Biomed Opt Express. 2024 Dec 11;16(1):126-141. doi: 10.1364/BOE.541570. eCollection 2025 Jan 1.

ABSTRACT

Gastric cancer is a leading cause of cancer-related deaths globally. As mortality rates continue to rise, predicting cancer survival using multimodal data-including histopathological images, genomic data, and clinical information-has become increasingly crucial. However, extracting effective predictive features from this complex data has posed challenges for survival analysis due to the high dimensionality and heterogeneity of histopathology images and genomic data. Furthermore, existing methods often lack sufficient interaction between intra- and inter-modal features, significantly impacting model performance. To address these challenges, we developed a deep learning-based multimodal feature fusion model, MultiDeepsurv, designed to predict the survival of gastric cancer patients by integrating histopathological images, clinical data, and gene expression data. Our approach includes a two-branch hybrid network, GLFUnet, which leverages the attention mechanism for enhanced pathology image representation learning. Additionally, we employ a graph convolutional neural network (GCN) to extract features from gene expression data and clinical information. To capture the correlations between different modalities, we utilize the SFusion fusion strategy that employs a self-attention mechanism to learn potential correlations across modalities. Finally, these deeply processed features are fed into Cox regression models for an end-to-end survival analysis. Comprehensive experiments and analyses conducted on a gastric cancer cohort from The Cancer Genome Atlas (TCGA) demonstrate that our proposed MultiDeepsurv model outperforms other methods in terms of prognostic accuracy, with a C-index of 0.806 and an AUC of 0.842.

PMID:39816158 | PMC:PMC11729289 | DOI:10.1364/BOE.541570

Biomed Opt Express. 2024 Dec 23;16(1):303-319. doi: 10.1364/BOE.547176. eCollection 2025 Jan 1.

ABSTRACT

Lung cancer with heterogeneity has a high mortality rate due to its late-stage detection and chemotherapy resistance. Liquid biopsy that discriminates tumor-related biomarkers in body fluids has emerged as an attractive technique for early-stage and accurate diagnosis. Exosomes, carrying membrane and cytosolic information from original tumor cells, impart themselves endogeneity and heterogeneity, which offer extensive and unique advantages in the field of liquid biopsy for cancer differential diagnosis. Herein, we demonstrate a Gramian angular summation field and MobileNet V2 (GASF-MobileNet)-assisted surface-enhanced Raman spectroscopy (SERS) technique for analyzing exosomes, aimed at precise diagnosis of lung cancer. Specifically, a composite substrate was synthesized for SERS detection of exosomes based on Ti3C2Tx Mxene and the array of gold-silver core-shell nanocubes (MGS), that combines sensitivity and signal stability. The employment of MXene facilitates the non-selective capture and enrichment of exosomes. To overcome the issue of potentially overlooking spatial features in spectral data analysis, 1-D spectra were first transformed into 2-D images through GASF. By using transformed images as the input data, a deep learning model based on the MobileNet V2 framework extracted spectral features from higher dimensions, which identified different non-small cell lung cancer (NSCLC) cell lines with an overall accuracy of 95.23%. Moreover, the area under the curve (AUC) for each category exceeded 0.95, demonstrating the great potential of integrating label-free SERS with deep learning for precise lung cancer differential diagnosis. This approach allows routine cancer management, and meanwhile, its non-specific analysis of SERS signatures is anticipated to be expanded to other cancers.

PMID:39816152 | PMC:PMC11729284 | DOI:10.1364/BOE.547176

Biomed Opt Express. 2024 Dec 3;16(1):28-41. doi: 10.1364/BOE.539721. eCollection 2025 Jan 1.

