Artificial Intelligence-based Models Research Articles

Artificial intelligence-based modeling and optimization of microbial electrolysis cell-assisted anaerobic digestion fed with alkaline-pretreated waste-activated sludge

Microbial electrolysis cell-assisted anaerobic digestion (MEC-AD) is a promising emerging strategy to enhance simultaneously waste treatment and biomethane recovery from various biowastes, particularly waste-activated sludge (WAS). However, MEC-AD is still in the early stages of development, with numerous experimental studies but no modeling or optimization. Thus, to provide an effective modeling and optimization tool for this process, this study proposed applying artificial intelligence for the first time. The literature-based experimental data of MEC-AD fed with alkaline-pretreated waste-activated sludge (al-WAS) were used for this purpose. Accordingly, a two-hidden-layer artificial neural network (ANN) with topology 2–25–34–6, obtained from the response surface methodology, showed the best agreement between actual and predicted data with a low mean squared error of 0.0579 and a high R-value of 0.9870. This best ANN model was then optimized by particle swarm optimization. As a result, an Eapp of 0.63 V was found to be optimal for al-WAS-fed MEC-AD with a highest net energy output of 41.3 KJ/L-reactor (∼2.6 MJ/kg-WAS) and highest net monetary value of 0.72 $/L-reactor (∼45 $/kg-WAS), which enhanced around 160 % and 300 % compared to AD alone. These findings can support decision-making for managers and operators in wastewater treatment, biomass waste management, and renewable energy sectors.

Application of Soft Computing Techniques for Predicting Thermal Conductivity of Rocks

Due to the different challenges in rock sampling and in measuring their thermal conductivity (TC) in the field and laboratory, the determination of the TC of rocks using non-invasive methods is in demand in engineering projects. The relationship between TC and non-destructive tests has not been well-established. An investigation of the most important variables affecting the TC values for rocks was conducted in this study. Currently, the black-boxed models for TC prediction are being replaced with artificial intelligence-based models, with mathematical equations to fill the gap caused by the lack of a tangible model for future studies and developments. In this regard, two models were developed based on which gene expression programming (GEP) algorithms and non-linear multivariable regressions (NLMR) were utilized. When comparing the performances of the proposed models to that of other previously published models, it was revealed that the GEP and NLMR models were able to produce more accurate predictions than other models were. Moreover, the high value of R-squared (equals 0.95) for the GEP model confirmed its superiority.

AI-based ensemble modeling of landfill leakage employing a lysimeter, climatic data and transfer learning

• Field measurement of EC adjoined by artificial intelligence-based model to conduct leachate induced pollution in landfill. • To monitor the EC, waste temperature, and moisture, two lysimeter tests were conducted. • ANN, ANFIS and Emotional ANN (EANN) models were developed to determine the parameters affecting the EC. • Transfer learning method were used to predict the missing EC in a lysimeter. Predicting leachate pollutants is of prime importance in detecting the amount of pollution in water resources adjacent to sources of leakage. In this study, Electrical Conductivity (EC) as a physicochemical water pollution parameter with the possibility of portable measurement was used as an indicator of leachate quality for the Tychy-Urbanowice operating and closed landfill complex. In order to simulate landfill conditions, two lysimeter experiments were conducted simultaneously. Using sensors mounted in the lysimeters, from the end of November 2018 to the end of December 2019, EC, waste temperature and waste moisture were measured for the open lysimeter and only waste moisture for the closed lysimeter. Additionally, meteorological data obtained from the nearest synoptic station and soil moisture and temperature acquired from the GLDAS satellite were employed as external data to analyze various conditions. Thereafter, Artificial Neural Network (ANN), Neuro-Fuzzy Inference System (ANFIS), and Emotional ANN (EANN) models were developed to determine the parameters affecting the EC value recorded for the open lysimeter and subsequently, predict the missing EC parameter of the closed lysimeter by employing the transfer learning method. Following that, in order to improve the precision of EC predictions, ensemble techniques were applied to the outputs of the models that were developed. The results showed that the moisture of the lysimeters made a significant contribution to the EC value prediction. It is worth mentioning that among ANN, ANFIS, and EANN, the EANN model yielded more precise results in EC estimation, with the average DC above 0.80 and 0.90 for individual and ensembled modeling in both the training and verification phases, respectively.

