

Proteins, as the fundamental macromolecules of life, play critical roles in various biological processes. Recent advancements in intelligent protein function prediction methods leverage sequences, structures, and biomedical literature data. Among them, function prediction methods for protein sequences remain an enduring and popular research direction. Existing studies have failed to effectively utilize the multilevel attribute features reflected in protein sequences. This limitation hinders the enrichment of protein descriptions needed for high-precision prediction of protein functions. To address this, we propose DeepMVD, a novel deep learning model that enhances prediction accuracy by dynamically fusing multiview features. DeepMVD employs specialized modules to extract unique features from each view and utilizes an adaptive fusion mechanism for optimal integration. Evaluation of the CAFA4 data set shows that DeepMVD significantly outperforms existing state-of-the-art models in terms of BP, MF, and CC terminology, all obtaining the highest Fmax (0.523, 0.712, 0.740). Ablation studies confirm the model's robustness. Source code and data sets are available at http://swanhub.co/scl/DeepMVD.

PMID:40053671 | DOI:10.1021/acs.jcim.4c02216

IEEE J Biomed Health Inform. 2025 Mar 7;PP. doi: 10.1109/JBHI.2025.3548940. Online ahead of print.

ABSTRACT

MiRNA subcellular localizations (MSLs) are essential for uncovering and understanding miRNA functions in various biological processes. Several computational methods have been proposed for measuring MSL. However, existing methods only rely on manually crafted features based on sequence without considering RNA 3D structure information, and most methods often rely on single-model approaches, which fail to capture the full complexity of biological systems, further hindering predictive accuracy and performance. In this study, we introduce a deep learning-based approach, MMFmiRLocEL, which integrates multi-model fusion and ensemble learning for MSL identification. To the best of our knowledge, MMFmiRLocEL is the first method to combine sequence, structure, and function three information for MSL prediction. Specifically, it employs RNA 3D structure generated by the predicted structural model to construct a structure-based approach for MSL prediction. It also develops a sequence-based prediction method using sequence features and convolutional neural networks, while constructing a function-based prediction method using miRNA-disease association networks and deep residual neural networks. Furthermore, a multi-model fusion approach, employing weighted ensemble strategies, integrates sequence, structure, and function models to enhance the robustness and accuracy of MSL identification. Experimental results demonstrate that MMFmiRLocEL outperforms existing state-of-the-art methods, and then ablation analysis confirmed the significant contribution of the multi-model fusion mechanism to improve the prediction performance.

PMID:40053625 | DOI:10.1109/JBHI.2025.3548940

PLoS One. 2025 Mar 7;20(3):e0317240. doi: 10.1371/journal.pone.0317240. eCollection 2025.

ABSTRACT

BACKGROUND: Breast cancer (BC) continues to pose a substantial global health concern, necessitating continuous advancements in therapeutic approaches. Neoadjuvant chemotherapy (NAC) has gained prominence as a key therapeutic strategy, and there is growing interest in the predictive utility of Background Parenchymal Enhancement (BPE) in evaluating the response of breast tumors to NAC. However, the analysis of BPE as a predictive biomarker, along with the techniques used to model BPE changes for accurate and timely predictions of treatment response presents several obstacles. This systematic review aims to thoroughly investigate recent advancements in the analytical methodologies for BPE analysis, and to evaluate their reliability and effectiveness in predicting breast tumor response to NAC, ultimately contributing to the development of personalized and effective therapeutic strategies.

METHODS: A comprehensive and structured literature search was conducted across key electronic databases, including Cochrane Database of Systematic Reviews, Google Scholar, PubMed, and IEEE Xplore covering articles published up to May 10, 2024. The inclusion criteria targeted studies focusing on breast cancer cohorts treated with NAC, involving both pre-treatment and at least one post-treatment breast dynamic contrast-enhanced Magnetic Resonance Imaging (DCE-MRI) scan, and analyzing BPE utility in predicting breast tumor response to NAC. Methodological quality assessment and data extraction were performed to synthesize findings and identify commonalities and differences among various BPE analytical approaches.

RESULTS: The search yielded a total of 882 records. After meticulous screening, 78 eligible records were identified, with 13 studies ultimately meeting the inclusion criteria for the systematic review. Analysis of the literature revealed a significant evolution in BPE analysis, from early studies focusing on single time-point BPE analysis to more recent studies adopting longitudinal BPE analysis. The review uncovered several gaps that compromise the accuracy and timeliness of existing longitudinal BPE analysis methods, such as missing data across multiple imaging time points, manual segmentation of the whole-breast region of interest, and over reliance on traditional statistical methods like logistic regression for modeling BPE and pathological complete response (pCR).

CONCLUSION: This review provides a thorough examination of current advancements in analytical approaches for BPE analysis in predicting breast tumor response to NAC. The shift towards longitudinal BPE analysis has highlighted significant gaps, suggesting the need for alternative analytical techniques, particularly in the realm of artificial intelligence (AI). Future longitudinal BPE research work should focus on standardization in longitudinal BPE measurement and analysis, through integration of deep learning-based approaches for automated tumor segmentation, and implementation of advanced AI technique that can better accommodate varied breast tumor responses, non-linear relationships and complex temporal dynamics in BPE datasets, while also handling missing data more effectively. Such integration could lead to more precise and timely predictions of breast tumor responses to NAC, thereby enhancing personalized and effective breast cancer treatment strategies.

PMID:40053513 | DOI:10.1371/journal.pone.0317240