Search

Background & Aims: Malnutrition is persistent in 50%-75% of children with congenital heart disease (CHD) after surgery, and early prediction is crucial for nutritional intervention. The aim of this study was to develop and validate machine learning (ML) models to predict the malnutrition status of children with CHD. We used explainable ML methods to provide insight into the model's predictions and outcomes., Methods: This prospective cohort study included consecutive children with CHD admitted to the hospital from December 2017 to May 2020. The cohort data were divided into the training and test data sets based on the follow-up time. The outcome of the study was CHD child malnutrition 1 year after surgery, the primary outcome was an underweight status, and the secondary outcomes were stunted and wasting status. We used five ML algorithms with multiple features to construct prediction models, and the performance of these ML models was measured by an area under the receiver operating characteristic curve (AUC) analysis. We also used the permutation importance and SHapley Additive exPlanations (SHAP) to determine the importance of the selected features and interpret the ML models., Results: We enrolled 536 children with CHD who underwent complete repair. The proportions of children with an underweight, stunted, or wasting status 1 year after surgery were 18.1% (97/536), 12.1% (65/536), and 17.5% (94/536), respectively. All patients contributed to the generation of 115 useable features, which allowed us to build models to predict malnutrition. Five prediction algorithms were used, and the XGBoost model achieved the greatest AUC in all outcomes. The results obtained from the permutation importance and SHAP analyses showed that the 1-month postoperative WAZ-score, discharge WAZ score and preoperative WAZ score were the top 3 important features in predicting an underweight status in the XGBoost algorithm. Regarding the stunted status, the top 3 important features were the 1-month postoperative HAZ score, discharge HAZ score, and aortic clamping time. Regarding the wasting status, the top 3 important features were the hospital length of stay, formula intake, and discharge WHZ-score. We also used a narrative case report as an example to describe the clinical manifestations and predicted the primary outcomes of two children., Conclusions: We developed an ML model (XGBoost) that provides accurate early predictions of malnutrition 1-year postoperatively in children with CHD. Because the ML model is explainable, it may better enable clinicians to better understand the reasoning underlying the outcome. Our study could aid in determining individual treatment and nutritional follow-up strategies for children with CHD., Competing Interests: Conflict of interest The authors read and understood the journal's policy regarding the declaration of interests and declare that there are no conflicts of interest., (Copyright © 2021 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.)

Abstract It is accepted that cancer stem cells (CSCs) are key to the occurrence, progression, drug resistance, and recurrence of bladder cancer (BLCA). Here, we aimed to characterize the landscapes of CSCs and investigate the biological and clinical signatures based on a prognostic model constructed by genes associated with CSCs. The malignant epithelial cells were discovered and sorted into six clusters through single cell analysis. C2 was identified as the CSCs. The signaling involved in the interactions between C2, cancer-associated fibroblasts (CAFs), and immune cells mainly consisted of MK, THBS, ANGPTL, VISFATIN, JAM, and ncWNT pathways. The CSC-like prognostic index (CSCLPI) constructed by the random survival forest was a reliable risk factor for BLCA and had a stable and powerful effect on predicting the overall survival of patients with BLCA. The level of CAFs was higher among patients with higher CSCLPI scores, suggesting that CAFs play a significant role in regulating biological characteristics. The CSCLPI-developed survival prediction nomogram has the potential to be applied clinically to predict the 1-, 2-, 3-, and 5-year overall survival of patients with BLCA. The CSCLPI can be used for prognostic prediction and drug treatment evaluation in the clinic.

Demodulation and decoding are pivotal for the eLoran system's timing and information transmission capabilities. This paper proposes a novel demodulation algorithm leveraging a multiclass support vector machine (MSVM) for pulse position modulation (PPM) of eLoran signals. Firstly, the existing demodulation method based on envelope phase detection (EPD) technology is reviewed, highlighting its limitations. Secondly, a detailed exposition of the MSVM algorithm is presented, demonstrating its theoretical foundations and comparative advantages over the traditional method and several other methods proposed in this study. Subsequently, through comprehensive experiments, the algorithm parameters are optimized, and the parallel comparison of different demodulation methods is carried out in various complex environments. The test results show that the MSVM algorithm is significantly superior to traditional methods and other kinds of machine learning algorithms in demodulation accuracy and stability, particularly in high-noise and -interference scenarios. This innovative algorithm not only broadens the design approach for eLoran receivers but also fully meets the high-precision timing service requirements of the eLoran system. [ABSTRACT FROM AUTHOR]

