Large Datasets Research Articles

Enhanced Early Detection of Thyroid Abnormalities using a Hybrid Deep Learning Model: A Sequential CNN and K-Means Clustering Approach

The thyroid gland, often referred to as the butterfly gland due to its shape, is located in the neck and plays a crucial role in regulating metabolism. It is susceptible to various health conditions, including hypothyroidism, hyperthyroidism, thyroid cancer, and thyroid nodules. Early detection of these conditions is essential for accurate diagnosis and effective treatment. Detecting thyroid nodules using machine learning and deep learning techniques presents a challenging yet promising research avenue. The choice of model depends on the characteristics of the patient's thyroid data, the dataset size, and the available computational resources. Hybrid models can be employed to handle complex data more effectively. In this study, a sequential Convolutional Neural Network (CNN) model was developed due to its capability to automate feature extraction and focus on Regions-of-Interest (ROIs) for detecting thyroid abnormalities. The proposed model achieved an accuracy of 81.5%, with a precision of 97.4% and a sensitivity of 83.1%, indicating its robustness in classifying images as benign or malignant. The confusion matrix provided further performance insights. Data segmentation was enhanced using K-means clustering for its scalability and efficiency in processing large medical image datasets. Compared to traditional models, the proposed hybrid approach demonstrated a significant improvement in diagnostic accuracy and precision, achieving performance gains of approximately 15-20% over baseline methods. These advancements underscore the potential of integrating machine learning and deep learning in medical diagnostics, paving the way for more reliable and efficient diagnostic tools for healthcare professionals.

SHINE: A Scalable Heterogeneous Inductive Graph Neural Network for Large Imbalanced Datasets

SHINE: A Scalable Heterogeneous Inductive Graph Neural Network for Large Imbalanced Datasets

Main challenges on the curation of large scale datasets for pancreas segmentation using deep learning in multi-phase CT scans: Focus on cardinality, manual refinement, and annotation quality

Accurate segmentation of the pancreas in computed tomography (CT) holds paramount importance in diagnostics, surgical planning, and interventions. Recent studies have proposed supervised deep-learning models for segmentation, but their efficacy relies on the quality and quantity of the training data. Most of such works employed small-scale public datasets, without proving the efficacy of generalization to external datasets. This study explored the optimization of pancreas segmentation accuracy by pinpointing the ideal dataset size, understanding resource implications, examining manual refinement impact, and assessing the influence of anatomical subregions. We present the AIMS-1300 dataset encompassing 1,300 CT scans. Its manual annotation by medical experts required 938 h. A 2.5D UNet was implemented to assess the impact of training sample size on segmentation accuracy by partitioning the original AIMS-1300 dataset into 11 smaller subsets of progressively increasing numerosity. The findings revealed that training sets exceeding 440 CTs did not lead to better segmentation performance. In contrast, nnU-Net and UNet with Attention Gate reached a plateau for 585 CTs. Tests on generalization on the publicly available AMOS-CT dataset confirmed this outcome. As the size of the partition of the AIMS-1300 training set increases, the number of error slices decreases, reaching a minimum with 730 and 440 CTs, for AIMS-1300 and AMOS-CT datasets, respectively. Segmentation metrics on the AIMS-1300 and AMOS-CT datasets improved more on the head than the body and tail of the pancreas as the dataset size increased. By carefully considering the task and the characteristics of the available data, researchers can develop deep learning models without sacrificing performance even with limited data. This could accelerate developing and deploying artificial intelligence tools for pancreas surgery and other surgical data science applications.

