Performance Of Prediction Models Research Articles

P-261 A comparative analysis of artificial intelligence models for blastulation in In vitro fertilization: single-frame vs. full-frame approach

Abstract Study question How does the performance of an blastulation prediction model using full-frame data over 3 days compare to a single-frame model using day 3 data? Summary answer The single-frame model, based on a day 3 embryo image, demonstrated comparable predictive performance to the full-frame model What is known already Pregnancy rates following blastocyst transfer have been reported to be higher than those after cleavage transfer. However, not all cleavage embryos reach the blastocyst stage, leading to the exploration of predictive models using embryo images and artificial intelligence. Clinical practice often involves the decision to culture embryos to the blastocyst stage on day 3 based on embryo quality assessments. Recently, there has been interest in using morphokinetics data for predicting blastulation and pregnancy outcomes. While the importance of morphokinetics in IVF outcomes is acknowledged, the specific time frames and features contributing significantly to accurate predictions remain unclear. Study design, size, duration This retrospective study analyzed a dataset of 15,615 timelapse images from 906 patients at a single IVF center. The Single-Frame Model used a single image collected at 66.7 hours post-ICSI, and the Full-Frame Model used 145 frames collected from fertilization to 66.7 hours post-ICSI. Blastulation criteria included reaching a blastocyst on day 5, deemed suitable for transfer or freezing by an experienced embryologist. Participants/materials, setting, methods We trained the data using ResNet for the one-frame model and 3D CNN for the full-frame model. The ResNet model used pre-trained weights, while the 3D CNN model was built from scratch with no pre-trained weights. The models were evaluated using the AUROC, accuracy, and F1 score metrics. The statistical significance of the AUC performance differences was assessed using DeLong’s test. Main results and the role of chance Our study revealed similar AUC performance between the Single-Frame and Full-Frame Models for day 3 embryos. The Single-Frame Model exhibited an AUROC of 0.76, accuracy of 0.71, and an F1 score of 0.75, while the Full-Frame Model demonstrated comparable performance with an AUROC of 0.75, accuracy of 0.70, and an F1 score of 0.72. DeLong’s test was employed to assess the statistical significance of the difference in AUROC, yielding a non-significant p-value of 0.4884. This study underscores that the Single-Frame Model provides similar predictive performance for blastulation compared to the Full-Frame Model. Limitations, reasons for caution Our analysis was limited to images taken at 66.7 hours post-ICSI to mirror clinical practice. Further investigations at different time points are necessary to determine the optimal time for blastulation prediction. Additional analysis of the Day 1 and Day 2 single frames may offer further insights. Wider implications of the findings Accurate prediction of blastulation on day 3 is pivotal for embryo survival and IVF success. Our study challenges the assumption that more data improves AI models, highlighting the importance of day 3 observations. Further research is warranted to identify key time points and features for better predictive accuracy. Trial registration number not applicable

P-144 Utilizing artificial intelligence to uncover Day 5 morphokinetic features associated with single blastocyst transfer implantation outcome