Modeling the flow behavior of Haynes 214 superalloy during hot deformation using mathematical and artificial intelligence-based models

This work proposes, enhances, and compares various mathematical and artificial intelligence-based models for the modeling and prediction of the flow behavior of Haynes 214 superalloy at hot deformation. The utilized models are as follows: Johnson-Cook (JC), three modifications of JC (M1_JC, M2_JC, and M3_JC), Artificial Neural Network (ANN), and Subtractive Clustering-Fuzzy Interference System (SC-FIS). The predictions of the flow behavior are evaluated and assessed using various statistical error measures, namely, correlation coefficient (R), Relative Error (RE), and Root Mean Square Error (RMSE). The results showed that the M3_JC is the best addressed mathematical model in terms of flow prediction accuracy, while the Artificial Intelligence (AI) based SC-FIS model outperformed all of the six addressed mathematical and AI-based models with an R value of 0.999, RE range of − 0.79–1.15% and an RMSE value of as low as 0.89 MPa.

Artificial Intelligence for the Prediction of In-Hospital Clinical Deterioration: A Systematic Review.

OBJECTIVES:To analyze the available literature on the performance of artificial intelligence-generated clinical models for the prediction of serious life-threatening events in non-ICU adult patients and evaluate their potential clinical usage.DATA SOURCES:The PubMed database was searched for relevant articles in English literature from January 1, 2000, to January 23, 2022. Search terms, including artificial intelligence, machine learning, deep learning, and deterioration, were both controlled terms and free-text terms.STUDY SELECTION:We performed a systematic search reporting studies that showed performance of artificial intelligence-based models with outcome mortality and clinical deterioration.DATA EXTRACTION:Two review authors independently performed study selection and data extraction. Studies with the same outcome were grouped, namely mortality and various forms of deterioration (including ICU admission, adverse events, and cardiac arrests). Meta-analysis was planned in case sufficient data would be extracted from each study and no considerable heterogeneity between studies was present.DATA SYNTHESIS:In total, 45 articles were included for analysis, in which multiple methods of artificial intelligence were used. Twenty-four articles described models for the prediction of mortality and 21 for clinical deterioration. Due to heterogeneity of study characteristics (patient cohort, outcomes, and prediction models), meta-analysis could not be performed. The main reported measure of performance was the area under the receiver operating characteristic (AUROC) (n = 38), of which 33 (87%) had an AUROC greater than 0.8. The highest reported performance in a model predicting mortality had an AUROC of 0.935 and an area under the precision-recall curve of 0.96.CONCLUSIONS:Currently, a growing number of studies develop and analyzes artificial intelligence-based prediction models to predict critical illness and deterioration. We show that artificial intelligence-based prediction models have an overall good performance in predicting deterioration of patients. However, external validation of existing models and its performance in a clinical setting is highly recommended.

Experimental and Probabilistic Investigations of the Effect of Fly Ash Dosage on Concrete Compressive Strength and Stress-strain Relationship

The effect of fly ash (FA) dosage on concrete’s compressive strength and stress-strain relationship is investigated in two steps in this article. First, an experimental program was conducted on concrete mixtures designed with 0% (control batch of 30 MPa mean cylinder compressive strength), 10, 20, 30, and 40% of ordinary Portland cement (OPC) mass replaced by FA, which is taken from a new source in an Asia country. The test results showed that compared to other investigated dosages, concrete using 20%FA/OPC mass-replacement gained the most improvement in the 28-day compressive strength and tensile split strength, as well as the compressive strength development. Second, a probabilistic investigation was conducted using Dropout Neural Network, Bayesian Neural Network, and Gaussian Process models. These artificial intelligence-based models were compared to other models reviewed from the literature, showing relatively good results in terms of the statistical metric R2, which are 0.92, 0.9, and 0.88, respectively. The three models were tested and validated with a dataset of 1032 experimental results on FAC collected from the literature. When testing with the experimental results obtained in the first step, a good correlation between the predicted values and the experimental results was observed within the confidence interval of (5%, 95%), showing the reliability of the proposed models. Thus, the stress-strain relationship of fly ash concrete can also be investigated in a probabilistic manner. It is proved in this study that among the proposed models, Dropout Neural Network has the best balance between performance and time complexity.