Modelling soil prokaryotic traits across environments with the trait sequence database ampliconTraits and the R package MicEnvMod

We present a comprehensive, customizable workflow for inferring prokaryotic phenotypic traits from marker gene sequences and modelling the relationships between these traits and environmental factors, thus overcoming the limited ecological interpretability of marker gene sequencing data. We created the trait sequence database ampliconTraits, constructed by cross-mapping species from a phenotypic trait database to the SILVA sequence database and formatted to enable seamless classification of environmental sequences using the SINAPS algorithm. The R package MicEnvMod enables modelling of trait – environment relationships, combining the strengths of different model types and integrating an approach to evaluate the models' predictive performance in a single framework. Traits could be accurately predicted even for sequences with low sequence identity (80 %) with the reference sequences, indicating that our approach is suitable to classify a wide range of environmental sequences. Validating our approach in a large trans-continental soil dataset, we showed that trait distributions were robust to classification settings such as the bootstrap cutoff for classification and the number of discrete intervals for continuous traits. Using functions from MicEnvMod, we revealed precipitation seasonality and land cover as the most important predictors of genome size. We found Pearson correlation coefficients between observed and predicted values up to 0.70 using repeated split sampling cross validation, corroborating the predictive ability of our models beyond the training data. Predicting genome size across the Iberian Peninsula, we found the largest genomes in the northern part. Potential limitations of our trait inference approach include dependence on the phylogenetic conservation of traits and limited database coverage of environmental prokaryotes. Overall, our approach enables robust inference of ecologically interpretable traits combined with environmental modelling allowing to harness traits as bioindicators of soil ecosystem functioning.

Automated detection of Bornean white-bearded gibbon (Hylobates albibarbis) vocalizations using an open-source framework for deep learning.

Passive acoustic monitoring is a promising tool for monitoring at-risk populations of vocal species, yet, extracting relevant information from large acoustic datasets can be time-consuming, creating a bottleneck at the point of analysis. To address this, an open-source framework for deep learning in bioacoustics to automatically detect Bornean white-bearded gibbon (Hylobates albibarbis) "great call" vocalizations in a long-term acoustic dataset from a rainforest location in Borneo is adapted. The steps involved in developing this solution are described, including collecting audio recordings, developing training and testing datasets, training neural network models, and evaluating model performance. The best model performed at a satisfactory level (F score = 0.87), identifying 98% of the highest-quality calls from 90 h of manually annotated audio recordings and greatly reduced analysis times when compared to a human observer. No significant difference was found in the temporal distribution of great call detections between the manual annotations and the model's output. Future work should seek to apply this model to long-term acoustic datasets to understand spatiotemporal variations in H. albibarbis' calling activity. Overall, a roadmap is presented for applying deep learning to identify the vocalizations of species of interest, which can be adapted for monitoring other endangered vocalizing species.

Investigating the shared genetic architecture between psychiatric disorders and executive function

BackgroundEvidence for widespread comorbidity of executive dysfunctions with psychiatric disorders suggests common mechanisms underlying their pathophysiology. However, the shared genetic architectures between psychiatric disorders and executive functions remain poorly understood. MethodsLeveraging large GWAS datasets of European ancestry on bipolar disorder (BIP, N=353,899), major depressive disorder (MDD, N=674,452), schizophrenia (SCZ, N=130,644) from the Psychiatric Genomics Consortium (PGC) and the Integrative Psychiatric Research (iPSYCH), and a common factor of executive functions (N=427,037) from UK Biobank, we systematically investigated the shared genomic architectures between psychiatric disorders and executive function with a set of statistical genetic, functional genomic, and gene-level analyses. ResultsOur study demonstrated substantial genetic overlaps and significant genetic correlations between psychiatric disorders and executive function. Executive function has an estimated 95.9%, 98.1%, and 99.2% of phenotype-influencing variants, as well as 50, 23, and 130 genomic loci shared with BIP, MDD, and SCZ, respectively. SNP heritability enrichment suggests that genetic architecture of psychiatric disorders and executive function implicate in brain frontal cortex and prefrontal glutamergic neurons 1 (ExPFC1) and ExPFC2. Functional genomic analysis of shared variants identified 12 functional regulatory variants that regulate gene expression by affecting the binding affinities of five transcription factors. Additionally, functional characterization analyses of shared genes revealed potential common biological mechanisms related to synaptic processes and fetal brain development. ConclusionsOur findings provide evidence for extensive shared genetic architectures between psychiatric disorders and executive function and have valuable implications for future mechanistic investigations and drug development efforts.