ABSTRACT

The study aimed to identify differences in the biochemical composition of corneal stroma lenses across varying degrees of myopia using Raman spectrum characteristics. Corneal stroma lens samples from 38 patients who underwent small incision lens extraction (SMILE) surgery, were categorized into low (n = 9, spherical power ≧ -3.00D), moderate (n = 23, spherical power < -3.00D and > -6.00D), and high myopia (n = 6, spherical power ≦-6.00D) groups. A custom-built microscopic confocal Raman system (MCRS) was used to collect Raman spectra, which were processed by smoothing, denoising, and baseline calibrating to refine raw data. Independent sample t-tests were used to analyze spectral feature peaks among sample types. Significant differences (P < 0.001) were found in multiple Raman spectral characteristic peaks (854 cm-1, 937 cm-1, 1002 cm-1, 1243 cm-1, 1448 cm-1, and 2940 cm-1) between low and high myopia samples, particularly at 2940 cm-1. Differences were also found between low and moderate, and moderate and high myopia samples, although fewer than between low and high myopia samples. The three-classification model, particularly with PLS-KNN training, exhibited superior discriminative performance with accuracy rates of 95%. Similarly, the two-classification model for low and high myopia achieved high accuracy with PLS-KNN (94.4%) compared to PCA-KNN (93.3%). PLS dimensionality reduction slightly outperformed PCA, enhancing classification accuracy. In addition, in both reduction methods, the KNN algorithm demonstrated the highest accuracy and performance. The optimal PLS-KNN classification model showed AUC values of 0.99, 0.98, and 1.00 for ROC curves corresponding to low, moderate, and high myopia, respectively. Classification accuracy rates were 89.7% and 96.9%, and 100% for low and high myopia, respectively. For the two-classification model, accuracy reached 94.4% with an AUC of 0.98, indicating strong performance in distinguishing between high and low myopic corneal stroma. We found significant biochemical differences such as collagen, lipids, and nucleic acids in corneal stroma lenses across varying degrees of myopia, suggesting that Raman spectroscopy holds substantial potential in elucidating the pathogenesis of myopia.

PMID:39816144 | PMC:PMC11729285 | DOI:10.1364/BOE.539721

Narra J. 2024 Dec;4(3):e1512. doi: 10.52225/narra.v4i3.1512. Epub 2024 Dec 19.

ABSTRACT

Psoriasis is a chronic skin condition with challenges in the accurate assessment of its severity due to subtle differences between severity levels. The aim of this study was to evaluate deep learning models for automated classification of psoriasis severity. A dataset containing 1,546 clinical images was subjected to pre-processing techniques, including cropping and applying noise reduction through median filtering. The dataset was categorized into four severity classes: none, mild, moderate, and severe, based on the Psoriasis Area and Severity Index (PASI). It was split into 1,082 images for training (70%) and 463 images for validation and testing (30%). Five modified deep convolutional neural networks (DCNN) were evaluated, including ResNet50, VGGNet19, MobileNetV3, MnasNet, and EfficientNetB0. The data were validated based on accuracy, precision, sensitivity, specificity, and F1-score, which were weighted to reflect class representation; Pairwise McNemar's test, Cochran's Q test, Cohen's Kappa, and Post-hoc test were performed on the model performance, where overall accuracy and balanced accuracy were determined. Findings revealed that among the five deep learning models, ResNet50 emerged as the optimum model with an accuracy of 92.50% (95%CI: 91.2-93.8%). The precision, sensitivity, specificity, and F1-score of this model were found to be 93.10%, 92.50%, 97.37%, and 92.68%, respectively. In conclusion, ResNet50 has the potential to provide consistent and objective assessments of psoriasis severity, which could aid dermatologists in timely diagnoses and treatment planning. Further clinical validation and model refinement remain required.

PMID:39816098 | PMC:PMC11731931 | DOI:10.52225/narra.v4i3.1512

BioData Min. 2025 Jan 16;18(1):4. doi: 10.1186/s13040-024-00419-4.

ABSTRACT

BACKGROUND: Understanding the molecular properties of chemical compounds is essential for identifying potential candidates or ensuring safety in drug discovery. However, exploring the vast chemical space is time-consuming and costly, necessitating the development of time-efficient and cost-effective computational methods. Recent advances in deep learning approaches have offered deeper insights into molecular structures. Leveraging this progress, we developed a novel multi-view learning model.

RESULTS: We introduce a graph-integrated model that captures both local and global structural features of chemical compounds. In our model, graph attention layers are employed to effectively capture essential local structures by jointly considering atom and bond features, while multi-head attention layers extract important global features. We evaluated our model on nine MoleculeNet datasets, encompassing both classification and regression tasks, and compared its performance with state-of-the-art methods. Our model achieved an average area under the receiver operating characteristic (AUROC) of 0.822 and a root mean squared error (RMSE) of 1.133, representing a 3% improvement in AUROC and a 7% improvement in RMSE over state-of-the-art models in extensive seed testing.