As an efficacious solution to remedying the imbalance of medical resources, the online medical platform has burgeoned expeditiously. Apt allotment of medical resources on the medical platform can facilitate patients in reasonably selecting physicians and time slots, coordinating doctors' clinical arrangements, and generating precise projections of medical platform service volume to enhance patient satisfaction and alleviate physicians' workload. To this end, grounded in the data‐driven method, this paper assembles an exhaustive feature set encompassing hospital features, physician features, and patient features. Through feature selection, appropriate features are screened, and machine learning algorithms are leveraged to accurately forecast doctors' online consultation volume. Subsequently, to glean the influence relationship between online medical services and offline medical services, this paper introduces features of offline medical services such as hospital registration volume and regional gross domestic product (GDP) to solve the prediction of offline medical service volume using online medical information. The findings signify that online data feature prediction can pinpoint superior machine learning models for online medical platform service volume (with the optimal accuracy up to 96.89%). Online features exert a positive effect on predicting offline medical service volume, but the accuracy declines to some degree (the optimal accuracy is 73%). Physicians with favorable reputations on the online platform are more susceptible to attain higher offline appointment volumes when online consultation volume is a vital feature impacting offline appointment volume. [ABSTRACT FROM AUTHOR]

In order to quickly and non-destructively detect imperfect maize kernels, and to effectively enhance the efficiency of maize quality detection during collection and storage, identification models for imperfect maize kernels were constructed by using hyperspectral imaging (HSI) combined with machine learning techniques. Hyperspectral images of maize kernels in the range of 380–1000 nm were collected, and 10 spectral characteristic wavelengths (variables) were selected by using variable combination cluster analysis (VCPA). For grayscale images corresponding to these characteristic wavelengths, 3 texture features were extracted by using Tamura algorithm. Additionally, 3 color features and 4 morphological features were extracted through color moment analysis and regional geometry calculation, respectively. Based on the data of spectral features, image features and fusion features (spectral and image features), partial least squares regression (PLSR) and extreme learning machine (ELM) algorithms were respectively used to establish identification models for imperfect maize kernels. The results demonstrated that the overall average recognition accuracy of the ELM models was 91.60%, slightly surpassing the 91.29% achieved by the PLSR models. Notably, the ELM model based on the fusion features exhibited the highest recognition accuracy for heat-damaged kernels, achieving an accuracy rate of 97.22% in the test set. Therefore, the classification models established in this study proved to be feasible for the rapid and accurate identification of imperfect maize kernels. This can provide valuable technical support for the research and the development of non-destructive, rapid inspection equipment for imperfect kernels, as well as online batch detection. [ABSTRACT FROM AUTHOR]

Rapid and non-destructive identification of the suitable storage status of paddy during storage is crucial for controlling the quality of stored grains, which can provide high-quality raw grains for rice processing. Near-infrared (NIR) spectroscopy combined with neighborhood rough set (NRS) and multiple classification methods were used to identify the different storage statuses of paddy. The NIR data were collected in the range of 1000–1800 nm, and three storage statuses from suitable storage to severely unsuitable storage were divided using the measured fatty acid value of paddy. The spectral features were selected using NRS, successive projection algorithm and variable combination population analysis methods. Random forest (RF), extreme learning machine, and soft independent modeling of class analogy classifiers coupled with spectral features were used to establish classification models to distinguish the different storage statuses of paddy. The comparison results indicated that the optimal wavelengths selected by NRS combined with the RF classifier to construct the NRS-RF series models led to satisfactory identification results, with high correct classification rates of 96.31% and 93.68% in the calibration and test sets, respectively; the indicators of sensitivity and specificity ranged from 0.93 to 0.99. Therefore, the combination of NIR technology with NRS and RF algorithms for identifying the storage status of paddy was feasible, as this would be more helpful for rapidly evaluating the changes of stored paddy quality. The proposed method from this study is expected to provide support for the development of non-destructive equipment for the accurate detection of the quality of stored paddy. [ABSTRACT FROM AUTHOR]