Improving inter-session performance via relevant session-transfer for multi-session motor imagery classification

Motor imagery (MI)-based brain-computer interfaces (BCIs) using electroencephalography (EEG) have found practical applications in external device control. However, the non-stationary nature of EEG signals remains to obstruct BCI performance across multiple sessions, even for the same user. In this study, we aim to address the impact of non-stationarity, also known as inter-session variability, on multi-session MI classification performance by introducing a novel approach, the relevant session-transfer (RST) method. Leveraging the cosine similarity as a benchmark, the RST method transfers relevant EEG data from the previous session to the current one. The effectiveness of the proposed RST method was investigated through performance comparisons with the self-calibrating method, which uses only the data from the current session, and the whole-session transfer method, which utilizes data from all prior sessions. We validated the effectiveness of these methods using two datasets: a large MI public dataset (Shu Dataset) and our own dataset of gait-related MI, which includes both healthy participants and individuals with spinal cord injuries. Our experimental results revealed that the proposed RST method leads to a 2.29 % improvement (p < 0.001) in the Shu Dataset and up to a 6.37 % improvement in our dataset when compared to the self-calibrating method. Moreover, our method surpassed the performance of the recent highest-performing method that utilized the Shu Dataset, providing further support for the efficacy of the RST method in improving multi-session MI classification performance. Consequently, our findings confirm that the proposed RST method can improve classification performance across multiple sessions in practical MI-BCIs.

Family firms and carbon emissions

This study examines the relationship between family firms and carbon emissions using a large cross-country dataset of 6600 non-financial firms over the period 2010–2019. We find that family firms emit less carbon than non-family firms, especially after the Paris Agreement. Several factors contribute to this outcome, including governance structure, the degree of family control, R&D spending, and the issuance of green patents. Our study also shows that despite lower carbon emissions, family firms have lower environmental scores, primarily due to their reduced public commitment to emission reduction. Both environmental scores and carbon emissions increase when non-family CEOs are appointed and when family ownership decreases, indicating that agency conflicts may influence these outcomes.

Application of transformers in stomatological imaging: A review

Stomatology extensively utilizes detailed imaging to assist clinicians, traditionally requiring labor-intensive manual analysis, which significantly adds to their workload. Transformers are revolutionary neural network in deep learning, gaining substantial momentum in computer vision tasks. Their introduction into medical imaging, particularly in processing large image datasets, has been marked by remarkable precision and efficiency, establishing them as a pivotal tool in emerging research. However, the application of Transformers in stomatological imaging is still in its infancy. Current studies primarily focus on segmenting specific anatomical features such as teeth and jawbones, with some clinical implementations. Yet, the comprehensive analytical potential of Transformers in this field remains largely untapped. This paper presents an introductory examination of Transformers, coupled with an initial synthesis and assessment of its dental applications across various areas. It will highlight the observed advantages and limitations in dental contexts and conclude with a discussion on future research directions. This serves as a foundational guide for in-depth future investigations in this area.

Dynamic prediction and optimization of tunneling parameters with high reliability based on a hybrid intelligent algorithm

In this paper, a hybrid intelligent framework comprising Bayesian optimization (BO), gradient boosting with categorical features (CatBoost) and the nondominated sorting genetic algorithm-III (NSGA-III) was proposed to support multiobjective optimization of shield construction parameters without large sample datasets, improve the shield performance, and ensure reliable and interpretable results. First, with the use of the specific tunneling energy, advancing speed and cutter wear as objective functions, a BO-CatBoost prediction model for shield construction parameters and various objectives was constructed, and the key influencing factors were identified via the SHapley Additive exPlanations (SHAP) method. Then, a BO-CatBoost-NSGA-III model was developed to obtain Pareto solutions under different scenarios involving the adjustment of the key influencing factors. Finally, adopting the Wuhan Metro as the background, the accuracy, stability, and generalizability of the constructed algorithm were verified. The results indicated that (1) the developed BO-CatBoost algorithm is superior to 9 other algorithms. The R2 values of the proposed approach were 0.976 and 0.901-0.976 on the test set. (2) The developed BO-CatBoost-NSGA-III algorithm could be used to obtain Pareto solutions under different scenarios via the adjustment of the key influencing factors with the SHAP method, and the optimal solutions could facilitate improvements in the advancing speed, specific tunneling energy and cutter wear of 3.45%, 6.09%, and 0.52%, respectively, with an overall average reliability of 90.5%. (3) By comparing various prediction algorithms, optimization schemes of different objectives and geological conditions, the accuracy, stability, and generalizability of the constructed algorithm were verified. The developed BO-CatBoost-NSGA-III framework could enable dynamic adjustment of shield construction parameters for decision-making purposes in the event of conflicting shield construction objectives and exhibits generality.