CONCLUSION: MultiChem highlights the importance of integrating both local and global structural information in predicting molecular properties, while also assessing the stability of the models across multiple datasets using various random seed values.

IMPLEMENTATION: The codes are available at https://github.com/DMnBI/MultiChem .

PMID:39815309 | DOI:10.1186/s13040-024-00419-4

Respir Res. 2025 Jan 15;26(1):17. doi: 10.1186/s12931-024-03093-6.

ABSTRACT

H3 lysine 4 trimethylation (H3K4me3) modification and related regulators extensively regulate various crucial transcriptional courses in health and disease. However, the regulatory relationship between H3K4me3 modification and anti-tumor immunity has not been fully elucidated. We identified 72 independent prognostic genes of lung adenocarcinoma (LUAD) whose transcriptional expression were closely correlated with known 27 H3K4me3 regulators. We constructed three H3K4me3 modification patterns utilizing the expression profiles of the 72 genes, and patients classified in each pattern exhibited unique tumor immune infiltration characteristics. Using the principal component analysis (PCA) of H3K4me3-related patterns, we constructed a H3K4me3 risk score (H3K4me3-RS) system. The deep learning analysis using 12,159 cancer samples from 26 cancer types and 725 cancer samples from 5 immunotherapy cohorts revealed that H3K4me3-RS was significantly correlated with cancer immune tolerance and sensitivity. Importantly, this risk-score system showed satisfactory predictive performance for the ICB therapy responses of patients suffering from several cancer types, and we identified that SLAMF9 was one of the immunosuppressive phenotype and immunotherapy resistance-determined genes of H3K4me3-RS. The mice melanoma model showed Slamf9 knockdown remarkably restrained cancer progression and enhanced the efficacy of anti-CTLA-4 and anti-PD-L1 therapies by elevating CD8 + T cell infiltration. This study provided a new H3K4me3-associated biomarker system to predict tumor immunotherapy response and suggested the preclinical rationale for investigating the roles of SLAMF9 in cancer immunity regulation and treatment.

PMID:39815269 | DOI:10.1186/s12931-024-03093-6

BMC Med Inform Decis Mak. 2025 Jan 15;25(1):26. doi: 10.1186/s12911-025-02859-2.

ABSTRACT

PURPOSE: Identifying patients who may benefit from multiple drilling are crucial. Hence, the purpose of the study is to utilize radiomics and deep learning for predicting no-collapse survival in patients with femoral head osteonecrosis.

METHODS: Patients who underwent multiple drilling were enrolled. Radiomics and deep learning features were extracted from pelvic radiographs and selected by LASSO-COX regression, radiomics and DL signature were then built. The clinical variables were selected through univariate and multivariate Cox regression analysis, and the clinical, radiomics, DL and DLRC model were constructed. Model performance was evaluated using the concordance index (C-index), area under the receiver operating characteristic curve (AUC), net reclassification index (NRI), integrated discrimination improvement (IDI), calibration curves, and Decision Curve Analysis (DCA).

RESULTS: A total of 144 patients (212 hips) were included in the study. ARCO classification, bone marrow edema, and combined necrotic angle were identified as independent risk factors for collapse. The DLRC model exhibited superior discrimination ability with higher C-index of 0.78 (95%CI: 0.73-0.84) and AUC values (0.83 and 0.87) than other models. The DLRC model demonstrated superior predictive performance with a higher C-index of 0.78 (95% CI: 0.73-0.84) and area under the curve (AUC) values of 0.83 for 3-year survival and 0.87 for 5-year survival, outperforming other models. The DLRC model also exhibited favorable calibration and clinical utility, with Kaplan-Meier survival curves revealing significant differences in survival rates between high-risk and low-risk cohorts.

CONCLUSION: This study introduces a novel approach that integrates radiomics and deep learning techniques and demonstrates superior predictive performance for no-collapse survival after multiple drilling. It offers enhanced discrimination ability, favorable calibration, and strong clinical utility, making it a valuable tool for stratifying patients into high-risk and low-risk groups. The model has the potential to provide personalized risk assessment, guiding treatment decisions and improving outcomes in patients with osteonecrosis of the femoral head.

PMID:39815247 | DOI:10.1186/s12911-025-02859-2

BMC Bioinformatics. 2025 Jan 16;26(1):17. doi: 10.1186/s12859-024-06020-0.