Developing reliable biomarkers is important for screening Alzheimer's disease (AD) and monitoring its progression. Although EEG is non-invasive direct measurement of brain neural activity and has potentials for various neurologic disorders, vulnerability to noise, difficulty in clinical interpretation and quantification of signal information have limited its clinical application. There have been many research about machine learning (ML) adoption with EEG, but the accuracy of detecting AD is not so high or not validated with Aβ PET scan. We developed EEG-ML algorithm to detect brain Aβ pathology among subjective cognitive decline (SCD) or mild cognitive impairment (MCI) population, and validated it with Aβ PET. 19-channel resting-state EEG and Aβ PET were collected from 311 subjects: 196 SCD(36 Aβ +, 160 Aβ −), 115 MCI(54 Aβ +, 61Aβ −). 235 EEG data were used for training ML, and 76 for validation. EEG features were standardized for age and sex. Multiple important features sets were selected by 6 statistics analysis. Then, we trained 8 multiple machine learning for each important features set. Meanwhile, we conducted paired t-test to find statistically different features between amyloid positive and negative group. The best model showed 90.9% sensitivity, 76.7% specificity and 82.9% accuracy in MCI + SCD (33 Aβ +, 43 Aβ −). Limited to SCD, 92.3% sensitivity, 75.0% specificity, 81.1% accuracy (13 Aβ +, 24 Aβ −). 90% sensitivity, 78.9% specificity and 84.6% accuracy for MCI (20 Aβ +, 19 Aβ −). Similar trends of EEG power have been observed from the group comparison between Aβ + and Aβ −, and between MCI and SCD: enhancement of frontal/ frontotemporal theta; attenuation of mid-beta in centroparietal areas. The present findings suggest that accurate classification for beta-amyloid accumulation in the brain based on QEEG alone could be possible, which implies that QEEG is a promising biomarker for beta-amyloid. Since QEEG is more accessible, cost-effective, and safer than amyloid PET, QEEG-based biomarkers may play an important role in the diagnosis and treatment of AD. We expect specific patterns in QEEG could play an important role to predict future progression of cognitive impairment in the preclinical stage of AD. Further feature engineering and validation with larger dataset is recommended. [ABSTRACT FROM AUTHOR]

To reduce CO2 emissions from shipping activities to, from, and within the European Union (EU) area, a system of monitoring, reporting, and verification (MRV) of CO2 emissions from ships are implemented in 2015 by the EU. Although the MRV records in 2018 and 2019 have been published, there are scarce studies on the MRV system especially from a quantitative perspective, which restrains the potential of the MRV. To bridge this gap, this paper first analyzes and compares MRV records in 2018 and 2019, and then develops machine learning models for annual average fuel consumption prediction for each ship type combining ship features from an external database. The performance of the prediction models is accurate, with the mean absolute percentage error (MAPE) on the test set no more than 12% and the average R-squared of all the models at 0.78. Based on the analysis and prediction results, model meanings, implications, and extensions are thoroughly discussed. This study is a pioneer to analyze the emission reports in the MRV system from a quantitative perspective. It also develops the first fuel consumption prediction models from a macro perspective using the MRV data. It can contribute to the promotion of green shipping strategies. [ABSTRACT FROM AUTHOR]

The gasoline compression ignition mode is an advanced combustion mode that achieves high efficiency with low emissions. This study constructs a global sensitivity analysis (GSA) by coupling machine learning (ML) algorithms and conducts a comprehensive investigation into the statistical correlations between research octane number (RON), fuel sensitivity (S), three initial ambient parameters, and the spray combustion characteristics of gasoline-like fuels. Eighty sets of CFD numerical simulation data were used for training the machine learning model. Results show liquid length (LL) is most sensitive to changes in ambient pressure while ambient temperature significantly influences vapor penetration length (VPL). Regarding combustion behaviors, ignition delay (ID) and flame lift-off length (LOL) show the highest sensitivity to ambient temperature. The occurrence of distinct ignition points in the domain also advances from 1.75 ms to 0.42 ms as the ambient temperature rises from 1000K to 1200K. Additionally, the ambient temperature has higher-order interaction effects with ambient pressure on LOL. The production of NOx in the field is also more closely correlated with ambient temperature, with higher temperatures promoting its generation. Furthermore, ambient temperature exhibits a positive interaction effect with both ambient pressure and oxygen concentration on NOx emissions. • Global sensitivity analysis method combined with machine learning was used. • The machine learning algorithms for predicting different targets were evaluated. • Initial thermodynamic environments had more influence on combustion and emission than fuel properties. • NOx was primarily generated in the high-temperature region at the spray's periphery. [ABSTRACT FROM AUTHOR]