An Effective Res-Progressive Growing Generative Adversarial Network-Based Cross-Platform Super-Resolution Reconstruction Method for Drone and Satellite Images

In recent years, accurate field monitoring has been a research hotspot in the domains of aerial remote sensing and satellite remote sensing. In view of this, this study proposes an innovative cross-platform super-resolution reconstruction method for remote sensing images for the first time, aiming to make medium-resolution satellites capable of field-level detection through a super-resolution reconstruction technique. The progressive growing generative adversarial network (PGGAN) model, which has excellent high-resolution generation and style transfer capabilities, is combined with a deep residual network, forming the Res-PGGAN model for cross-platform super-resolution reconstruction. The Res-PGGAN architecture is similar to that of the PGGAN, but includes a deep residual module. The proposed Res-PGGAN model has two main benefits. First, the residual module facilitates the training of deep networks, as well as the extraction of deep features. Second, the PGGAN structure performs well in cross-platform sensor style transfer, allowing for cross-platform high-magnification super-resolution tasks to be performed well. A large pre-training dataset and real data are used to train the Res-PGGAN to improve the resolution of Sentinel-2’s 10 m resolution satellite images to 0.625 m. Three evaluation metrics, including the structural similarity index metric (SSIM), the peak signal-to-noise ratio (PSNR), and the universal quality index (UQI), are used to evaluate the high-magnification images obtained by the proposed method. The images generated by the proposed method are also compared with those obtained by the traditional bicubic method and two deep learning super-resolution reconstruction methods: the enhanced super-resolution generative adversarial network (ESRGAN) and the PGGAN. The results indicate that the proposed method outperforms all the comparison methods and demonstrates an acceptable performance regarding all three metrics (SSIM/PSNR/UQI: 0.9726/44.7971/0.0417), proving the feasibility of cross-platform super-resolution image recovery.

Mobility and persistence of pesticides and emerging contaminants in age-dated and redox-classified groundwater under a range of land use types

Understanding groundwater contamination patterns is hampered by the heterogeneous groundwater age and redox status over the depth range typically sampled for identifying pesticides and emerging contaminants threats. This study explores depth patterns of groundwater age and redox status across various land use types, unraveling spatial and temporal trends of pesticides and emerging contaminants using data from groundwater quality monitoring in the south of the Netherlands. The Netherlands is an ideal testing ground due to its high population density and widespread groundwater contamination from multiple sources. 146 multi-level observation wells were age-dated using 3H/3He, and contaminant concentrations were analyzed based on recharge year, land use type, and redox conditions, mitigating uncertainties from spatial and depth-dependent variations in both groundwater age and redox status. Redox-recharge year diagrams were developed to visually evaluate contaminant patterns in relation to these factors and to assess concentration patterns in relation to contamination history. Most detections of pesticides, metabolites, and emerging contaminants occurred in the youngest recharge periods (2000–2010 and 2010–2020) and in agricultural areas. However, certain contaminants, including BAM, desphenyl-chloridazon, short-chain PFCAs, PFOA, and EDTA, were consistently found in older water and Fe- or SO4-reduced conditions, indicating their mobility and persistence in the regional groundwater system. Comparing the presence of contaminants in specific redox classes and recharge periods with known application or leaching history provides insights into retardation (e.g., PFOS) and degradation (e.g., 2-hydroxy-atrazine, benzotriazole), explaining lower detection frequencies in earlier recharge periods. Identifying recharge years from age-dated groundwater helps relate contaminants to farmland application or river water recharge periods, revealing leaching history and contamination origins. The presented framework has the potential to enhance the interpretation of large groundwater datasets from dedicated, short-screened observation wells, such as those from the Danish GRUMO network, the Dutch monitoring networks, and parts of the US National Water Quality Program.