Investigation of quantitative and qualitative changes in groundwater of Ardebil plain using ensemble artificial intelligence-based modeling

Abstract Groundwater is an essential source to supply water for various sectors. This paper aimed to predict the quantitative and qualitative changes in groundwater over time and to evaluate the efficiency of different modeling methods. This study is based on three steps. In the first step, quantitative and qualitative piezometers were clustered by the Growing Neural Gas Network (GNG) method, and the central piezometer of each cluster was used on behalf of each cluster. In the second step, four different Artificial Intelligence (AI) models were applied, namely Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), Support Vector Machine (SVM), and Emotional Artificial Neural Network (EANN). As a post-processing approach three different ensemble methods were used: simple average ensemble (SAE), weighted average ensemble (WAE), and nonlinear neural network ensemble (NNE). In the third step, the outputs of single AI models were used to enhance the evaluation results. Therefore, the results demonstrate that the NNE led to reach the better performance for three GWL, TDS, and TH parameters up to 37, 29, and 23% on average, respectively. Study results will lead to the improvement of AI applications in groundwater research and will benefit groundwater development plans.

Artificial intelligence-based modeling of extruded aluminum beams subjected to patch loading

Artificial intelligence (AI) has become a reliable tool for the solution of structural engineering problems. This paper aims at developing prediction models for the behavior of extruded aluminum beams under patch loading via symbolic regression (SR) and artificial neural networks (ANN). The models are constructed employing an extensive dataset calculated numerically through nonlinear finite element analysis. This dataset covers a wide range of geometric parameters and considers aluminum alloys with different strain hardening characteristics. Explicit formulations for the resistance of extruded aluminum beams to patch loading are developed by fitting this dataset using SR and ANN, the results are compared with those of the resistance model in the EC9:1-1. Finally, the performance of the proposed AI models is also evaluated. Results confirmed the superior accuracy of the proposed models.

Simulating and modeling CO2 flux emitted from decomposed oil palm root cultivated at tropical peatland as affected by water content and residence time

Determining the oil palm dead roots contribution to total (R<sub>t</sub>) and heterotrophic (R<sub>h</sub>) respiration as a source of greenhouse gas/GHG emission in tropical peatland is urgently required, as well as predicting their magnitude to cope with difficulties of direct in-situ measurement. This study is designed to simulate the CO<sub>2</sub> flux emitted from oil palm dead roots/R<sub>dr</sub> in tropical peatland as affected by water content/WC and residence time/RT. The dead oil palm roots were cleaned, treated with control/15, 100, 150, 300, and 450%WC, and then incubated for three months. CO<sub>2</sub> flux measurement, C, N, and CN ratio determination were conducted every month. This study demonstrated the importance R<sub>dr</sub> among other CO<sub>2</sub> emission sources, ranging from 0.05-2.3 Mg CO<sub>2</sub> ha<sup>-1</sup> year<sup>-1</sup> with an average of 0.7 Mg CO<sub>2</sub> ha<sup>-1</sup> year<sup>-1</sup>. R<sub>dr</sub> contribution for literature R<sub>t</sub> and R<sub>h</sub> were around 0.3 to 1.3 and 0.9 to 3.5%, respectively. As a product of microbial respiration, R<sub>dr</sub> was affected by WC and RT, supported by analysis of variance, linear mixed effect model/REML, and multivariate analysis. 100-150%WC resulting in significant and highest R<sub>dr</sub>, whereas the increase (300-450%WC) or decrease (15%WC) would generate lower emission. R<sub>dr</sub> culminated in the first month after incubation; meanwhile, it declined in the following months. This study also emphasized non-linear relationships between CO<sub>2</sub> flux and other root properties, which can be modeled conveniently using non-linear approach, particularly using polynomial and artificial intelligence-based models. The simulation presented in this study served as an initial attempt to separate R<sub>dr</sub> from R<sub>h</sub>, as well as to predict CO<sub>2</sub> flux with reasonable accuracy and interpretable methods.