Trajectory-based fishing vessel behavior recognition is currently a vibrant and dynamic area of investigation within transportation and maritime research. It can provide crucial technical support for combating illegal, unreported, and unregulated (IUU) fishing. Nevertheless, existing studies typically rely on extensive datasets of pre-labeled data to train machine learning models. The acquisition of these labeled datasets is frequently a time-consuming and complicated endeavor, which presents challenges in the practical application. Hence, this paper develops a semi-supervised approach for fishing vessel behavior identification using a small amount of manually labeled trajectory samples. Besides, given that the common vessel behavioral feature parameters provide limited insight into fishing activities, this paper innovatively proposes a set of descriptive trajectory feature parameters. Based on the morphological characteristics of fishing activities and domain knowledge, the proposed trajectory features excel in discerning the intricate pattern of fishing vessel trajectories. Combined with a set of Long Short-Term Memory (LSTM)-based encoding trajectory features, a composed feature engineering scheme that leverages both human intelligence and machine intelligence is proposed. The clustering experiment result confirms that our trajectory feature scheme can successfully recognize different fishing vessel behaviors while the conventional trajectory features can't. In the consequently classification tests, the 90% accuracy achieved by the composed feature scheme exhibits a 10% increase compared to solely using deep learning-based encoding features. This novel trajectory feature engineering method together with the semi-supervised machine learning structure, not reliant on extensive labeled datasets, can be tailored and applied to mobility pattern identification across various domains. • Develop a semi-supervised approach for fishing vessel behavior identification. • Propose a set of descriptive trajectory feature parameters based on domain knowledge. • Combine both human and machine intelligence to compose the feature engineering. • The developed approach identifies fishing vessel behavior with over 90% accuracy. [ABSTRACT FROM AUTHOR]

Efficient tunnel ventilation is essential for ensuring construction safety and protecting personnel health during tunnel construction. This study proposes a demand-controlled ventilation (DCV) method on the basis of deep learning algorithm to both improve pollutant removal efficiency and reduce energy consumption. The DCV method utilizes a two-layer bidirectional long short-term memory algorithm (BiLSTM) to predict pollutant concentrations. The air volume is dynamically adjusted based on the gaseous pollutant removal requirements. The coefficient of ventilation performance (COVP) is proposed to evaluate the performance of two ventilation methods (DCV and constant air-volume ventilation (CAV)) through computational fluid dynamics (CFD) simulations. The results show that the DCV results in a lower maximum average CO concentration and higher removal efficiency in the heading area (372.3 mg/m3) than the CAV does (404.1 mg/m3). The fan's energy consumption of DCV is 64.6% lower than that of CAV during a 1000 s ventilation period. The COVPs for both methods exhibit temporal variation and achieves their maximums (2.25 for DCV and 0.741 for CAV) after reaching the constraint conditions (air volume threshold). The DCV method expedites pollutant elimination, reduces construction waiting period, and minimizes energy consumption, providing a novel application of a deep learning algorithm in construction engineering. • Demand-controlled ventilation method (DCV) based on the BiLSTM algorithm. • The ventilation time of DCV is 62 s shorter than the current ventilation method. • The DCV method results in a 64.6% decrease in fan energy consumption at 1000 s. • The COVPs for both methods are 2.25 (DCV) and 0.741 (constant air-volume). • DCV offers a new decontaminating and energy-saving route for engineering practice. [ABSTRACT FROM AUTHOR]

[Display omitted] • The spectra-image features of defective maize kernels were obtain by HSI. • CNN and spectral-spatial attention were combined to recognize the HSI features. • The CNN-Spl-Spal-At based on fusion features had the excellent performance. • The identification results of various defective kernels were visualized. Rapid, effective and non-destructive detection of the defective maize kernels is crucial for their high-quality storage in granary. Hyperspectral imaging (HSI) coupled with convolutional neural network (CNN) based on spectral and spatial attention (Spl-Spal-At) module was proposed for identifying the different types of maize kernels. The HSI data within 380–1000 nm of six classes of sprouted, heat-damaged, insect-damaged, moldy, broken and healthy kernels was collected. The CNN-Spl-At, CNN-Spal-At and CNN-Spl-Spal-At models were established based on the spectra, images and their fusion features as inputs for the recognition of different kernels. Further compared the performances of proposed models and conventional models were built by support vector machine (SVM) and extreme learning machine (ELM). The results indicated that the recognition ability of CNN with attention series models was significantly better than that of SVM and ELM models and fused features were more conducive to expressing the appearance of different kernels than single features. And the CNN-Spl-Spal-At model had an optimal recognition result with high average classification accuracy of 98.04 % and 94.56 % for the training and testing sets, respectively. The recognition results were visually presented on the surface image of kernels with different colors. The CNN-Spl-Spal-At model was built in this study could effectively detect defective maize kernels, and it also had great potential to provide the analysis approaches for the development of non-destructive testing equipment based on HSI technique for maize quality. [ABSTRACT FROM AUTHOR]