451. PREDICTIVE FACTORS FOR PNEUMONIA AND ANASTOMOTIC LEAK AFTER TRANSTHORACIC ESOPHAGECTOMY FOR CANCER AND FAILURE TO RESCUE ANALYSIS ON 1355 PATIENTS

Abstract Background Pneumonia (PN) and anastomotic leak (AL) represent the most feared complications after transthoracic esophagectomy for cancer, the former more commonly observed after open and the latter after minimally-invasive (MI) procedures. Studies comparing these two complications with focus on predictive risk factors and procedure-related failure-to-rescue (FTR) are sparse. Aim of the study was to explore such aspects with a large multicentric dataset. Methods All patients who underwent an Ivor Lewis transthoracic esophagectomy for cancer between 2008 and 2022 in four tertiary centers were queried. Predictive risk factors for pneumonia, anastomotic leak and in-hospital mortality were analyzed with univariable models. FTR was defined as in-hospital death following a major complication. Patients were then classified and compared according to the type of operation [open, hybrid-laparoscopic, hybrid-robotic, standard minimally-invasive (MIE) or robotic minimally-invasive (RAMIE)]. Pneumonia and anastomotic leak associated FTR was calculated and compared across the groups. Machine-learning models (neuronal nets, support vector machines and random forests classifier) were applied to predict complications and FTR. Results In total, 1,355 patients were included. Significant (p&amp;lt;0.05) predictors for PN (n=211, 15.6%) were ASA=4, preexisting cardiac and pulmonary conditions, conversion (thoracotomy), neoadjuvant therapy and age. Those for AL (n=160, 11.8%) were preexisting cardiac condition, conversion (thoracotomy) and operation time. Overall FTR-rate was 3.6%. Age&amp;gt;65y, preexisting cardiac and pulmonary condition, AL, PN, postoperative bleeding, sepsis, pulmonary embolism, myocardial infarction, arrhythmia and need for blood transfusion were the risk factors significantly associated with in-hospital mortality. PN- and AL- associated FTR rates did not significantly differ among the different operation types (Figure 1). Machine-learning models were unable to predict complications or FTR properly. Conclusions Our study showed that there is a greater number of factors associated with pneumonia rather than anastomotic leak, probably making the prevention of this complication more challenging to achieve. However, pneumonia- and leak-associated FTR-rates did not significantly differ among the groups, demonstrating that the implementation of minimally-invasive procedures can be safely carried out, if experience is provided.

The use of AI technologies in cancer care

AI is already an essential tool in the handling of large data sets in epidemiology and basic research.Significant contributions to radiological diagnosis are emerging alongside increasing use of digital pathology. The future lies in integrating this information together with clinical data relevant to each individual patient. Linkage with clinical protocols will enable personalised management options to be presented to the oncologist of the future.Radiotherapy has the distinction of being the first area to have a NICE approved AI based recommendation. There is the opportunity to revolutionise the workflow with many tasks currently undertaken by clinicians taken over by AI based systems for volume outlining, planning and quality assurance.Education and training will be essential to understand the AI processes and inputs. Clinicians will however have to feel confident interrogating the AI derived information and in communicating AI-derived treatment plans to patients.

