Hybrid approach for permeability prediction in porous media: combining FFT simulations with machine learning

Permeability of subsurface porous media is one of the primary factors that affect fluid transport in porous rock. However, accurate prediction of rock permeability is a challenging task due to its intricate pore network. Development of digital rocks provides an effective approach to reveal and characterize the pore network. In this paper, a combination of digital rock petrophysics and ensemble machine learning (ML) models is proposed to improve the permeability prediction of subsurface porous media. The permeability of the numerically generated porous samples as outputs was determined by the lattice Boltzmann method (LBM). The five most important parameters (porosity, tortuosity, fractal dimension, average pore diameter, and coordination number) were selected as inputs for the permeability prediction. To improve the accuracy, feature selection and ML methods comparisons were conducted. Three feature selection methods based on expert knowledge, correlation coefficient, and importance score were compared. Moreover, a comparison was performed on six ML methods (support vector machine, artificial neural network, decision tree, random forest, gradient-boosting machine, and Bayesian ridge regression) that were optimized by particle swarm optimization (PSO). The results indicated that (1) the feature selection based on the expert knowledge obtained a higher performance than the groups based on the correlation coefficient and importance score, implying the importance of expert knowledge on feature selection, and thus on ML performance; (2) artificial neural network with hyperparameter tuning achieved the best performance in predicting permeability; and (3) the optimized ML method outperformed the empirical equations in predicting permeability. In conclusion, this study provides a fast and reliable approach predicting permeability of subsurface porous media based on numerically generated porous images. Moreover, the proposed framework can be further extended to determine other petrophysical properties, for example, the relative permeability and thermal conductivity.

The complexity and heterogeneity of pore structure present significant challenges in accurate permeability estimation. Commonly used empirical formulas neglect its microscopic and topological characteristics, thus lacking accuracy and adaptability. While machine learning (ML) and deep learning (DL) models demonstrate promising performance, but encounter challenges of data availability, computational cost, and model interpretability. The present study aims to develop a more robust and accurate permeability prediction model via knowledge extraction from ML model. We first establish an ML model between permeability and the geometry‐topology characteristics of porous media using Extreme Gradient Boosting (XGBoost) algorithm. The data set used to fit ML model is prepared from 458 samples of different types of porous media. Using the SHapley Additive exPlanations (SHAP) value, the influence of each feature on permeability prediction is quantified. It is found that the closeness centrality (topology feature), tortuosity, porosity (macroscopic features) and throat diameter, throat length, pore diameter (pore network features) are vital for permeability prediction. Guided by partial dependence calculation, the unknown function relationship between permeability and the top six important features is established. The novel permeability prediction model incorporating topology feature improves the prediction accuracy and demonstrates strong applicability across diverse data sets. This new model presents an optimal balance between simplicity and performance, rendering it a compelling alternative for permeability prediction in porous media. The research provides a novel referable framework of knowledge extraction via ML to reveal the important features and establish the potential relationship that can be extended and applied in other research fields.

The thesis evaluates the machine learning (ML) and deep learning (DL) approaches’ performance in accurately detecting glaucoma based on optical coherence tomography tabular data and images from individuals of different ethnicities. While numerous studies have employed ML and DL techniques for glaucoma identification, their performance has not been evaluated across diverse ethnic groups. In addition, a DL approach utilizing the Swin Transformer architecture trained on the thickness map images of the retinal nerve fiber layer (RNFL) was also evaluated. This Swin transformer DL model demonstrated an AUC of 0.97 in the internal testing dataset (Asian) and 0.88 in the external testing dataset (Caucasian). However, like the ML classifiers trained on measured data, the DL approach which was trained on raw thickness map images also exhibited poor reproducibility across different datasets. To address these issues, a cross-sectional study design was employed to investigate both ML and DL’s model performance in glaucoma detection using OCT data from individuals of different ethnicities. The study included 514 Asian participants, consisting of 257 with glaucoma and 257 controls, to develop ML and DL classifiers. The trained classifiers were subsequently evaluated on two separate participant groups comprising 356 Asians and 138 Caucasians. Two machine learning classifiers were created using the two types of RNFL thickness, one using the original values extracted from OCT machines (measured RNFL), and the other generated from the compensation model. The compensation model is a multivariate regression trained on normal individuals. It corrects the 12-clock RNFL thicknesses for multiple demographic and anatomical parameters. Additionally, a deep learning model was developed using the Swin Transformer architecture based on the measured RNFL thickness map images from OCT. Explainable artificial intelligence techniques (CAM and SHAP) were utilized to better interpret the results. Performance metrics such as the area under the receiver operating characteristic curve (AUC), accuracy and sensitivity were employed to examine the effectiveness of different glaucoma detection models. Both machine learning (AUC = 0.96) and deep learning (AUC = 0.97) models demonstrated superior performance compared to the raw measured data (baseline, AUC = 0.93), in the internal testing dataset (Asian). However, in the external testing dataset (Caucasian), ML models utilizing the compensated data (AUC = 0.93) exhibited significantly better performance compared to ML models using the original measured data (AUC = 0.83) and the baseline (AUC = 0.82). Furthermore, the ML and DL models trained on measured data exhibited inadequate generalization ability across different ethnicities, whereas the ML model using the compensated data maintained its performance in the external testing dataset. These findings caution against the indiscriminate application of ML and DL models to patient cohorts of different ethnicities. They also suggest that incorporating the compensation model into the development of ML models may enhance their performance in glaucoma detection across diverse ethnicities. Overall, the study highlights the importance of accounting for anatomical variations across different ethnic groups when developing machine-learning models for glaucoma detection using OCT data.