It is significant to identify the moldy status of stored maize by fungi infection in the early stage. Hyperspectral imaging (HSI) combined with the sparse auto-encoders (SAE) and convolutional neural network (CNN) algorithms was used to classify the moldy grades of maize kernels. The HSI data were obtained in the range of 400–1000 nm, and four grades from health to heavy mildew were distinguished using the measured fungal spores of maize. The depth spectral features were represented using SAE and the image features were extracted by CNN. K nearest neighbors, support vector machine (SVM), and partial least squares discriminant analysis classifiers were combined with the spectral and image features to establish classification models to identify the different moldy grades of maize kernels. The comparison results indicated that the fusion of SAE and CNN combined with the SVM classifier to construct the SAE-CNN-SVM model had the most satisfactory identification result with high correct recognition rates of 99.47% and 98.94% for the training and testing sets, respectively, and the values of sensitivity and specificity were 0.95–1. The moldy grades were presented intuitively on the maize image based on pixels or kernel-wise. Therefore, the HSI with the SAE-CNN-SVM model had good recognition ability for the early detection of moldy maize kernels, which could potentially provide technical support for the development of online detection of moldy maize kernels during storage. [ABSTRACT FROM AUTHOR]

The use of positron emission tomography (PET) as the initial or sole biomarker of β-amyloid (Aβ) brain pathology may inhibit Alzheimer's disease (AD) drug development and clinical use due to cost, access, and tolerability. We developed a qEEG-ML algorithm to predict Aβ pathology among subjective cognitive decline (SCD) and mild cognitive impairment (MCI) patients, and validated it using Aβ PET. We compared QEEG data between patients with MCI and those with SCD with and without PET-confirmed beta-amyloid plaque. We compared resting-state eyes-closed electroencephalograms (EEG) patterns between the amyloid positive and negative groups using relative power measures from 19 channels (Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, O2), divided into eight frequency bands, delta (1–4 Hz), theta (4–8 Hz), alpha 1 (8–10 Hz), alpha 2 (10–12 Hz), beta 1 (12–15 Hz), beta 2 (15–20 Hz), beta 3 (20–30 Hz), and gamma (30–45 Hz) calculated by FFT and denoised by iSyncBrain®. The resulting 152 features were analyzed using a genetic algorithm strategy to identify optimal feature combinations and maximize classification accuracy. Guided by gene modeling methods, we treated each channel and frequency band of EEG power as a gene and modeled it with every possible combination within a given dimension. We then collected the models that showed the best performance and identified the genes that appeared most frequently in the superior models. By repeating this process, we converged on a model that approximates the optimum. We found that the average performance increased as this iterative development of the genetic algorithm progressed. We ultimately achieved 85.7% sensitivity, 89.3% specificity, and 88.6% accuracy in SCD amyloid positive/negative classification, and 83.3% sensitivity, 85.7% specificity, and 84.6% accuracy in MCI amyloid positive/negative classification. [ABSTRACT FROM AUTHOR]