Abstract Study question Can AI aid in studing embryos’ continuous development in Day5 and identifying morphokinetic features (blastocyst growth and spontaneous collapse etc.), to improve implantation potential prediction? Summary answer The AI system utilized time-lapse images to measure newly identified morphokinetic features, establishing a significant correlation with clinical pregnancy outcomes. What is known already In Single Blastocyst Transfer (SBT), the conventional Gardner evaluation system may inadequately differentiate between embryos of comparable morphology scores in specific cases. To address this, integrating new morphokinetic features is essential for enhanced blastocyst evaluation. Spontaneous collapse (SC) during blastocyst expansion is a phenomenon that was revealed relatively recently following the clinical application of time-lapse monitoring in IVF laboratories. Previous studies link embryo size and SC to quality, yet these correlations require further validation. In the critical selection process utilizing time-lapse imaging (TLI) systems, AI technology is indispensable for developing automated and interpretable tools to ensure precise quantitative measurements. Study design, size, duration 255 patients underwent IVF with Day5 single blastocyst transfer in fresh cycles at the Center for Reproductive Medicine, Nanjing Drum Tower Hospital, from April 2021 to August 2022, and were analyzed in this retrospective study. Embryos were cultured in time-lapse incubators (CCM-iBIS, ASTEC). On Day 3 embryos were transferred to sequential media (G2, Vitrolife, Sweden) and cultured in the same incubator until Day5. In this cohort, there were 167 clinical pregnancies and 88 non-pregnant cases. Participants/materials, setting, methods An AI system was developed to calculate the development curves of areas in each part of the embryo (ZP,blastocel, ICM and TE) based on TLI images from 92 to 116 phi(±2). T-test and correlation analysis were used to select effective features. The dataset was divided into training and testing subset in an 80:20 ratio. The performance of different prediction models were compared by Delong test. Main results and the role of chance In this study, an AI system automatically generated blastocyst development curves, from which 35 morphokinetic features were extracted. These features were carefully defined to comprehensively capture the dynamics of embryo size development and spontaneous collapse (SC). Six parameters, including maximum size, max/min ratio of blastocyst area (with and without ZP region), speed of expansion after tSB, and number of minor SC, exhibited statistical correlation with clinical pregnancy outcomes. Using these parameters, an XGBoost model achieved an AUC of 0.71 for pregnancy prediction, surpassing (p = 0.0307, p < 0.05) the AUC of 0.48 from the model trained with manual Gardner evaluation. These findings underscore the potential of AI-generated morphokinetic features of D5 in enhancing blastocyst evaluation systems. Limitations, reasons for caution The limited sample size used in this retrospective experiment may lead to an underrepresentation of certain significant features. Wider implications of the findings In contrast to prior findings, most SC-based features in our study did not significantly correlate with clinical pregnancy outcome. Further research is needed to understand their causes and implications. Integrating AI with TLI offers promise for improving single blastocyst transfer outcomes, emphasizing quantifiability and interpretability in approach. Trial registration number NA

Predicting the 28-day prognosis of acute-on-chronic liver failure patients based on machine learning

BackgroundWe aimed to establish a prognostic predictive model based on machine learning (ML) methods to predict the 28-day mortality of acute-on-chronic liver failure (ACLF) patients, and to evaluate treatment effectiveness. MethodsACLF patients from six tertiary hospitals were included for analysis. Features for ML models’ development were selected by LASSO regression. Models’ performance was evaluated by area under the curve (AUC) and accuracy. Shapley additive explanation was used to explain the ML model. ResultsOf the 736 included patients, 587 were assigned to a training set and 149 to an external validation set. Features selected included age, hepatic encephalopathy, total bilirubin, PTA, and creatinine. The eXtreme Gradient Boosting (XGB) model outperformed other ML models in the prognostic prediction of ACLF patients, with the highest AUC and accuracy. Delong's test demonstrated that the XGB model outperformed Child-Pugh score, MELD score, CLIF-SOFA, CLIF-C OF, and CLIF-C ACLF. Sequential assessments at baseline, day 3, day 7, and day 14 improved the predictive performance of the XGB-ML model and can help clinicians evaluate the effectiveness of medical treatment. ConclusionsWe established an XGB-ML model to predict the 28-day mortality of ACLF patients as well as to evaluate the treatment effectiveness.

A predictive model combining connectomics and entropy biomarkers to discriminate long-term vagus nerve stimulation efficacy for pediatric patients with drug-resistant epilepsy.

To predict the vagus nerve stimulation (VNS) efficacy for pediatric drug-resistant epilepsy (DRE) patients, we aim to identify preimplantation biomarkers through clinical features and electroencephalogram (EEG) signals and thus establish a predictive model from a multi-modal feature set with high prediction accuracy. Sixty-five pediatric DRE patients implanted with VNS were included and followed up. We explored the topological network and entropy features of preimplantation EEG signals to identify the biomarkers for VNS efficacy. A Support Vector Machine (SVM) integrated these biomarkers to distinguish the efficacy groups. The proportion of VNS responders was 58.5% (38/65) at the last follow-up. In the analysis of parieto-occipital α band activity, higher synchronization level and nodal efficiency were found in responders. The central-frontal θ band activity showed significantly lower entropy in responders. The prediction model reached an accuracy of 81.5%, a precision of 80.1%, and an AUC (area under the receiver operating characteristic curve) of 0.838. Our results revealed that, compared to nonresponders, VNS responders had a more efficient α band brain network, especially in the parieto-occipital region, and less spectral complexity of θ brain activities in the central-frontal region. We established a predictive model integrating both preimplantation clinical and EEG features and exhibited great potential for discriminating the VNS responders. This study contributed to the understanding of the VNS mechanism and improved the performance of the current predictive model.