Deep neural networks (DNNs) have shown success in image classification, with high accuracy in recognition of everyday objects. Performance of DNNs has traditionally been measured assuming human accuracy is perfect. In specific problem domains, however, human accuracy is less than perfect and a comparison between humans and machine learning (ML) models can be performed. In recognising everyday objects, humans have the advantage of a lifetime of experience, whereas DNN models are trained only with a limited image dataset. We have tried to compare performance of human learners and two DNN models on an image dataset which is novel to both, i.e. histological images. We thus aim to eliminate the advantage of prior experience that humans have over DNN models in image classification. Ten classes of tissues were randomly selected from the undergraduate first year histology curriculum of a Medical School in North India. Two machine learning (ML) models were developed based on the VGG16 (VML) and Inception V2 (IML) DNNs, using transfer learning, to produce a 10-class classifier. One thousand (1000) images belonging to the ten classes (i.e. 100 images from each class) were split into training (700) and validation (300) sets. After training, the VML and IML model achieved 85.67 and 89% accuracy on the validation set, respectively. The training set was also circulated to medical students (MS) of the college for a week. An online quiz, consisting of a random selection of 100 images from the validation set, was conducted on students (after obtaining informed consent) who volunteered for the study. 66 students participated in the quiz, providing 6557 responses. In addition, we prepared a set of 10 images which belonged to different classes of tissue, not present in training set (i.e. out of training scope or OTS images). A second quiz was conducted on medical students with OTS images, and the ML models were also run on these OTS images. The overall accuracy of MS in the first quiz was 55.14%. The two ML models were also run on the first quiz questionnaire, producing accuracy between 91 and 93%. The ML models scored more than 80% of medical students. Analysis of confusion matrices of both ML models and all medical students showed dissimilar error profiles. However, when comparing the subset of students who achieved similar accuracy as the ML models, the error profile was also similar. Recognition of ‘stomach’ proved difficult for both humans and ML models. In 04 images in the first quiz set, both VML model and medical students produced highly equivocal responses. Within these images, a pattern of bias was uncovered–the tendency of medical students to misclassify ‘liver’ tissue. The ‘stomach’ class proved most difficult for both MS and VML, producing 34.84% of all errors of MS, and 41.17% of all errors of VML model; however, the IML model committed most errors in recognising the ‘skin’ class (27.5% of all errors). Analysis of the convolution layers of the DNN outlined features in the original image which might have led to misclassification by the VML model. In OTS images, however, the medical students produced better overall score than both ML models, i.e. they successfully recognised patterns of similarity between tissues and could generalise their training to a novel dataset. Our findings suggest that within the scope of training, ML models perform better than 80% medical students with a distinct error profile. However, students who have reached accuracy close to the ML models, tend to replicate the error profile as that of the ML models. This suggests a degree of similarity between how machines and humans extract features from an image. If asked to recognise images outside the scope of training, humans perform better at recognising patterns and likeness between tissues. This suggests that ‘training’ is not the same as ‘learning’, and humans can extend their pattern-based learning to different domains outside of the training set.