Stock market index prediction based on market trend using LSTM

The stock market data analysis has received interest as a result of technological advancements and the investigation of new machine learning models, since these models provide a platform for traders and business people to choose gaining stocks. The business price prediction is a challenging and extremely complex process due to the impact of several factors on company prices. The numerous patterns that the stock market goes, they have been the focus of extensive research and analysis by numerous experts. There are several large data sets accessible, an artificial intelligence and machine learning techniques are developing quickly, and because of the machine’s improved computational power, complex stock price prediction algorithms can be developed. This paper presents stock market index prediction based on market trend using long short-term memory (LSTM). Using built-in application programmable interface (API), Yahoo Finance offers a simple method to programmatically retrieve any historical stock prices of an organization using the ticker name. The standard and poor’s 500 index (S&amp;amp;P 500 index) include the firms that have been taken into consideration here. Utilizing the selected input variable, single-layer and multi-layer LSTM models are implemented, and the measurement parameters of mean absolute error (MAE), root mean square error (RMSE), and correlation coefficient (R) are used to compare each performance. Nearly all of the real closing price’s curve and the prediction curve’s closing price for test data overlap. A potential stock investor may benefit significantly from such a prediction by using it to make well-informed choices that would increase his earnings.

Prophet modeling for oil production forecasting in an enhanced oil recovery field

Accurate daily oil production forecasting is essential for efficient reservoir management and investment strategy. Forecasting oil production in enhanced oil recovery (EOR) and conformance-dominated fields is a complex process due to the nonlinear, voluminous, and often uncertain nature of reservoir parameters and hidden factors. As a result, conventional tools such as decline curve analysis frequently fail to accurately predict daily oil production in conformance-controlled areas. In contrast, machine learning works efficiently for large datasets, even if the parameter values are unknown. The current study employs a Prophet time series forecasting method for five oil production wells in an EOR applied field, but it fails to achieve the desired sweep efficiency. This study compares the results of conventional decline curve analysis (DCA) and popular autoregressive integrated moving average time series forecasting methods with the Prophet model. This is the first attempt to use Prophet for oil well production forecasting, where polymer flooding is used. In all, 60% of the data are used for training, and the remaining 40% are used for testing. The Prophet shows the best performance for all the wells. This study is also the first to handle shut-in data using the Prophet model for oil production. Well-2 achieves the highest accuracy after incorporating shut-in results, with an R2 score of 92%. The result shows that though the DCA performs reasonably well with higher linearity and trend stationary data, Prophet modeling shows superior results than conventional DCA for all EOR applied producing wells.

SSKG: Subject stream knowledge graph, a new approach for event detection from text

Events play a crucial role in shaping societal meaning and discourse. They are recorded in text data streams from social networks or online sources and analyzing them is vital for predicting and managing future occurrences. Methods for event detection from text data streams have been developed to enhance accuracy and efficiency in analyzing large datasets. This study focuses on events published as text data streams in Telegram. Two main perspectives are considered for better results: firstly, major events can overshadow the detection of smaller ones with fewer narrative instances. Secondly, due to diverse narratives surrounding each event, establishing meaningful connections between narrative sets is essential. In this research, two new concepts, “subject stream” and “stream graph,” are introduced. The subject stream aims to process topics to mitigate the impact of bursty events on detecting others. The stream graph models data stream and identifies various narratives associated with each event for better identification and categorization. Combining these approaches accurately represents events and their characteristics in the real world. Implementing the system in three versions demonstrates the effectiveness of using the “stream graph” alongside the “subject stream,” resulting in improved execution speed and accuracy. Evaluation results show a 6% enhancement in topic recall.

A variable fidelity approach for predicting aerodynamic wall quantities of hypersonic vehicles using the ConvNeXt Encoder-Decoder framework

Computational fluid dynamics (CFD) simulations for obtaining three-dimensional hypersonic vehicle aerodynamic characteristics are resource-intensive. Deep learning offers a promising alternative for predicting aerodynamic wall quantities but typically requires a large dataset, which conflicts with the intention to reduce CFD costs. To address this, we propose a novel prediction model leveraging variable fidelity data to alleviate computational demands. The model utilizes an encoder-decoder architecture with ConvNeXt blocks as operators, processing variable fidelity data as inputs and outputs. We also developed a U-Net with residual blocks (ResUNet) for performance comparison. Transformation and topology patching techniques were applied to tackle the challenges posed by large grid volumes and complex topologies in three-dimensional vehicles. Results demonstrate that the ConvNeXt Encoder-Decoder predicts peak regions more accurately than ResUNet and aligns closely with CFD in low-value areas. The ConvNeXt Encoder-Decoder maintains maximum heat flux errors below 1% and viscous drag errors below 3.81% across varying angles of attack. It exhibits superior fitting performance with minimal deviation from CFD compared to ResUNet. Prediction accuracy decreases under multiple inflow changes compared to the angle of attack variations alone. Heat flux predictions show high consistency with CFD, with relative errors below 6.38%, whereas friction predictions exhibit higher errors with 10.28% maximum error and 0.28% minimum error. The model accurately predicts friction variation trends in peak regions at the blunt edge's central plane but performs poorly in low-value areas. In summary, the ConvNeXt Encoder-Decoder accurately predicts the wall quantities under varying multiple inflow conditions.