ABSTRACT

BACKGROUND: Pacific Biosciences (PacBio) circular consensus sequencing (CCS), also known as high fidelity (HiFi) technology, has revolutionized modern genomics by producing long (10 + kb) and highly accurate reads. This is achieved by sequencing circularized DNA molecules multiple times and combining them into a consensus sequence. Currently, the accuracy and quality value estimation provided by HiFi technology are more than sufficient for applications such as genome assembly and germline variant calling. However, there are limitations in the accuracy of the estimated quality scores when it comes to somatic variant calling on single reads.

RESULTS: To address the challenge of inaccurate quality scores for somatic variant calling, we introduce TopoQual, a novel tool designed to enhance the accuracy of base quality predictions. TopoQual leverages techniques including partial order alignments (POA), topologically parallel bases, and deep learning algorithms to polish consensus sequences. Our results demonstrate that TopoQual corrects approximately 31.9% of errors in PacBio consensus sequences. Additionally, it validates base qualities up to q59, which corresponds to one error in 0.9 million bases. These improvements will significantly enhance the reliability of somatic variant calling using HiFi data.

CONCLUSION: TopoQual represents a significant advancement in genomics by improving the accuracy of base quality predictions for PacBio HiFi sequencing data. By correcting a substantial proportion of errors and achieving high base quality validation, TopoQual enables confident and accurate somatic variant calling. This tool not only addresses a critical limitation of current HiFi technology but also opens new possibilities for precise genomic analysis in various research and clinical applications.

PMID:39815230 | DOI:10.1186/s12859-024-06020-0

BMC Med Res Methodol. 2025 Jan 15;25(1):10. doi: 10.1186/s12874-024-02430-z.

ABSTRACT

PURPOSE: The process of searching for and selecting clinical evidence for systematic reviews (SRs) or clinical guidelines is essential for researchers in Traditional Chinese medicine (TCM). However, this process is often time-consuming and resource-intensive. In this study, we introduce a novel precision-preferred comprehensive information extraction and selection procedure to enhance both the efficiency and accuracy of evidence selection for TCM practitioners.

METHODS: We integrated an established deep learning model (Evi-BERT combined rule-based method) with Boolean logic algorithms and an expanded retrieval strategy to automatically and accurately select potential evidence with minimal human intervention. The selection process is recorded in real-time, allowing researchers to backtrack and verify its accuracy. This innovative approach was tested on ten high-quality, randomly selected systematic reviews of TCM-related topics written in Chinese. To evaluate its effectiveness, we compared the screening time and accuracy of this approach with traditional evidence selection methods.

RESULTS: Our finding demonstrated that the new method accurately selected potential literature based on consistent criteria while significantly reducing the time required for the process. Additionally, in some cases, this approach identified a broader range of relevant evidence and enabled the tracking of selection progress for future reference. The study also revealed that traditional screening methods are often subjective and prone to errors, frequently resulting in the inclusion of literature that does not meet established standards. In contrast, our method offers a more accurate and efficient way to select clinical evidence for TCM practitioners, outperforming traditional manual approaches.

CONCLUSION: We proposed an innovative approach for selecting clinical evidence for TCM reviews and guidelines, aiming to reduce the workload for researchers. While this method showed promise in improving the efficiency and accuracy of evidence-based selection, its full potential required further validation. Additionally, it may serve as a useful tool for editors to assess manuscript quality in the future.

PMID:39815209 | DOI:10.1186/s12874-024-02430-z

Mol Imaging Biol. 2025 Jan 15. doi: 10.1007/s11307-025-01981-x. Online ahead of print.

ABSTRACT

PURPOSE: We aim to perform radiogenomic profiling of breast cancer tumors using dynamic contrast magnetic resonance imaging (MRI) for the estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) genes.

METHODS: The dataset used in the current study consists of imaging data of 922 biopsy-confirmed invasive breast cancer patients with ER, PR, and HER2 gene mutation status. Breast MR images, including a T1-weighted pre-contrast sequence and three post-contrast sequences, were enrolled for analysis. All images were corrected using N4 bias correction algorithms. Based on all images and tumor masks, a bounding box of 128 × 128 × 68 was chosen to include all tumor regions. All networks were implemented in 3D fashion with input sizes of 128 × 128 × 68, and four images were input to each network for multi-channel analysis. Data were randomly split into train/validation (80%) and test set (20%) with stratification in class (patient-wise), and all metrics were reported in 20% of the untouched test dataset.