Compared with invasive fractional flow reserve (FFR), coronary CT angiography (cCTA) is limited in detecting hemodynamically relevant lesions. cCTA-based FFR (CT-FFR) is an approach to overcome this insufficiency by use of computational fluid dynamics. Applying recent innovations in computer science, a machine learning (ML) method for CT-FFR derivation was introduced and showed improved diagnostic performance compared to cCTA alone. We sought to investigate the influence of stenosis location in the coronary artery system on the performance of ML-CT-FFR in a large, multicenter cohort. Three hundred and thirty patients (75.2% male, median age 63 years) with 502 coronary artery stenoses were included in this substudy of the MACHINE (Machine Learning Based CT Angiography Derived FFR: A Multi-Center Registry) registry. Correlation of ML-CT-FFR with the invasive reference standard FFR was assessed and pooled diagnostic performance of ML-CT-FFR and cCTA was determined separately for the following stenosis locations: RCA, LAD, LCX, proximal, middle, and distal vessel segments. ML-CT-FFR correlated well with invasive FFR across the different stenosis locations. Per-lesion analysis revealed improved diagnostic accuracy of ML-CT-FFR compared with conventional cCTA for stenoses in the RCA (71.8% [95% confidence interval, 63.0%–79.5%] vs. 54.8% [45.7%–63.8%]), LAD (79.3 [73.9–84.0] vs. 59.6 [53.5–65.6]), LCX (84.1 [76.0–90.3] vs. 63.7 [54.1–72.6]), proximal (81.5 [74.6–87.1] vs. 63.8 [55.9–71.2]), middle (81.2 [75.7–85.9] vs. 59.4 [53.0–65.6]) and distal stenosis location (67.4 [57.0–76.6] vs. 51.6 [41.1–62.0]). In a multicenter cohort with high disease prevalence, ML-CT-FFR offered improved diagnostic performance over cCTA for detecting hemodynamically relevant stenoses regardless of their location. [ABSTRACT FROM AUTHOR]

The Qilian Mountains (QLM) are an important ecological barrier in western China. High-precision land cover data products are the basic data for accurately detecting and evaluating the ecological service functions of the QLM. In order to study the land cover in the QLM and performance of different remote sensing classification algorithms for land cover mapping based on the Google Earth Engine (GEE) cloud platform, the higher spatial resolution remote sensing images of Sentinel-1 and Sentinel-2; digital elevation data; and three remote sensing classification algorithms, including the support vector machine (SVM), the classification regression tree (CART), and the random forest (RF) algorithms, were used to perform supervised classification of Sentinel-2 images of the QLM. Furthermore, the results obtained from the classification process were compared and analyzed by using different remote sensing classification algorithms and feature-variable combinations. The results indicated that: (1) the accuracy of the classification results acquired by using different remote sensing classification algorithms were different, and the RF had the highest classification accuracy, followed by the CART and the SVM; (2) the different feature variable combinations had different effects on the overall accuracy (OA) of the classification results and the performance of the identification and classification of the different land cover types; and (3) compared with the existing land cover products for the QLM, the land cover maps obtained in this study had a higher spatial resolution and overall accuracy. [ABSTRACT FROM AUTHOR]

Source apportionment can be an effective tool in mitigating soil pollution but its efficacy is often limited by a lack of information on the factors that influence the accumulation of pollutants at a site. In response to this limitation and focusing on a suite of heavy metals identified as priorities for pollution control, the study established a comprehensive pollution control framework using factor identification coupled with spatial agglomeration for agricultural soils in an industrialized part of Zhejiang Province, China. In addition to elucidating the key role of industrial and traffic activities on heavy metal accumulation through implementing a receptor model, specific influencing factors were identified using a random forest model. The distance from the soil sample location to the nearest likely industrial source was the most important factor in determining cadmium and copper concentrations, while distance to the nearest road was more important for lead and zinc pollution. Soil parent materials, pH, organic matter, and clay particle size were the key factors influencing accumulation of arsenic, chromium, and nickel. Spatial auto-correlation between levels of soil metal pollution and industrial agglomeration can enable a more targeted approach to pollution control measures. Overall, the approach and results provide a basis for improved accuracy in source apportionment, and thus improved soil pollution control, at the regional scale. [Display omitted] • Provision of a framework that links heavy metals in soils with their source. • Identification of environmental influence over heavy metal contamination of soils. • Spatially integrated industrial agglomerations and heavy metal-contaminated soils. • Basis for targeted measures aimed at mitigating pollution and health risks. [ABSTRACT FROM AUTHOR]