Optimization of Power Prediction of BP Network with Improved Pelican Algorithm

Abstract In view of the large fluctuation of photovoltaic output power affected by different weather, accurate prediction of photovoltaic output power is particularly important for the safe and stable operation of power system. Firstly, the pelican optimization algorithm ( POA ) is improved in the following three aspects : adding Circle chaotic map to make the population evenly distributed, introducing mutation factor to expand the search range of pelican when approaching prey, adding adaptive weight and firefly disturbance to avoid falling into local optimum in the water surface flight stage ; then, in order to improve the prediction accuracy of BP algorithm, the improved pelican algorithm ( IPOA ) is used to optimize the weights and thresholds of BP neural network, and the IPOA-BP photovoltaic power prediction model is built to improve the accuracy of power prediction. Finally, this paper tests the prediction performance of IPOA-BP, POA-BP and basic BP power prediction models in sunny, cloudy and rainy days through experiments. The experimental results demonstrate that the IPOA-BP prediction model outperforms both the POA-BP and traditional BP neural network models under various weather conditions.

A Prognostic Model Incorporating Relevant Peripheral Blood Inflammation Indicator to Predict Postherpetic Neuralgia in Patients with Acute Herpes Zoster.

To determine the risk of postherpetic neuralgia (PHN) in patients with acute herpes zoster (HZ), this study developed and validated a novel clinical prediction model by incorporating a relevant peripheral blood inflammation indicator. Between January 2019 and June 2023, 209 patients with acute HZ were categorized into the PHN group (n = 62) and the non-PHN group (n = 147). Univariate and multivariate logistic regression analyses were conducted to identify risk factors serving as independent predictors of PHN development. Subsequently, a nomogram prediction model was established, and the discriminative ability and calibration were evaluated using the receiver operating characteristic curve, calibration plots, and decision curve analysis (DCA). The nomogram model was internally verified through the bootstrap test method. According to univariate logistic regression analyses, five variables, namely age, hypertension, acute phase Numeric Rating Scale (NRS-11) score, platelet-to-lymphocyte ratio (PLR), and systemic immune inflammation index, were significantly associated with PHN development. Multifactorial analysis further unveiled that age (odds ratio (OR) [95% confidence interval (CI)]: 2.309 [1.163-4.660]), acute phase NRS-11 score (OR [95% CI]: 2.837 [1.294-6.275]), and PLR (OR [95% CI]: 1.015 [1.010-1.022]) were independent risk factors for PHN. These three predictors were integrated to establish the prediction model and construct the nomogram. The area under the receiver operating characteristic curve (AUC) for predicting the PHN risk was 0.787, and the AUC of internal validation determined using the bootstrap method was 0.776. The DCA and calibration curve also indicated that the predictive performance of the nomogram model was commendable. In this study, a risk prediction model was developed and validated to accurately forecast the probability of PHN after HZ, thereby demonstrating favorable discrimination, calibration, and clinical applicability.

Feasibility of differentiating gait in Parkinson's disease and spinocerebellar degeneration using a pose estimation algorithm in two-dimensional video

BackgroundAlthough pose estimation algorithms have been used to analyze videos of patients with Parkinson's disease (PD) to assess symptoms, their feasibility for differentiating PD from other neurological disorders that cause gait disturbances has not been evaluated yet. We aimed to determine whether it was possible to differentiate between PD and spinocerebellar degeneration (SCD) by analyzing video recordings of patient gait using a pose estimation algorithm. MethodsWe videotaped 82 patients with PD and 61 patients with SCD performing the timed up-and-go test. A pose estimation algorithm was used to extract the coordinates of 25 key points of the participants from these videos. A transformer-based deep neural network (DNN) model was trained to predict PD or SCD using the extracted coordinate data. We employed a leave-one-participant-out cross-validation method to evaluate the predictive performance of the trained model using accuracy, sensitivity, and specificity. As there were significant differences in age, weight, and body mass index between the PD and SCD groups, propensity score matching was used to perform the same experiment in a population that did not differ in these clinical characteristics. ResultsThe accuracy, sensitivity, and specificity of the trained model were 0.86, 0.94, and 0.75 for all participants and 0.83, 0.88, and 0.78 for the participants extracted by propensity score matching. ConclusionThe differentiation of PD and SCD using key point coordinates extracted from gait videos and the DNN model was feasible and could be used as a collaborative tool in clinical practice and telemedicine.