RESULTS: For ER prediction, SEResNet50 achieved an AUC mean of 0.695 (CI95%: 0.610-0.775), a sensitivity of 0.564, and a specificity of 0.787. For PR prediction, ResNet34 achieved an AUC mean of 0.658 (95% CI: 0.573-0.741), a sensitivity of 0.593, and a specificity of 0.734. For HER2 prediction, SEResNext101 achieved an AUC mean of 0.698 (95% CI: 0.560-0.822), a sensitivity of 0.750, and a specificity of 0.625.

CONCLUSION: The current study demonstrated the feasibility of imaging gene-phenotype decoding in breast tumors using MR images and deep learning algorithms with moderate performance.

PMID:39815134 | DOI:10.1007/s11307-025-01981-x

J Gastroenterol. 2025 Jan 15. doi: 10.1007/s00535-024-02209-5. Online ahead of print.

ABSTRACT

BACKGROUND: The automated classification of Helicobacter pylori infection status is gaining attention, distinguishing among uninfected (no history of H. pylori infection), current infection, and post-eradication. However, this classification has relatively low performance, primarily due to the intricate nature of the task. This study aims to develop a new multistage deep learning method for automatically classifying H. pylori infection status.

METHODS: The proposed multistage deep learning method was developed using a training set of 538 subjects, then tested on a validation set of 146 subjects. The classification performance of this new method was compared with the findings of four physicians.

RESULTS: The accuracy of our method was 87.7%, 83.6%, and 95.9% for uninfected, post-eradication, and currently infected cases, respectively, whereas that of the physicians was 81.7%, 76.5%, and 90.3%, respectively. When including the patient's H. pylori eradication history information, the classification accuracy of the method was 92.5%, 91.1%, and 98.6% for uninfected, post-eradication, and currently infected cases, respectively, whereas that of the physicians was 85.6%, 85.1%, and 97.4%, respectively.

CONCLUSION: The new multistage deep learning method shows potential for an innovative approach to gastric cancer screening. It can evaluate individual subjects' cancer risk based on endoscopic images and reduce the burden of physicians.

PMID:39815116 | DOI:10.1007/s00535-024-02209-5

Eur J Hum Genet. 2025 Jan 15. doi: 10.1038/s41431-025-01787-z. Online ahead of print.

ABSTRACT

The facial gestalt (overall facial morphology) is a characteristic clinical feature in many genetic disorders that is often essential for suspecting and establishing a specific diagnosis. Therefore, publishing images of individuals affected by pathogenic variants in disease-associated genes has been an important part of scientific communication. Furthermore, medical imaging data is also crucial for teaching and training deep-learning models such as GestaltMatcher. However, medical data is often sparsely available, and sharing patient images involves risks related to privacy and re-identification. Therefore, we explored whether generative neural networks can be used to synthesize accurate portraits for rare disorders. We modified a StyleGAN architecture and trained it to produce artificial condition-specific portraits for multiple disorders. In addition, we present a technique that generates a sharp and detailed average patient portrait for a given disorder. We trained our GestaltGAN on the 20 most frequent disorders from the GestaltMatcher database. We used REAL-ESRGAN to increase the resolution of portraits from the training data with low-quality and colorized black-and-white images. To augment the model's understanding of human facial features, an unaffected class was introduced to the training data. We tested the validity of our generated portraits with 63 human experts. Our findings demonstrate the model's proficiency in generating photorealistic portraits that capture the characteristic features of a disorder while preserving patient privacy. Overall, the output from our approach holds promise for various applications, including visualizations for publications and educational materials and augmenting training data for deep learning.

PMID:39815041 | DOI:10.1038/s41431-025-01787-z

Geroscience. 2025 Jan 15. doi: 10.1007/s11357-024-01502-8. Online ahead of print.