AI-Based Particle Position Prediction Near Southwestern Area of Jeju Island

Positions of five drifting buoys deployed on August 2020 near southwestern area of Jeju Island and numerically predicted velocities were used to develop five Artificial Intelligence-based models (AI models) for the prediction of particle tracks. Five AI models consisted of three machine learning models (Extra Trees, LightGBM, and Support Vector Machine) and two deep learning models (DNN and RBFN). To evaluate the prediction accuracy for six models, the predicted positions from five AI models and one numerical model were compared with the observed positions from five drifting buoys. Three skills (MAE, RMSE, and NCLS) for the five buoys and their averaged values were calculated. DNN model showed the best prediction accuracy in MAE, RMSE, and NCLS.

Design of an Artificial Neural Network-Based Model for Prediction Solar Radiation Utilizing Measured Weather Datasets

Forecasting solar radiation plays an important role in the field of energy meteorology, as it provides the energy value expected to be produced by the solar plants on a specific day and time of the year. In this paper, a new and reliable artificial intelligence-based model for solar radiation prediction is presented using Artificial Neural Network (ANN). The proposed model is built utilizing real atmospheric affecting measured values according to their locational weather station. In the training process, the Levenberg–Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG) are used. The mean absolute error (MAE) and the root mean square error (RMSE) are used to evaluate the model accuracy. Results of the investigation show that the proposed model provides the lowest error rate when using the (BR) training algorithm for predicting the average daily solar radiation.

Abstract 6221: Tandem use of multi-omics and artificial intelligence identified novel prognostic and therapeutic biomarkers for head and neck squamous cell carcinoma

Abstract Introduction: Head and neck squamous cell carcinoma (HNSCC) is categorized into Human Papilloma Virus positive (HPV+ve) and negative (HPV-ve) tumors, with the latter being more aggressive and displaying poor overall survival (OS). There is a need to identify aggressive biomarkers for patient stratification and improved patient care. Materials and Methods: Using the multi-omics data, including somatic mutations, copy number alterations, RNA sequencing, miRNA sequencing, and methylation from 513 HNSCC patients, we performed an integrative clustering analysis to differentiate HPV+ve and HPV-ve tumors and identify novel prognostic biomarkers for survival and treatment response. Using artificial intelligence (AI) based machine learning approach by implementing scikit-learn, we developed a model to predict the patient's OS. Results: Our clustering analysis robustly segregated HNSCC patients into HPV+ve and HPV-ve tumors. Artificial intelligence-based machine learning model identified a 30 gene signature panel specific to HPV-ve tumors precisely predicting poor OS with an accuracy of 90.5±0.8% (95% CI) and AUC of 80.2%. Furthermore, the multivariate Cox regression survival model identified 12 genes as independent prognostic biomarkers in HPV-ve tumors, including overexpression of novel genes NT5E and TRIML2. Interestingly, higher expression of NT5E and TRIML2 due to promoter hypomethylation predicted therapy resistance in 62.8% and 65.1% of HNSCC patients, respectively. The prognostic signature panel of 30 genes was further validated using three additional expression data sets (n = 450) of HNSCC. Conclusion: Our multi-omics and AI analysis revealed a novel HPV-ve molecular cluster with implications for predicting disease aggressiveness, therapeutic response, and OS in HNSCC patients. Citation Format: Ajaz A. Bhat, Tariq Masoodi, Deepika Mishra, Mayank Singh, Sabah Nisar, Sheema Hashem, Sana K. Baba, Puneet Bagga, Ravinder Reddy, Davide Bedognetti, Shahab Uddin, Wael El-Rifai, Muzafar A. Macha, Mohammad Haris. Tandem use of multi-omics and artificial intelligence identified novel prognostic and therapeutic biomarkers for head and neck squamous cell carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6221.