Growing degree‐days do not explain moth species' distributions at broad scales

AbstractGrowing degree‐days (GDD), an estimate of an organism's growing season length, has been shown to be an important predictor of Lepidopteran species' distributions and could be influencing Lepidopteran range shifts to climate change. Yet, one understudied simplification in this literature is that the same thermal threshold is used in the calculations of GDD for all species instead of a species‐specific threshold. By characterizing the phenological process influenced by climate, a species‐specific estimate of GDD should improve the accuracy of species distribution models (SDMs). To test this hypothesis, we used published, experimentally estimated thermal thresholds and modeled the current geographic distribution of 30 moth species native to North America. We found that the predictive performance of models based on a species‐specific estimate of GDD was indistinguishable from models based on a standard estimate of GDD. This is likely because GDD was not an important predictor of these species' distributions. Our findings suggest that experimentally estimated thermal thresholds may not always scale up to be predictive at broad scales and that more work is needed to leverage the data from lab experiments into SDMs to accurately predict species' range shifts in response to climate change.

Evaluation of the performance of both machine learning models using PET and CT radiomics for predicting recurrence following lung stereotactic body radiation therapy: A single-institutional study.

Predicting recurrence following stereotactic body radiotherapy (SBRT) for non-small cell lung cancer provides important information for the feasibility of the individualized radiotherapy and allows to select the appropriate treatment strategy based on the risk of recurrence. In this study, we evaluated the performance of both machine learning models using positron emission tomography (PET) and computed tomography (CT) radiomic features for predicting recurrence after SBRT. Planning CT and PET images of 82 non-small cell lung cancer patients who performed SBRT at our hospital were used. First, tumors were delineated on each CT and PET of each patient, and 111 unique radiomic features were extracted, respectively. Next, the 10 features were selected using three different feature selection algorithms, respectively. Recurrence prediction models based on the selected features and four different machine learning algorithms were developed, respectively. Finally, we compared the predictive performance of each model for each recurrence pattern using the mean area under the curve (AUC) calculated following the 0.632+ bootstrap method. The highest performance for local recurrence, regional lymph node metastasis, and distant metastasis were observed in models using Support vector machine with PET features (mean AUC=0.646), Naive Bayes with PET features (mean AUC=0.611), and Support vector machine with CT features (mean AUC=0.645), respectively. We comprehensively evaluated the performance of prediction model developed for recurrence following SBRT. The model in this study would provide information to predict the recurrence pattern and assist in making treatment strategies.

Model Prediksi Mortalitas Pembedahan Pasien Usia Lanjut yang Menjalani Pembedahan Elektif di Rumah Sakit Cipto Mangunkusumo

Introduction. Elderly patients are at high risk of mortality after elective surgery. The aim of this study was to obtain and evaluate the performance of a mortality prediction model for elderly patients undergoing elective surgery. Methods. The design of this study was a retrospective cohort using medical record data from 747 elderly subjects who underwent elective surgery in the period January 2015–December 2017 at Cipto Mangunkusumo Hospital (RSCM), Jakarta. This study used multivariate analysis with logistic regression to determine significant predictors that were included in the prediction model. The performance of the prediction model was assessed using the Hosmer-Lameshow test, and its discrimination ability was determined by calculating the Area Under the Curve (AUC). Results. Of the total 747 elderly subjects, the elective surgery mortality rate was 14.5%. The predictor variables were functional status [ADL 9-11, OR 1.808 (95% CI 0.848-3.854); ADL 0-8, OR 3.382 (95% CI 1.724-6.634)], comorbidities [CCI 3-4, OR 12.206 (95% CI 5.317-28.018); CCI >5, OR 15.820 (95% CI 6.701-37.347)], albumin level <3 g/dL [OR 3.777 (95% CI 2.105-6.779)], type of surgery [grade II, OR 3.827 (95% CI 1.849- 7.923); grade III, OR 6.560 (95% CI 3.378-12.739)], and ASA status with an ASA score > 3 [OR 5.106 (95% CI 1.841-14.159)] were further included in the components of the surgical mortality predictor scoring system. The mortality prediction model was categorized into low risk (score < 7; probability of mortality 2.33%), medium risk (score 7-10; probability of mortality 25.22%), and high risk (score > 10; probability of mortality 74.67%). The prediction model showed good discrimination [AUC score 0.900 (95% CI: 0.873-0.927)] and good calibration (p=0.718 on Hosmer-Lameshow test). Conclusion. The prediction model of mortality among elderly patients undergoing elective surgery, incorporating factors like functional capacity, comorbidities, preoperative serum albumin concentrations, surgical procedure type, and ASA classification, showed good performance.