ABSTRACT

Aging is influenced by a complex interplay of multifarious factors, including an individual's genetics, environment, and lifestyle. Notably, high altitude may impact aging and age-related diseases through exposures such as hypoxia and ultraviolet (UV) radiation. To investigate this, we mined risk exposure data (summary exposure value), disease burden data (disability-adjusted life years (DALYs)), and death rates and life expectancy from the Global Health Data Exchange (GHDx) and National Data Management Center for Health of Ethiopia for each subnational region of Ethiopia, a country with considerable differences in the living altitude. We conducted a cross-sectional clinical trial involving 227 highland and 202 lowland dwellers from the Tigray region in Northern Ethiopia to gain a general insight into the biological aging at high altitudes. Notably, we observed significantly lower risk exposure rates and a reduced disease burden as well as increased life expectancy by lower mortality rates in higher-altitude regions of Ethiopia. When assessing biological aging using facial photographs, we found a faster rate of aging with increasing elevation, likely due to greater UV exposure. Conversely, analysis of nuclear morphologies of peripheral blood mononuclear cells (PBMCs) in blood smears with five different senescence predictors revealed a significant decrease in DNA damage-induced senescence in both monocytes and lymphocytes with increasing elevation. Overall, our findings suggest that disease and DNA damage-induced senescence decreases with altitude in agreement with the idea that oxidative stress may drive aging.

PMID:39815037 | DOI:10.1007/s11357-024-01502-8

Nature. 2025 Jan 15. doi: 10.1038/s41586-024-08435-4. Online ahead of print.

ABSTRACT

Molecular recognition events between proteins drive biological processes in living systems1. However, higher levels of mechanistic regulation have emerged, in which protein-protein interactions are conditioned to small molecules2-5. Despite recent advances, computational tools for the design of new chemically induced protein interactions have remained a challenging task for the field6,7. Here we present a computational strategy for the design of proteins that target neosurfaces, that is, surfaces arising from protein-ligand complexes. To develop this strategy, we leveraged a geometric deep learning approach based on learned molecular surface representations8,9 and experimentally validated binders against three drug-bound protein complexes: Bcl2-venetoclax, DB3-progesterone and PDF1-actinonin. All binders demonstrated high affinities and accurate specificities, as assessed by mutational and structural characterization. Remarkably, surface fingerprints previously trained only on proteins could be applied to neosurfaces induced by interactions with small molecules, providing a powerful demonstration of generalizability that is uncommon in other deep learning approaches. We anticipate that such designed chemically induced protein interactions will have the potential to expand the sensing repertoire and the assembly of new synthetic pathways in engineered cells for innovative drug-controlled cell-based therapies10.

PMID:39814890 | DOI:10.1038/s41586-024-08435-4

PLoS One. 2025 Jan 15;20(1):e0316291. doi: 10.1371/journal.pone.0316291. eCollection 2025.

ABSTRACT

Cognitive Radio (CR) technology enables wireless devices to learn about their surrounding spectrum environment through sensing capabilities, thereby facilitating efficient spectrum utilization without interfering with the normal operation of licensed users. This study aims to enhance spectrum sensing in multi-user cooperative cognitive radio systems by leveraging a hybrid model that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. A novel multi-user cooperative spectrum sensing model is developed, utilizing CNN's local feature extraction capability and LSTM's advantage in handling sequential data to optimize sensing accuracy and efficiency. Furthermore, a multi-head self-attention mechanism is incorporated to improve information flow, enhancing the model's adaptability and robustness in dynamic and complex environments. Simulation experiments were conducted to quantitatively evaluate the performance of the proposed model. The results demonstrate that the CNN-LSTM model achieves low sensing error rates across various numbers of secondary users (16, 24, 32, 40, 48), with a particularly low sensing error of 9.9658% under the 32-user configuration. Additionally, when comparing the sensing errors of different deep learning models, the proposed model consistently outperformed others, showing a 12% lower sensing error under low-power conditions (100 mW). This study successfully develops a CNN-LSTM-based cooperative spectrum sensing model for multi-user cognitive radio systems, significantly improving sensing accuracy and efficiency. By integrating CNN and LSTM technologies, the model not only enhances sensing performance but also improves the handling of long-term dependencies in time-series data, offering a novel technical approach and theoretical support for cognitive radio research. Moreover, the introduction of the multi-head self-attention mechanism further optimizes the model's adaptability to complex environments, demonstrating significant potential for practical applications.

PMID:39813223 | DOI:10.1371/journal.pone.0316291

PLoS One. 2025 Jan 15;20(1):e0310465. doi: 10.1371/journal.pone.0310465. eCollection 2025.