Exploring the Efficiencies of Spectral Isolation for Intelligent Wear Monitoring of Micro Drill Bit Automatic Regrinding In-Line Systems

Despite the increasing digitalization of equipment diagnostic/condition monitoring systems, it remains a challenge to accurately harness discriminant information from multiple sensors with unique spectral (and transient) behaviors. High-precision systems such as the automatic regrinding in-line equipment provide intelligent regrinding of micro drill bits; however, immediate monitoring of the grinder during the grinding process has become necessary because ignoring it directly affects the drill bit’s life and the equipment’s overall utility. Vibration signals from the frame and the high-speed grinding wheels reflect the different health stages of the grinding wheel and can be exploited for intelligent condition monitoring. The spectral isolation technique as a preprocessing tool ensures that only the critical spectral segments of the inputs are retained for improved diagnostic accuracy at reduced computational costs. This study explores artificial intelligence-based models for learning the discriminant spectral information stored in the vibration signals and considers the accuracy and cost implications of spectral isolation of the critical spectral segments of the signals for accurate equipment monitoring. Results from one-dimensional convolutional neural networks (1D-CNN) and multi-layer perceptron (MLP) neural networks, respectively, reveal that spectral isolation offers a higher condition monitoring accuracy at reduced computational costs. Experimental results using different 1D-CNN and MLP architectures reveal 4.6% and 7.5% improved diagnostic accuracy by the 1D-CNNs and MLPs, respectively, at about 1.3% and 5.71% reduced computational costs, respectively.

Application of random forest for modelling of surface water salinity

Modeling surface water quality using artificial intelligence-based models is essential in projecting suitable mitigation measures. However, it remains a challenge and requires further research to enhance the modeling accuracy. For this aim, this article presents a methodology to optimize the modeling inputs and reduce the associated complexity. The proposed approach employs Random Forest for modeling surface water salinity in terms of electrical conductivity (EC) and total dissolved solids (TDS) in the upper Indus River basin, one of the major rivers in Asia. Various water quality parameters measured monthly over a historical 30-year period were utilized in the modeling process. Various statistical indicators were used to evaluate the model performance. Random Forest process is suitable technique to simulate the salinity of surface water bodies, and effective tool in minimizing the modeling complexity and elaborating proper management and mitigation measures.

Comparative study of SARS COVID-19 using X-ray and CT scan images using deep learning techniques

Coronavirus disease (Covid-19) first came to light in 2019. It has spread across the world, which has led to a very strong ephemeral. As a result of its infection rate and severity, it has emerged as one of the significant ultimatums of the present generation. Since then, many academics and clinicians have worked on artificial intelligence-based models to predict Covid-19. It is easier to assess and cure the disease if it is detected and expected in an early stage. Moreover, using deep learning model techniques can decrease the time spent on traditional practices. As a result, our study stresses the use of deep learning patterns for interpreting medical images in Covid-19 patients. In addition, a comparison of numerous strategies for developing a deep learning model has been carried out. The goal is to create a deep learning architecture and compare and analyze performance metrics like accuracy, recall, precision, and F-score for deep learning methods used to detect SARS Covid-19 in an automated manner.