Risk factor analysis and nomogram prediction model construction of postoperative complications of thoracoscopic non-small cell lung cancer.

A series of complications will inevitably occur after thoracoscopic pulmonary resection. How to avoid or reduce postoperative complications is an important research area in the perioperative treatment of thoracic surgery. This study analyzed the risk factors for thoracoscopic postoperative complications of non-small cell lung cancer (NSCLC) and established a nomogram prediction model in order to provide help for clinical decision-making. Patients with NSCLC who underwent thoracoscopic surgery from January 2017 to December 2021 were selected as study subjects. The relationship between patient characteristics, surgical factors, and postoperative complications was collected and analyzed. Based on the results of the statistical regression analysis, a nomogram model was constructed, and the predictive performance of the nomogram model was evaluated. A total of 872 patients who met the study criteria were included in the study. A total of 171 patients had complications after thoracoscopic surgery, accounting for 19.6% of the study population. Logistic regression analysis showed that thoracic adhesion, history of respiratory disease, and lymphocyte-monocyte ratio (LMR) were independent risk factors for complications after thoracoscopic surgery (P<0.05). Variables with P<0.1 in logistic regression analysis were included in the nomogram model. The verification results showed that the area under curve (AUC) of the model was 0.734 [95% confidence interval (CI): 0.693-0.775], and the calibration curve showed that the model had good differentiation. The decision curve analysis (DCA) curve showed that this model has good clinical application value. In subgroup analysis of complications, gender, history of respiratory disease, body mass index (BMI), type of surgical procedure, thoracic adhesion, and Time of operation were identified as significant risk factors for prolonged air leak (PAL) after surgery. Tumor location and forced expiratory volume in the first second (FEV1) were identified as important risk factors for postoperative pulmonary infection. N stage and thoracic adhesion were identified as significant risk factors for postoperative pleural effusion. The AUC for PAL was 0.823 (95% CI: 0.768-0.879). The AUC of postoperative pulmonary infection was 0.714 (95% CI: 0.627-0.801). The AUC of postoperative pleural effusion was 0.757 (95% CI: 0.650-0.864). The calibration curve and DCA curve indicated that the model had good predictive performance and clinical application value. This study analyzed the risk factors affecting the postoperative complications of NSCLC through thoracoscopic surgery, and the nomogram model built based on the influencing factors has certain significance for the identification and reduction of postoperative complications.

Analysis of Water Quality Prediction in the Yangtze River Delta under the River Chief System

Water quality prediction is essential for effective water resource management and pollution prevention. In China, research on predictive analytics for various water bodies has not kept pace with environmental needs. This study addresses this gap by conducting a comprehensive analysis and modeling of water quality monitoring data from multiple distributed water bodies specifically within the Yangtze River Delta. Using a novel approach, this paper introduces a distributed water quality prediction system enhanced by a CNN-LSTM joint model. This model synergistically combines convolutional neural networks (CNN) and long short-term memory (LSTM) networks to robustly extract and utilize spatiotemporal data, thereby significantly improving the accuracy of predicting dynamic water quality trends. Notably, the excellent predictive performance of the joint model enables its prediction results to achieve RMSE and MAPE as low as 1.08% and 6.8%, respectively. Empirical results from this study highlight the system’s superior predictive performance. Based on these findings, this paper offers targeted recommendations for water quality monitoring, treatment, and management strategies tailored to the specific needs of the Yangtze River Delta. These contributions are poised to aid policymakers and environmental managers in making more informed decisions.