ABSTRACT

Accurate energy demand forecasting is critical for efficient energy management and planning. Recent advancements in computing power and the availability of large datasets have fueled the development of machine learning models. However, selecting the most appropriate features to enhance prediction accuracy and robustness remains a key challenge. This study proposes an ensemble approach that integrates a genetic algorithm with multiple forecasting models to optimize feature selection. The genetic algorithm identifies the optimal subset of features from a dataset that includes historical energy consumption, weather variables, and temporal characteristics. These selected features are then used to train three base learners: Long Short-Term Memory (LSTM), Bi-directional Long Short-Term Memory (BiLSTM), and Gated Recurrent Unit (GRU). The predictions from these models are combined using a stacking ensemble technique to generate the final forecast. To enhance model evaluation, we divided the dataset into weekday and weekend subsets, allowing for a more detailed analysis of energy consumption patterns. To ensure the reliability of our findings, we conducted ten simulations and applied the Wilcoxon Signed Rank Test to the results. The proposed model demonstrated exceptional precision, achieving a Root Mean Square Error (RMSE) of 130.6, a Mean Absolute Percentage Error (MAPE) of 0.38%, and a Mean Absolute Error (MAE) of 99.41 for weekday data. The model also maintained high accuracy for weekend predictions, with an RMSE of 137.41, a MAPE of 0.42%, and an MAE of 105.67. This research provides valuable insights for energy analysts and contributes to developing more sophisticated demand forecasting methods.

PMID:39813218 | DOI:10.1371/journal.pone.0310465

PLoS One. 2025 Jan 15;20(1):e0316258. doi: 10.1371/journal.pone.0316258. eCollection 2025.

ABSTRACT

The Covid-19 pandemic has sparked renewed attention to the risks of online misinformation, emphasizing its impact on individuals' quality of life through the spread of health-related myths and misconceptions. In this study, we analyze 6 years (2016-2021) of Italian vaccine debate across diverse social media platforms (Facebook, Instagram, Twitter, YouTube), encompassing all major news sources-both questionable and reliable. We first use the symbolic transfer entropy analysis of news production time-series to dynamically determine which category of sources, questionable or reliable, causally drives the agenda on vaccines. Then, leveraging deep learning models capable to accurately classify vaccine-related content based on the conveyed stance and discussed topic, respectively, we evaluate the focus on various topics by news sources promoting opposing views and compare the resulting user engagement. Our study uncovers misinformation not as a parasite of the news ecosystem that merely opposes the perspectives offered by mainstream media, but as an autonomous force capable of even overwhelming the production of vaccine-related content from the latter. While the pervasiveness of misinformation is evident in the significantly higher engagement of questionable sources compared to reliable ones (up to 11 times higher in median value), our findings underscore the need for consistent and thorough pro-vax coverage to counter this imbalance. This is especially important for sensitive topics, where the risk of misinformation spreading and potentially exacerbating negative attitudes toward vaccines is higher. While reliable sources have successfully promoted vaccine efficacy, reducing anti-vax impact, gaps in pro-vax coverage on vaccine safety led to the highest engagement with anti-vax content.

PMID:39813203 | DOI:10.1371/journal.pone.0316258

PLoS One. 2025 Jan 15;20(1):e0316832. doi: 10.1371/journal.pone.0316832. eCollection 2025.

ABSTRACT

There is growing interest in using deep learning models to automate wildlife detection in aerial imaging surveys to increase efficiency, but human-generated annotations remain necessary for model training. However, even skilled observers may diverge in interpreting aerial imagery of complex environments, which may result in downstream instability of models. In this study, we present a framework for assessing annotation reliability by calculating agreement metrics for individual observers against an aggregated set of annotations generated by clustering multiple observers' observations and selecting the mode classification. We also examined how image attributes like spatial resolution and texture influence observer agreement. To demonstrate the framework, we analyzed expert and volunteer annotations of twelve drone images of migratory waterfowl in New Mexico. Neither group reliably identified duck species: experts showed low agreement (43-56%) for several common species, and volunteers opted out of the task. When simplified into broad morphological categories, there was high agreement for cranes (99% among experts, 95% among volunteers) and ducks (93% among experts, 92% among volunteers), though agreement among volunteers was notably lower for classifying geese (75%) than among experts (94%). The aggregated annotation sets from the two groups were similar: the volunteer count of birds across all images was 91% of the expert count, with no statistically significant difference per image (t = 1.27, df = 338, p = 0.20). Bird locations matched 81% between groups and classifications matched 99.4%. Tiling images to reduce search area and maintaining a constant scale to keep size differences between classes consistent may increase observer agreement. Although our sample was limited, these findings indicate potential taxonomic limitations to aerial wildlife surveys and show that, in aggregate, volunteers can produce data comparable to experts'. This framework may assist other wildlife practitioners in evaluating the reliability of their input data for deep learning models.