Abstract Background Osteosarcoma (OSA) presents a clinical challenge and has a low 5‐year survival rate. Currently, the lack of advanced stratification models makes personalized therapy difficult. This study aims to identify novel biomarkers to stratify high‐risk OSA patients and guide treatment. Methods We combined 10 machine‐learning algorithms into 101 combinations, from which the optimal model was established for predicting overall survival based on transcriptomic profiles for 254 samples. Alterations in transcriptomic, genomic and epigenomic landscapes were assessed to elucidate mechanisms driving poor prognosis. Single‐cell RNA sequencing (scRNA‐seq) unveiled genes overexpressed in OSA cells as potential therapeutic targets, one of which was validated via tissue staining, knockdown and pharmacological inhibition. We characterized changes in multiple phenotypes, including proliferation, colony formation, migration, invasion, apoptosis, chemosensitivity and in vivo tumourigenicity. RNA‐seq and Western blotting elucidated the impact of squalene epoxidase (SQLE) suppression on signalling pathways. Results The artificial intelligence‐derived prognostic index (AIDPI), generated by our model, was an independent prognostic biomarker, outperforming clinicopathological factors and previously published signatures. Incorporating the AIDPI with clinical factors into a nomogram improved predictive accuracy. For user convenience, both the model and nomogram are accessible online. Patients in the high‐AIDPI group exhibited chemoresistance, coupled with overexpression of MYC and SQLE, increased mTORC1 signalling, disrupted PI3K–Akt signalling, and diminished immune infiltration. ScRNA‐seq revealed high expression of MYC and SQLE in OSA cells. Elevated SQLE expression correlated with chemoresistance and worse outcomes in OSA patients. Therapeutically, silencing SQLE suppressed OSA malignancy and enhanced chemosensitivity, mediated by cholesterol depletion and suppression of the FAK/PI3K/Akt/mTOR pathway. Furthermore, the SQLE‐specific inhibitor FR194738 demonstrated anti‐OSA effects in vivo and exhibited synergistic effects with chemotherapeutic agents. Conclusions AIDPI is a robust biomarker for identifying the high‐risk subset of OSA patients. The SQLE protein emerges as a metabolic vulnerability in these patients, providing a target with translational potential.

• A lightweight and efficient BLS-based crack detection model is proposed to automatically classify and identify crack images. • The bottleneck structure of MobileNetV3 is introduced as the convolution module of the proposed broad learning system for feature extraction. • By training and validating with public and realistic crack image datasets, the proposed model achieves fast and efficient incremental learning as training samples increase. • The sliding window algorithm is introduced to efficiently recognize and locate cracks in the large-size image. Concrete surface crack detection is an important means to ensure infrastructure security, including bridges and tunnels. In recent years, computer vision has been widely adopted for concrete surface crack detection, and the intelligent crack detection method based on computer vision has achieved good results with the breakthrough of the deep learning algorithm. However, there are also some challenges, such as complex architecture, high computing burden, and time-consuming training. In this paper, a lightweight broad learning system detection method called MobileNetV3-BLS is proposed. First, the inverted residual structure with the linear bottleneck of MobileNetV3 is introduced as the convolution module of the proposed broad learning system for feature extraction. Then the extracted features are employed to randomly generate both mapping and enhancement nodes in a broad learning system, which are used to train the crack detection model. The proposed MobileNetV3-BLS can improve crack detection accuracy and training speed compared with other classical deep learning networks. On the other hand, the proposed method is a dynamic network of which the weights can be updated quickly to achieve incremental learning by adding new nodes and samples. [ABSTRACT FROM AUTHOR]

• The nonlinear spatio-temporal correlations of the collected measurements from the distribution systems are utilized to develop an unsupervised deep LSTM-AE network to learn the normal dynamics of the distribution systems. • A false data injection attack detection method is developed and implemented by utilizing the developed LSTM-AE neural network model. • Wide-ranging comprehensive experiments on different systems are developed to evaluate the proposed approach performance with quantitative and comparative analysis. • The utilization of LSTM-AE provides two main advantages over other detection techniques: (i) better detection statistics for larger distribution systems, (ii) the unnecessity of labeled attack samples. The utilization of distributed generation units (DG) in power distribution systems has increased the complexity of system monitoring and operation. Numerous information and communication technologies have been adopted recently to overcome the challenges and complexities associated with the integration of DG units in distribution systems. However, these technologies have created wide opportunities for energy theft and other types of cyber-physical threats. False data injection attacks (FDIA) illustrate a challenging threat for distribution systems for these are very tough to detect in reality. In this manuscript, we propose a spatio-temporal learning algorithm that is able to acquire the normal dynamics of distribution systems to overcome possible FDIA. First, we use a long short-term memory autoencoder (LSTM-AE) in acquiring the usual dynamics. After that, we employ the unsupervised trained model in detecting the numerous potentials of FDIAs in distribution systems by assessing the residual error of every measurement sample. This developed method is purely data-driven. This enables it to be robust to the distribution systems' nonlinearities and uncertainties which overcomes the weaknesses of the proposed detection algorithms in the literature. The efficacy of the developed detection method is assessed via different test case scenarios with numerous basic and stealth FDIAs. [ABSTRACT FROM AUTHOR]