Outcome Prediction Using Multi-Modal Information: Integrating Large Language Model-Extracted Clinical Information and Image Analysis.

Survival prediction post-cystectomy is essential for the follow-up care of bladder cancer patients. This study aimed to evaluate artificial intelligence (AI)-large language models (LLMs) for extracting clinical information and improving image analysis, with an initial application involving predicting five-year survival rates of patients after radical cystectomy for bladder cancer. Data were retrospectively collected from medical records and CT urograms (CTUs) of bladder cancer patients between 2001 and 2020. Of 781 patients, 163 underwent chemotherapy, had pre- and post-chemotherapy CTUs, underwent radical cystectomy, and had an available post-surgery five-year survival follow-up. Five AI-LLMs (Dolly-v2, Vicuna-13b, Llama-2.0-13b, GPT-3.5, and GPT-4.0) were used to extract clinical descriptors from each patient's medical records. As a reference standard, clinical descriptors were also extracted manually. Radiomics and deep learning descriptors were extracted from CTU images. The developed multi-modal predictive model, CRD, was based on the clinical (C), radiomics (R), and deep learning (D) descriptors. The LLM retrieval accuracy was assessed. The performances of the survival predictive models were evaluated using AUC and Kaplan-Meier analysis. For the 163 patients (mean age 64 ± 9 years; M:F 131:32), the LLMs achieved extraction accuracies of 74%~87% (Dolly), 76%~83% (Vicuna), 82%~93% (Llama), 85%~91% (GPT-3.5), and 94%~97% (GPT-4.0). For a test dataset of 64 patients, the CRD model achieved AUCs of 0.89 ± 0.04 (manually extracted information), 0.87 ± 0.05 (Dolly), 0.83 ± 0.06~0.84 ± 0.05 (Vicuna), 0.81 ± 0.06~0.86 ± 0.05 (Llama), 0.85 ± 0.05~0.88 ± 0.05 (GPT-3.5), and 0.87 ± 0.05~0.88 ± 0.05 (GPT-4.0). This study demonstrates the use of LLM model-extracted clinical information, in conjunction with imaging analysis, to improve the prediction of clinical outcomes, with bladder cancer as an initial example.

Prediction of prognosis in patients with systemic sclerosis based on a machine-learning model.

The clinical manifestations of systemic sclerosis (SSc) are highly variable, resulting in varied outcomes and complications. Diverse fibrosis of the skin and internal organs, vasculopathy, and dysregulated immune system lead to poor and varied prognoses in patients with SSc subtypes. Therefore, this study aimed to develop a personalized tool for predicting the prognosis of patients with SSc. A cohort of 517 patients with SSc were recruited between January 2009 and November 2021 at Xijing Hospital in China, and 266 patients completed the follow-up and performed in the survival analysis. Risk factors for death were identified using Cox survival analysis and random survival forest-based machine-learning methods separately. The consistency index, area under the curve (AUC), and integrated Brier scores were used to compare the predictive performance of the different prognostic models. The results of Cox-based multivariate regression analysis suggested that pulmonary arterial hypertension, digital ulcer, and Modified Rodnan Skin Score (mRSS) were independent risk factors for poor prognosis in patients with SSc and significant risk factors in random survival forest (RSF) surveys. A nomogram was plotted to evaluate the prognostic risk to facilitate clinical assessment; the RSF model had better predictive performance than the Cox model, with 3- and 5-year AUCs of 0.74 and 0.78, respectively. Machine-learning models can help us better understand the prognosis of patients with SSc and comprehensively evaluate the clinical characteristics of each individual. The early identification of the characteristics of high-risk patients can improve the prognosis of those with SSc. Key Points • Regarding predictive performance, the random survival forest model was more effective than the Cox model and had unique advantages in analyzing nonlinear effects and variable importance. • Machine learning using the simple clinical features of patients with systemic sclerosis (SSc) to predict mortality can guide attending physicians, and the early identification of high-risk patients with SSc and referral to experts will assist rheumatologists in monitoring and management planning.