Explainable machine learning model for predicting the occurrence of postoperative malnutrition in children with congenital heart disease

Multi-phase micro-scale flow simulation in the electrodes of a PEM fuel cell by lattice Boltzmann method

A machine learning (ML) model would be difficult to apply in practice without knowing how the prediction is made. To address this issue, this paper uses XAI and engineers’ expertise to interpret an ML model for bridge scour risk prediction. By using data from the French National Railway Company (SNCF), an extreme gradient boosting (XGBoost) algorithm-based ML model was constructed at first. Later, XAI approaches were employed to have global and local explanations as well as explicit expressions for ML model interpretation. Meanwhile, a group of engineers from SNCF were asked to rank the input parameters based on their engineering judgment. In the end, feature importance obtained from XAI approaches and engineers’ survey was compared. It was found that for both XAI and engineers’ interpretations, observation of local scour around the bridge foundation is the most important feature for decision-making. The differences between XAI interpretations and human expertise emphasize the importance of knowledge in hydrology and hydromorphology for scour risk assessment, since currently engineers make decisions primarily based on observed damages (e.g., scour hole, crack). The results of this paper could make the ML model trustworthy by understanding how the prediction is made and provide valuable guidance for improving current inspection procedures.

Endotracheal tube (ET) misplacement is common in pediatric patients, which can lead to the serious complication. It would be helpful if there is an easy-to-use tool to predict the optimal ET depth considering in each patient’s characteristics. Therefore, we plan to develop a novel machine learning (ML) model to predict the appropriate ET depth in pediatric patients. This study retrospectively collected data from 1436 pediatric patients aged < 7 years who underwent chest x-ray examination in an intubated state. Patient data including age, sex, height weight, the internal diameter (ID) of the ET, and ET depth were collected from electronic medical records and chest x-ray. Among these, 1436 data were divided into training (70%, n = 1007) and testing (30%, n = 429) datasets. The training dataset was used to build the appropriate ET depth estimation model, while the test dataset was used to compare the model performance with the formula-based methods such as age-based method, height-based method and tube-ID method. The rate of inappropriate ET location was significantly lower in our ML model (17.9%) compared to formula-based methods (35.7%, 62.2%, and 46.6%). The relative risk [95% confidence interval, CI] of an inappropriate ET location compared to ML model in the age-based, height-based, and tube ID-based method were 1.99 [1.56–2.52], 3.47 [2.80–4.30], and 2.60 [2.07–3.26], respectively. In addition, compared to ML model, the relative risk of shallow intubation tended to be higher in the age-based method, whereas the risk of the deep or endobronchial intubation tended to be higher in the height-based and the tube ID-based method. The use of our ML model was able to predict optimal ET depth for pediatric patients only with basic patient information and reduce the risk of inappropriate ET placement. It will be helpful to clinicians unfamiliar with pediatric tracheal intubation to determine the appropriate ET depth.

Machine learning approaches for formation matrix volume prediction from well logs: Insights and lessons learned

For recent years, Machine Learning (ML) models have been proven to be useful in solving problems of a wide variety of fields such as medical, economic, manufacturing, transportation, energy, education, etc. With increased interest in ML models and advances in sensor technologies, ML models are being widely applied even in civil engineering domain. ML model enables analysis of large amounts of data, automation, improved decision making and provides more accurate prediction. While several state-of-the-art reviews have been conducted in each sub-domain (e.g., geotechnical engineering, structural engineering) of civil engineering or its specific application problems (e.g., structural damage detection, water&amp;#160;quality evaluation), little effort has been devoted to comprehensive review on ML models applied in civil engineering and compare them across sub-domains. A systematic, but domain-specific literature review framework should be employed to effectively classify and compare the models. To that end, this study proposes a novel review approach based on the hierarchical classification tree &amp;#8220;D-A-M-I-E (Domain-Application problem-ML models-Input data-Example case)&amp;#8221;. &amp;#8220;D-A-M-I-E&amp;#8221; classification tree classifies the ML studies in civil engineering based on the (1) domain of the civil engineering, (2) application problem, (3) applied ML models and (4) data used in the problem. Moreover, data used for the ML models in each application examples are examined based on the specific characteristic of the domain and the application problem. For comprehensive review, five different domains (structural engineering, geotechnical engineering, water engineering, transportation engineering and energy engineering) are considered and the ML application problem is divided into five different problems (prediction, classification, detection, generation, optimization). Based on the &amp;#8220;D-A-M-I-E&amp;#8221; classification tree, about 300 ML studies in civil engineering are reviewed. For each domain, analysis and comparison on following questions has been conducted: (1) which problems are mainly solved based on ML models, (2) which ML models are mainly applied in each domain and problem, (3) how advanced the ML models are and (4) what kind of data are used and what processing of data is performed for application of ML models. This paper assessed the expansion and applicability of the proposed methodology to other areas (e.g., Earth system modeling, climate science). Furthermore, based on the identification of research gaps of ML models in each domain, this paper provides future direction of ML in civil engineering based on the approaches of dealing data (e.g., collection, handling, storage, and transmission) and hopes to help application of ML models in other fields.