Development and Validation of an Artificial Intelligence-Based Model to Predict Gastroesophageal Reflux Disease After Sleeve Gastrectomy

PurposePrediction of the onset of de novo gastroesophageal reflux disease (GERD) after sleeve gastrectomy (SG) would be helpful in decision-making and selection of the optimal bariatric procedure for every patient. The present study aimed to develop an artificial intelligence (AI)-based model to predict the onset of GERD after SG to help clinicians and surgeons in decision-making.Materials and MethodsA prospectively maintained database of patients with severe obesity who underwent SG was used for the development of the AI model using all the available data points. The dataset was arbitrarily split into two parts: 70% for training and 30% for testing. Then ranking of the variables was performed in two steps. Different learning algorithms were used, and the best model that showed maximum performance was selected for the further steps of machine learning. A multitask AI platform was used to determine the cutoff points for the top numerical predictors of GERD.ResultsIn total, 441 patients (76.2% female) of a mean age of 43.7 ± 10 years were included. The ensemble model outperformed the other models. The model achieved an AUC of 0.93 (95%CI 0.88–0.99), sensitivity of 79.2% (95% CI 57.9–92.9%), and specificity of 86.1% (95%CI 70.5–95.3%). The top five ranked predictors were age, weight, preoperative GERD, size of orogastric tube, and distance of first stapler firing from the pylorus.ConclusionAn AI-based model for the prediction of GERD after SG was developed. The model had excellent accuracy, yet a moderate sensitivity and specificity. Further prospective multicenter trials are needed to externally validate the model developed.Graphical

Real-time Forecasting of Electrical Power System Loads using Moving Average-Extreme Learning Machine (MA-ELM) Algorithm

Load Forecasts are the primary factors which considered by electricity utility companies while planning power generation, power infrastructural development and load flows etc. Different forecasting techniques have been proposed from statistical to artificial intelligence-based models and the area of research is still growing. In our research work, considering the real time data of 33KV bus system which is having 34 buses and 54 lines. In this case, forecast the day ahead scheduling of various parameters such as load real power (Pload), voltage magnitude at each bus, apparent power flow between buses and total transmission losses for hourly basis and also forecasted the mentioned parameters for 5 days. The actual real time values are compared with forecasted values using two existing methods namely Extreme Learning Machine (ELM), moving average and proposed Moving Average–Extreme Learning Machine (MA-ELM) algorithm. In addition to this, forecasted the loads and losses for short term and long-term forecasting cases and verified through MATLAB programming.

Accuracy of an Artificial Intelligence-Based Model for Estimating Leftover Liquid Food in Hospitals: Validation Study.

BackgroundAn accurate evaluation of the nutritional status of malnourished hospitalized patients at a higher risk of complications, such as frailty or disability, is crucial. Visual methods of estimating food intake are popular for evaluating the nutritional status in clinical environments. However, from the perspective of accurate measurement, such methods are unreliable.ObjectiveThe accuracy of estimating leftover liquid food in hospitals using an artificial intelligence (AI)–based model was compared to that of visual estimation.MethodsThe accuracy of the AI-based model (AI estimation) was compared to that of the visual estimation method for thin rice gruel as staple food and fermented milk and peach juice as side dishes. A total of 576 images of liquid food (432 images of thin rice gruel, 72 of fermented milk, and 72 of peach juice) were used. The mean absolute error, root mean squared error, and coefficient of determination (R2) were used as metrics for determining the accuracy of the evaluation process. Welch t test and the confusion matrix were used to examine the difference of mean absolute error between AI and visual estimation.ResultsThe mean absolute errors obtained through the AI estimation approach were 0.63 for fermented milk, 0.25 for peach juice, and 0.85 for the total. These were significantly smaller than those obtained using the visual estimation approach, which were 1.40 (P<.001) for fermented milk, 0.90 (P<.001) for peach juice, and 1.03 (P=.009) for the total. By contrast, the mean absolute error for thin rice gruel obtained using the AI estimation method (0.99) did not differ significantly from that obtained using visual estimation (0.99). The confusion matrix for thin rice gruel showed variation in the distribution of errors, indicating that the errors in the AI estimation were biased toward the case of many leftovers. The mean squared error for all liquid foods tended to be smaller for the AI estimation than for the visual estimation. Additionally, the coefficient of determination (R2) for fermented milk and peach juice tended to be larger for the AI estimation than for the visual estimation, and the R2 value for the total was equal in terms of accuracy between the AI and visual estimations.ConclusionsThe AI estimation approach achieved a smaller mean absolute error and root mean squared error and a larger coefficient of determination (R2) than the visual estimation approach for the side dishes. Additionally, the AI estimation approach achieved a smaller mean absolute error and root mean squared error compared to the visual estimation method, and the coefficient of determination (R2) was similar to that of the visual estimation method for the total. AI estimation measures liquid food intake in hospitals more precisely than visual estimation, but its accuracy in estimating staple food leftovers requires improvement.