PMID:39813190 | DOI:10.1371/journal.pone.0316832

Network. 2025 Jan 15:1-40. doi: 10.1080/0954898X.2024.2449173. Online ahead of print.

ABSTRACT

Skin cancer is one of the most prevalent and harmful forms of cancer, with early detection being crucial for successful treatment outcomes. However, current skin cancer detection methods often suffer from limitations such as reliance on manual inspection by clinicians, inconsistency in diagnostic accuracy, and a lack of personalized recommendations based on patient-specific data. In our work, we presented a Personalized Recommendation System to handle Skin Cancer at an early stage based on Hybrid Model (PRSSCHM). Preprocessing, improved deep joint segmentation, feature extraction, and classification are the major steps to identify the stages of skin cancer. The input image is first preprocessed using the Gaussian filtering method. Improved deep joint segmentation is employed to segment the preprocessed image. A set of features including Median Binary Pattern (MBP), Gray Level Co-occurrence Matrix (GLCM), and Improved Local Direction Texture Pattern (ILDTP) are extracted in the next step. Finally, the hybrid classification includes Improved Bi-directional Long Short-Term Memory (Bi-LSTM) and Deep Belief Network (DBN) used for the classification process, where the training will be carried out by the Integrated Bald Eagle and Average and Subtraction Optimizer (IBEASO) algorithm via optimizing the weights of the models.

PMID:39813094 | DOI:10.1080/0954898X.2024.2449173

JAMA Cardiol. 2025 Jan 15. doi: 10.1001/jamacardio.2024.4991. Online ahead of print.

ABSTRACT

IMPORTANCE: Lung ultrasound (LUS) aids in the diagnosis of patients with dyspnea, including those with cardiogenic pulmonary edema, but requires technical proficiency for image acquisition. Previous research has demonstrated the effectiveness of artificial intelligence (AI) in guiding novice users to acquire high-quality cardiac ultrasound images, suggesting its potential for broader use in LUS.

OBJECTIVE: To evaluate the ability of AI to guide acquisition of diagnostic-quality LUS images by trained health care professionals (THCPs).

DESIGN, SETTING, AND PARTICIPANTS: In this multicenter diagnostic validation study conducted between July 2023 and December 2023, participants aged 21 years or older with shortness of breath recruited from 4 clinical sites underwent 2 ultrasound examinations: 1 examination by a THCP operator using Lung Guidance AI and the other by a trained LUS expert without AI. The THCPs (including medical assistants, respiratory therapists, and nurses) underwent standardized AI training for LUS acquisition before participation.

INTERVENTIONS: Lung Guidance AI software uses deep learning algorithms guiding LUS image acquisition and B-line annotation. Using an 8-zone LUS protocol, the AI software automatically captures images of diagnostic quality.

MAIN OUTCOMES AND MEASURES: The primary end point was the proportion of THCP-acquired examinations of diagnostic quality according to a panel of 5 masked expert LUS readers, who provided remote review and ground truth validation.

RESULTS: The intention-to-treat analysis included 176 participants (81 female participants [46.0%]; mean [SD] age, 63 [14] years; mean [SD] body mass index, 31 [8]). Overall, 98.3% (95% CI, 95.1%-99.4%) of THCP-acquired studies were of diagnostic quality, with no statistically significant difference in quality compared to LUS expert-acquired studies (difference, 1.7%; 95% CI, -1.6% to 5.0%).

CONCLUSIONS AND RELEVANCE: In this multicenter validation study, THCPs with AI assistance achieved LUS images meeting diagnostic standards compared with LUS experts without AI. This technology could extend access to LUS to underserved areas lacking expert personnel.

TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT05992324.

PMID:39813064 | DOI:10.1001/jamacardio.2024.4991