Feasibility study and optimization of limestone calcined clay composites for compressive strength using multi-layered explainable artificial intelligence models

Interpretable white box machine learning models to predict the compressive strength of Limestone Calcined Clay (LC3) cementitious composite have been developed and analyzed. For this, a large dataset consisting of more than 4000 numerical instances has been collected from the literature, which has been tuned and normalized. Mainly based on the proportion of the chemical composition of its main ingredients – Limestone (0–31 by %mass of LC3), Calcined Clay (0–70 by %mass of LC3), and Cement (30–100 by %mass of LC3) with a water-to-binder ratio ranging from 0.40 to 0.60, predictive models have been developed for the 28 days compressive strength (fc) of LC3. The performance of various predictive models is then compared across the dataset, and suggestions are made for their reliability and explainability. The models are further interpretated using multi-layered explanations invoking model-agnostic explainable artificial intelligence (XAI) technique. Optimization and interaction plots have also been generated and shown as three-dimensional (3D) visualization. Percent of calcined clay, cement and the water to binder ratio were found to be the most influencing parameters for the compressive strength of LC3. Results of this study indicate that the optimized blends of LC3 can be expressed with up to 50 % replacement of OPC without any significant compromise in fc, contributing to the development of sustainable cementitious composites. The major advantage of using this framework lies in the model agnostic approach, so different predictive models can be compared in a same manner and thus, this approach is applicable to interpretable forecasting models for binary, ternary, quaternary sustainable cementitious composites.

Self-supervised class-balanced active learning with uncertainty-mastery fusion

Deep active learning (DeepAL) offers a viable solution for enhancing the predictive performance of models with limited annotation costs. Popular DeepAL pipelines fail to (1) adequately and simultaneously draw on unlabeled and labeled samples to learn accurate data representations, and (2) progressively balance the distribution of selected samples across classes during the query. In this paper, we introduce a self-supervised DeepAL (SSAL) framework as a solution to the aforementioned challenges. First, SSAL simultaneously utilizes both labeled data and unlabeled data to learn high-quality representations. The method learns the image rotation degree for unlabeled data in a self-supervised manner, which can be jointly processed with the supervised learning. Second, the sample information is interpreted by fusing the discriminant uncertainty and the presumed mastery ability of each sample. This approach prevents the selection of many nonrepresentative boundary samples based solely on uncertainty and avoids the selection of aggregated samples based solely on clustering. Third, we propose a new sample selection strategy named class rebalanced group sampling. This strategy rebalances classes by the optimal size of query samples for each class, which is induced from the approximation between the population and the model distribution. Experiments were conducted on five public image datasets, namely CIFAR-10/100, FashionMNIST, SVHN, and TinyImageNet, and the results demonstrated the effectiveness of our approach. The code is available at https://github.com/FanSmale/SSAL.

Investigating the effects of cytokine biomarkers on HIV incidence: a case study for individuals randomized to pre-exposure prophylaxis vs. control.

Understanding and identifying the immunological markers and clinical information linked with HIV acquisition is crucial for effectively implementing Pre-Exposure Prophylaxis (PrEP) to prevent HIV acquisition. Prior analysis on HIV incidence outcomes have predominantly employed proportional hazards (PH) models, adjusting solely for baseline covariates. Therefore, models that integrate cytokine biomarkers, particularly as time-varying covariates, are sorely needed. We built a simple model using the Cox PH to investigate the impact of specific cytokine profiles in predicting the overall HIV incidence. Further, Kaplan-Meier curves were used to compare HIV incidence rates between the treatment and placebo groups while assessing the overall treatment effectiveness. Utilizing stepwise regression, we developed a series of Cox PH models to analyze 48 longitudinally measured cytokine profiles. We considered three kinds of effects in the cytokine profile measurements: average, difference, and time-dependent covariate. These effects were combined with baseline covariates to explore their influence on predictors of HIV incidence. Comparing the predictive performance of the Cox PH models developed using the AIC metric, model 4 (Cox PH model with time-dependent cytokine) outperformed the others. The results indicated that the cytokines, interleukin (IL-2, IL-3, IL-5, IL-10, IL-16, IL-12P70, and IL-17 alpha), stem cell factor (SCF), beta nerve growth factor (B-NGF), tumor necrosis factor alpha (TNF-A), interferon (IFN) alpha-2, serum stem cell growth factor (SCG)-beta, platelet-derived growth factor (PDGF)-BB, granulocyte macrophage colony-stimulating factor (GM-CSF), tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), and cutaneous T-cell-attracting chemokine (CTACK) were significantly associated with HIV incidence. Baseline predictors significantly associated with HIV incidence when considering cytokine effects included: age of oldest sex partner, age at enrollment, salary, years with a stable partner, sex partner having any other sex partner, husband's income, other income source, age at debut, years lived in Durban, and sex in the last 30 days. Overall, the inclusion of cytokine effects enhanced the predictive performance of the models, and the PrEP group exhibited reduced HIV incidences compared to the placebo group.