A new method relying on machine learning and resistivity to predict concentrations of petroleum hydrocarbon pollution in soil was proposed as a means of investigation and monitoring. Currently, determining pollutant concentrations in soil is primarily achieved through costly sampling and testing of numerous borehole samples, which carries the risk of further contamination by penetrating the aquifer. Additionally, conventional petroleum hydrocarbon geophysical surveys struggle to establish a correlation between survey results and pollutant concentration. To overcome these limitations, three machine learning models (KNN, RF, and XGBOOST) were combined with the geoelectrical method to predict petroleum hydrocarbon concentrations in the source area. The results demonstrate that the resistivity-based prediction method utilizing machine learning is effective, as validated by R-squared values of 0.91 and 0.94 for the test and validation sets, respectively, and a root mean squared error of 0.19. Furthermore, this study confirmed the feasibility of the approach using actual site data, along with a discussion of its advantages and limitations, establishing it as an inexpensive option to investigate and monitor changes in petroleum hydrocarbon concentration in soil. • The prediction of soil petroleum hydrocarbon concentration is achieved by machine learning and resistivity tomography method. • Building the mapping relationship between soil petroleum hydrocarbon pollution concentration and resistivity. • External validation data verify feasibility of predicting soil petroleum hydrocarbon contamination concentrations based on resistivity. [ABSTRACT FROM AUTHOR]

• The machine learning (ML) approaches are used for the impact displacement prediction of RC beams. • The ML methods are more applicable to evaluate the impact displacement than empirical models. • Ensemble ML methods are superior to classic ML ones in prediction performance. • XGBoost gives the best results on the impact displacement prediction among all models. The performance evaluation of structures under impact loading is of great importance for post-disaster emergency repair and maintenance. As the reinforced concrete (RC) beam is the mostly adopted structural component, the maximum impact displacement (MID) of an RC beam is one of the most important and common evaluation indexes. Currently, a number of empirical formulas and simplified dynamic models have been proposed to predict the MID of a reinforced concrete (RC) beam on the basis of limited experimental data. However, it is found that most existing models are only applicable to restrained scenarios, and meanwhile several models are hard to utilize. In order to ameliorate these issues, based on prevalent machine learning (ML) techniques, a data-driven approach is introduced for predicting the MID of an RC beam. In this study, 224 experiment samples are collected as a database and 12 representative ML algorithms are used to train the prediction model of an RC beam's MID. Furthermore, the permuted feature importance and partial dependence plot are adopted to interpret the mechanism behind the ML based model. [ABSTRACT FROM AUTHOR]

• MLSTM with attention mechanism is developed for improving monthly rainfall prediction. • Data decomposition, input selection and structure optimization are integrated to LSTM. • The forecasting performance is examined with 129 stations over Yangtze River basin. • Proposed MLSTM-AM improves monthly rainfall forecasting compared to reference models. In this study, a multiscale long short-term memory model with attention mechanism (MLSTM-AM) is proposed to improve the accuracy of monthly precipitation forecasting. In the MLSTM-AM model, à trous wavelet transform (ATWT) is first employed to decompose the standardized monthly precipitation anomaly and climate indices into a certain number of subseries with different time scales. Then at each of time scales, a long short-term memory (LSTM) model is constructed to predict the precipitation anomaly subseries by coupling with partial information (PI) algorithm for selecting the model inputs and attention mechanism (AM) for optimizing the model structure. Finally, all the predicted subseries are summed, and then the summed series is transformed as the monthly precipitation prediction. The proposed MLSTM-AM model is examined with 129 stations over the Yangtze River basin and compared with the MLSTM, LSTM and multiple linear regression (MLR). The MLSTM is formed by coupling the LSTM with the ATWT. For comparison, the inputs of all the models are identified by the PI algorithm. Results show that the prediction of all the models is more skillful in the western region than in the eastern region of the basin, which is strongly correlated with the spatial distribution of the variability of precipitation. The LSTM outperforms the MLR in terms of two performance evaluation indicators at most of stations. The MLSTM and MLSTM-AM improve monthly precipitation forecasts at all of 129 stations over the basin compared to the LSTM, and the MLSTM-AM provides the best prediction performance among all these models. [ABSTRACT FROM AUTHOR]