BackgroundInpatients with high risk of venous thromboembolism (VTE) usually face serious threats to their health and economic conditions. Many studies using machine learning (ML) models to predict VTE risk overlook the impact of class-imbalance problem due to the low incidence rate of VTE, resulting in inferior and unstable model performance, which hinders their ability to replace the Padua model, a widely used linear weighted model in clinic. Our study aims to develop a new VTE risk assessment model suitable for Chinese medical inpatients.Methods3284 inpatients in the medical department of Peking Union Medical College Hospital (PUMCH) from January 2014 to June 2016 were collected. The training and test set were divided based on the admission time and inpatients from May 2016 to June 2016 were included as the test dataset. We explained the class imbalance problem from a clinical perspective and defined a new term, “fuzzy population”, to elaborate and model this phenomenon. By considering the “fuzzy population”, a new ML VTE risk assessment model was built through population splitting. Sensitivity and specificity of our method was compared with five ML models (support vector machine (SVM), random forest (RF), gradient boosting decision tree (GBDT), logistic regression (LR), and XGBoost) and the Padua model.ResultsThe ‘fuzzy population’ phenomenon was explained and verified on the VTE dataset. The proposed model achieved higher specificity (64.94% vs. 63.30%) and the same sensitivity (90.24% vs. 90.24%) on test data than the Padua model. Other five ML models couldn’t simultaneously surpass the Padua’s sensitivity and specificity. Besides, our model was more robust than five ML models and its standard deviations of sensitivities and specificities were smaller. Adjusting the distribution of negative samples in the training set based on the ‘fuzzy population’ would exacerbate the instability of performance of five ML models, which limited the application of ML methods in clinic.ConclusionsThe proposed model achieved higher sensitivity and specificity than the Padua model, and better robustness than traditional ML models. This study built a population-split-based ML model of VTE by modeling the class-imbalance problem and it can be applied more broadly in risk assessment of other diseases.

Machine learning (ML) and deep learning (DL) models can play an important role when it comes to modelling complicated processes. Such capability is necessary for hydrological and climate-related applications. Generally, ML models utilize precipitation and temperature time series of a basin as input to develop a lumped rainfall-runoff model to simulate streamflow at the basin outlet. However, when it is divided into several sub-basins, Graph Neural Networks (GNN) can consider each sub-basin as a node and link them together using a connectivity matrix to account for spatial variations of hydroclimatic variables. In this study, GNN and various ML models with different types of architecture, ranging from neural networks, tree-based structure, and gradient boosting, were exploited for daily streamflow simulation over different case studies. For each case study, the basin was divided into a few sub-basins for which daily precipitation and temperature data were aggregated and used as input. For training GNN, the connection matrix of sub-basins was also used as input. Basically, 75% of historical records were utilized to train GNN and different ML models, e.g., artificial neural networks, support vector machine, decision tree, random forest, eXtreme Gradient Boosting (XGBoost), Light Gradient-Boosting Machine (LightGBM), and Category Boosting (CatBoost), while the rest was used for testing. Streamflow simulation was conducted with/without considering seasonality impact and lag times. The obtained results clearly demonstrate that considering seasonality and time lags can enhance accuracy of streamflow predictions based on Kling–Gupta efficiency (KGE). Furthermore, GNN with seasonality impact and time lags achieved promising results across different case studies with KGE&gt;0.85 for training and KGE&gt;0.59 for testing data, respectively. Among ML models, boosting models, e.g., LightGBM and XGBoost, performed slightly better than other ML models. for Finally, this comparative analysis provides valuable insights for ML/DL applications in climate change impact assessments.Acknowledgements: This research work was carried out as part of the TRANSCEND project with funding received from the European Union Horizon Europe Research and Innovation Programme under Grant Agreement No. 10108411.