Predictability of buprenorphine-naloxone treatment retention: A multi-site analysis combining electronic health records and machine learning.

Opioid use disorder (OUD) and opioid dependence lead to significant morbidity and mortality, yet treatment retention, crucial for the effectiveness of medications like buprenorphine-naloxone, remains unpredictable. Our objective was to determine the predictability of 6-month retention in buprenorphine-naloxone treatment using electronic health record (EHR) data from diverse clinical settings and to identify key predictors. This retrospective observational study developed and validated machine learning-based clinical risk prediction models using EHR data. Data were sourced from Stanford University's healthcare system and Holmusk's NeuroBlu database, reflecting a wide range of healthcare settings. The study analyzed 1800 Stanford and 7957 NeuroBlu treatment encounters from 2008 to 2023 and from 2003 to 2023, respectively. Predict continuous prescription of buprenorphine-naloxone for at least 6months, without a gap of more than 30 days. The performance of machine learning prediction models was assessed by area under receiver operating characteristic (ROC-AUC) analysis as well as precision, recall and calibration. To further validate our approach's clinical applicability, we conducted two secondary analyses: a time-to-event analysis on a single site to estimate the duration of buprenorphine-naloxone treatment continuity evaluated by the C-index and a comparative evaluation against predictions made by three human clinical experts. Attrition rates at 6months were 58% (NeuroBlu) and 61% (Stanford). Prediction models trained and internally validated on NeuroBlu data achieved ROC-AUCs up to 75.8 (95% confidence interval [CI] = 73.6-78.0). Addiction medicine specialists' predictions show a ROC-AUC of 67.8 (95% CI = 50.4-85.2). Time-to-event analysis on Stanford data indicated a median treatment retention time of 65 days, with random survival forest model achieving an average C-index of 65.9. The top predictor of treatment retention identified included the diagnosis of opioid dependence. US patients with opioid use disorder or opioid dependence treated with buprenorphine-naloxone prescriptions appear to have a high (∼60%) treatment attrition by 6months. Machine learning models trained on diverse electronic health record datasets appear to be able to predict treatment continuity with accuracy comparable to that of clinical experts.

Classification of Cardiovascular Arrhythmia Using Deep Learning Techniques: A Review

Deep Learning (DL), an offshoot of Machine Learning (ML) has emerged as a powerful and feasible solution for medical image analysis due to advancements in robust computer software and hardware technologies. It plays a key role in Cardiovascular disease (CVD) diagnosis by detecting anomalies in Electrocardiogram (ECG) signals. Cardiac arrhythmia, which refers to irregular heartbeat, may signal an early symptom of CVD and can lead to fatal outcomes if ignored. Accurate detection of arrhythmia is very challenging even for experts to distinguish between acute and chronic conditions in ECG readings. This triggered the focus of researchers to explore the application of Artificial Intelligence for ECG classification. Traditional machine learning methods use handcrafted features that require domain knowledge. The new era in DL makes the automatic detection of Cardiovascular Disease (CVD) possible. In this paper, an exhaustive review of DL-based techniques for ECG classification has been presented. Research findings in this survey indicate the challenges and issues with arrhythmia detection, such as single lead and multiple lead ECG signals, choice of the size of the training data set, and the number of arrhythmia classes, etc. The study also signifies that there is great scope for improving the performance of arrhythmia prediction models by employing hybrid ensemble learning, time series analysis using Recurrent Neural Network architectures and identification of unexplored classes of arrhythmia.

Deep-learning-based prediction of glaucoma conversion in normotensive glaucoma suspects

Background/aimsTo assess the performance of deep-learning (DL) models for prediction of conversion to normal-tension glaucoma (NTG) in normotensive glaucoma suspect (GS) patients.MethodsDatasets of 12 458 GS eyes were reviewed. Two...