NecroGlobalGCN: Integrating Micronecrosis Information in HCC Prognosis Prediction via Graph Convolutional Neural Networks

Background and ObjectiveHepatocellular carcinoma (HCC) ranks fourth in cancer mortality, underscoring the importance of accurate prognostic predictions to improve postoperative survival rates in patients. Although micronecrosis has been shown to have high prognostic value in HCC, its application in clinical prognosis prediction requires specialized knowledge and complex calculations, which poses challenges for clinicians. It would be of interest to develop a model to help clinicians make full use of micronecrosis to assess patient survival. MethodsTo address these challenges, we propose an HCC prognosis prediction model that integrates pathological micronecrosis information through Graph Convolutional Neural Networks (GCN). This approach enables GCN to utilize micronecrosis, which has been shown to be highly correlated with prognosis, thereby significantly enhancing prognostic stratification quality. We developed our model using 3622 slides from 752 patients with primary HCC from the FAH-ZJUMS dataset and conducted internal and external validations on the FAH-ZJUMS and TCGA-LIHC datasets, respectively. ResultsOur method outperformed the baseline by 8.18% in internal validation and 9.02% in external validations. Overall, this paper presents a deep learning research paradigm that integrates HCC micronecrosis, enhancing both the accuracy and interpretability of prognostic predictions, with potential applicability to other pathological prognostic markers. ConclusionsThis study proposes a composite GCN prognostic model that integrates information on HCC micronecrosis, collecting large dataset of HCC histopathological images. This approach could assist clinicians in analyzing HCC patient survival and precisely locating and visualizing necrotic tissues that affect prognosis. Following the research paradigm outlined in this paper, other prognostic biomarker integration models with GCN could be developed, significantly enhancing the predictive performance and interpretability of prognostic model.

Transfer learning of species co-occurrence patterns between plant communities

AimThe use of neural networks (NNs) is spreading to all areas of life, and Ecology is no exception. However, the data-hungry nature of NNs can leave out many small, valuable datasets. Here we show how to apply transfer learning to rescue small datasets that can be invaluable in understanding patterns of species co-occurrence. LocationSemiarid plant communities in Spain and México. Time period2016–2022. Major taxa studiedAngiosperms. MethodsBased on a large sample of plant species co-occurrence in vegetation patches in a semi-arid area of eastern Spain, we fit a generative artificial intelligence (AI) model that correctly reproduces which species live with which in these patches. Subsequently, we train the same type of model on two communities for which we only have smaller datasets (another semi-arid community in eastern Spain, and a tropical community in Mexico). ResultsWhen we transfer the knowledge learnt from the large dataset directly to the other two, the predictions improve for the community more similar to our reference one. As for the more dissimilar community, improving the accuracy of the transfer requires a further tuning of the model to the local data. In particular, the knowledge transferred relates primarily to species frequency and, to a lesser extent, to their phylogenetic relationships, which are known to be determinants of species interaction patterns. Main conclusionsThis AI-based approach can be performed for communities similar or not so similar to the reference community, opening the door to systematic transfer learning for accurate predictions on small datasets. Interestingly, this transfer operates by matching unrelated species between the origin and target datasets, implying that arbitrary datasets can then be transferred to, or even combined in order to augment each other, irrespective of the species involved, potentially allowing such models to be applied to a wide range of plant communities in different climates.