Skeletal muscle loss during treatment is associated with poor survival outcomes in patients with ovarian cancer. Although changes in muscle mass can be assessed on computed tomography (CT) scans, this labour-intensive process can impair its utility in clinical practice. This study aimed to develop a machine learning (ML) model to predict muscle loss based on clinical data and to interpret the ML model by applying SHapley Additive exPlanations (SHAP) method. This study included the data of 617 patients with ovarian cancer who underwent primary debulking surgery and platinum-based chemotherapy at a tertiary centre between 2010 and 2019. The cohort data were split into training and test sets based on the treatment time. External validation was performed using 140 patients from a different tertiary centre. The skeletal muscle index (SMI) was measured from pre- and post-treatment CT scans, and a decrease in SMI≥5% was defined as muscle loss. We evaluated five ML models to predict muscle loss, and their performance was determined using the area under the receiver operating characteristic curve (AUC) and F1 score. The features for analysis included demographic and disease-specific characteristics and relative changes in body mass index (BMI), albumin, neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR). The SHAP method was applied to determine the importance of the features and interpret the ML models. The median (inter-quartile range) age of the cohort was 52 (46-59) years. After treatment, 204 patients (33.1%) experienced muscle loss in the training and test datasets, while 44 (31.4%) patients experienced muscle loss in the external validation dataset. Among the five evaluated ML models, the random forest model achieved the highest AUC (0.856, 95% confidence interval: 0.854-0.859) and F1 score (0.726, 95% confidence interval: 0.722-0.730). In the external validation, the random forest model outperformed all ML models with an AUC of 0.874 and an F1 score of 0.741. The results of the SHAP method showed that the albumin change, BMI change, malignant ascites, NLR change, and PLR change were the most important factors in muscle loss. At the patient level, SHAP force plots demonstrated insightful interpretation of our random forest model to predict muscle loss. Explainable ML model was developed using clinical data to identify patients experiencing muscle loss after treatment and provide information of feature contribution. Using the SHAP method, clinicians may better understand the contributors to muscle loss and target interventions to counteract muscle loss.

ObjectiveThis retrospective, case-control study with internal validation evaluates the performance of machine learning (ML) and deep learning (DL) models in classifying pediatric patients at risk for anxiety disorders using structured electronic health records (EHRs) and area-based measures of health (ABMH). The aim is to enable proactive care by monitoring potential anxiety onset across developmental stages.MethodsWe trained a series of ML models (Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbors, XGBoost) and DL models (LSTM, GRU, RETAIN, Dipole) using structured EHR data from 30-day windows prior to diagnosis. Two datasets were used per age group: one with structured EHR data only, and another including both EHR and ABMH data. ML models were trained using short-term cross-sectional features, while DL models leveraged full longitudinal patient histories. Performance was assessed using AUROC, AUPRC, PPV, NPV, F1 score, and accuracy. Due to differences in input scope, model performance reflects both algorithmic and temporal design differences and is not intended as a direct comparison between ML and DL.ResultsML models offered strong baseline performance, with XGBoost achieving AUROC scores of 0.817 (EHR) and 0.816 (EHR+ABMH) for 8-year-olds. Adding ABMH features did not significantly improve performance. DL models, particularly RETAIN and Dipole, achieved the highest AUROC values (e.g., Dipole: 0.853 with EHR, 0.857 with EHR+ABMH for 8-year-olds), outperforming other DL and ML models within their respective design constraints.ConclusionBoth ML and DL models successfully identified likely anxiety onset using structured EHR data. DL models using longitudinal data achieved the highest performance, while XGBoost provided a robust ML baseline. The minimal impact of ABMH features highlights integration challenges, and performance variation across ages emphasizes the need for age-stratified modeling approaches.

ABSTRACTBackgroundCoronary artery plaque rupture (PR) is closely associated with immune‐inflammatory responses. The systemic inflammatory index (SII) and the systemic inflammatory response index (SIRI) have shown potential in predicting the occurrence of PR.ObjectiveThis study aims to establish a machine learning (ML) model that integrates baseline patient characteristics, SII, and SIRI to predict PR. The goal is to identify high‐risk PR patients before intravascular imaging examinations.MethodsWe included 337 patients with acute coronary syndrome who underwent emergency percutaneous coronary intervention and coronary optical coherence tomography (OCT) at the Affiliated Hospital of Zunyi Medical University, China, from May 2023 to October 2023. PR was determined by OCT images. Through manual feature selection, nine features, including SII and SIRI, were included, and an ML model was built using the XGBoost algorithm. Model performance was evaluated using receiver operating characteristic curves and calibration curves. SHAP values were used to assess the contribution of each feature to the model.ResultsThe ML model demonstrated a higher area under the curve value (AUC = 0.81) compared to using SII or SIRI alone for prediction. The ML model also showed good calibration. SHAP values revealed that the top three features in the ML model were SII, LDL‐C, and SIRI.ConclusionThe immuno‐inflammatory index, which integrates comprehensive clinical characteristics, can predict the occurrence of PR. However, large‐scale, multicenter studies are needed to confirm the generalizability of the predictive model.