Artificial intelligence-based early detection of acute kidney injury after cardiac surgery.

This study aims to improve the early detection of cardiac surgery-associated acute kidney injury using artificial intelligence-based algorithms. Data from consecutive patients undergoing cardiac surgery between 2008 and 2018 in our institution served as the source for artificial intelligence-based modelling. Cardiac surgery-associated acute kidney injury was defined according to the Kidney Disease Improving Global Outcomes criteria. Different machine learning algorithms were trained and validated to detect cardiac surgery-associated acute kidney injury within 12 h after surgery. Demographic characteristics, comorbidities, preoperative cardiac status and intra- and postoperative variables including creatinine and haemoglobin values were retrieved for analysis. From 7507 patients analysed, 1699 patients (22.6%) developed cardiac surgery-associated acute kidney injury. The ultimate detection model, 'Detect-A(K)I', recognizes cardiac surgery-associated acute kidney injury within 12 h with an area under the curve of 88.0%, sensitivity of 78.0%, specificity of 78.9% and accuracy of 82.1%. The optimal parameter set includes serial changes of creatinine and haemoglobin, operative emergency, bleeding-associated variables, cardiac ischaemic time and cardiac function-associated variables, age, diuretics and active infection, chronic obstructive lung and peripheral vascular disease. The 'Detect-A(K)I' model successfully detects cardiac surgery-associated acute kidney injury within 12 h after surgery with the best discriminatory characteristics reported so far.

Leveraging artificial intelligence to predict ERG gene fusion status in prostate cancer.

BackgroundTMPRSS2-ERG gene rearrangement, the most common E26 transformation specific (ETS) gene fusion within prostate cancer, is known to contribute to the pathogenesis of this disease and carries diagnostic annotations for prostate cancer patients clinically. The ERG rearrangement status in prostatic adenocarcinoma currently cannot be reliably identified from histologic features on H&E-stained slides alone and hence requires ancillary studies such as immunohistochemistry (IHC), fluorescent in situ hybridization (FISH) or next generation sequencing (NGS) for identification.MethodsObjectiveWe accordingly sought to develop a deep learning-based algorithm to identify ERG rearrangement status in prostatic adenocarcinoma based on digitized slides of H&E morphology alone.DesignSetting, and Participants: Whole slide images from 392 in-house and TCGA cases were employed and annotated using QuPath. Image patches of 224 × 224 pixel were exported at 10 ×, 20 ×, and 40 × for input into a deep learning model based on MobileNetV2 convolutional neural network architecture pre-trained on ImageNet. A separate model was trained for each magnification. Training and test datasets consisted of 261 cases and 131 cases, respectively. The output of the model included a prediction of ERG-positive (ERG rearranged) or ERG-negative (ERG not rearranged) status for each input patch.Outcome measurements and statistical analysis: Various accuracy measurements including area under the curve (AUC) of the receiver operating characteristic (ROC) curves were used to evaluate the deep learning model.Results and LimitationsAll models showed similar ROC curves with AUC results ranging between 0.82 and 0.85. The sensitivity and specificity of these models were 75.0% and 83.1% (20 × model), respectively.ConclusionsA deep learning-based model can successfully predict ERG rearrangement status in the majority of prostatic adenocarcinomas utilizing only H&E-stained digital slides. Such an artificial intelligence-based model can eliminate the need for using extra tumor tissue to perform ancillary studies in order to assess for ERG gene rearrangement in prostatic adenocarcinoma.