Dynamic-DeepHit: A Deep Learning Approach for Dynamic Survival Analysis With Competing Risks Based on Longitudinal Data

Deep neural network base competing risk in predicting heart failure patient's survival

Objectives

Heart failure (HF) is a complicated disease with several competing risks of interest, such as HF death and other causes. This study compares a Deep Neural Network Competing Risks (DNNCR) and Random Survival Forest (RSF) model to evaluate the predictive performance of time-to-event outcomes in HF patients with competing risks.

Methods

This study represents the retrospective analysis of 435 HF patients admitted to RCMRC, Tehran, Iran, between March and August 2018. After a five-year follow-up in 2023, predictions were analyzed based on Cause of death. This study employed RSF and DNN methods to account for competing risks in survival analysis. Then, model fitness was applied using C-index and IBS.

Results

The C-index of the results shows that DNNCR is superior to RSF in predicting survival outcomes for HF and other causes of death. Precisely, the C-index was 0.65 (0.04) for HF and 0.63 (0.02) for other causes of death in the DNNCR model, while in RSF, the C-index was 0.65 (0.04) for HF and 0.61 (0.03) for Other Causes. Additionally, calibration results showed via the IBS the finest performance of the DNNCR model at 0.16 for HF, followed by other causes with an IBS of 0.18.

Conclusions

The study shows that the DNNCR model outperforms RSF in predicting survival outcomes for HF patients, particularly in the presence of competing risks. The improved accuracy enables physicians to identify high-risk individuals and tailor treatment plans accordingly. Future research could utilize more diverse datasets to enhance DNNCR performance and integrate these models into clinical tools.

Graphical Abstract

Deep Learning for Predicting Acute Exacerbation and Mortality of Interstitial Lung Disease

Rationale: Some patients with interstitial lung disease (ILD) have a high mortality rate or experience acute exacerbation of ILD (AE-ILD) that results in increased mortality. Early identification of these high-risk patients and accurate prediction of the onset of these important events is important to determine treatment strategies. Although various factors that affect disease behavior among patients with ILD hinder the accurate prediction of these events, the use of longitudinal information may enable better prediction. Objectives: To develop a deep learning (DL) model to predict composite outcomes defined as the first occurrence of AE-ILD and mortality using longitudinal data. Methods: Longitudinal clinical and environmental data were retrospectively collected from consecutive patients with ILD at two specialty centers between January 2008 and December 2015. A DL model was developed to predict composite outcomes using longitudinal data from 80% of patients from the first center, which was then validated using data from the remaining 20% of patients and the second center. The developed model was compared with the univariate Cox proportional hazard (CPH) model using the ILD gender-age-physiology (ILD-GAP) score and multivariate CPH model at the time of ILD diagnosis. Results: AE-ILD was reported in 218 patients among the 1,175 patients enrolled, whereas 380 died without developing AE-ILD. The truncated concordance index (C-index) values of univariate/multivariate CPH models for composite outcomes within 12, 24, and 36 months after prediction were 0.789/0.843, 0.788/0.853, and 0.787/0.853 in internal validation, and 0.650/0.718, 0.652/0.756, and 0.640/0.756 in external validation, respectively. At 12 months after ILD diagnosis, the DL model outperformed the univariate CPH model and multivariate CPH model for composite outcomes within 12 months, with concordance index values of 0.842, 0.840, and 0.839 in internal validation, and 0.803, 0.744, and 0.746 in external validation, respectively. Neutrophils, C-reactive protein, ILD-GAP score, and exposure to suspended particulate matter were strongly associated with the composite outcomes. Conclusions: The DL model can accurately predict the incidence of AE-ILD or mortality using longitudinal data.

Predicting radiation-induced hypothyroidism in nasopharyngeal carcinoma patients using a deep learning model

Background

Radiation-induced hypothyroidism (RIHT) is a common complication in nasopharyngeal carcinoma patients. Predicting its onset is crucial for effective management and early intervention. This study aims to develop a model based on deep learning survival analysis to predict RIHT in nasopharyngeal carcinoma patients.

Methods

This retrospective study included 535 nasopharyngeal carcinoma patients between January 2015 and October 2020. Cox regression, LASSO-Cox analyses and Spearman correlation test were employed to identify significant predictors. Two deep learning and two machine learning algorithms were trained, tuned, and compared against traditional Cox and NTCP models by C-index, Brier score, and decision curve analysis.

Results

The study observed a 41.7 % incidence of RIHT, with a median time to onset of 15 months. AJCC N stage, thyroid volume and specific dose-volume parameters were identified as potential predictors. DeepSurv model outperformed traditional ones (C-index: DeepSurv 0.75, traditional models ≤ 0.63). While other models were competitive at early post-treatment intervals, deep learning models demonstrated superior performance over time. Calibration and decision curve analysis corroborated the enhanced predictive capability of DeepSurv. Feature importance analysis highlighted thyroid V30 and V50 as the most significant predictors.

Conclusions

DeepSurv demonstrated superior predictive performance for RIHT in nasopharyngeal carcinoma patients compared to traditional models. Deep learning-based predictions offer high accuracy, which may enable personalized patient management and have great potentials in mitigating the risk of RIHT. These findings suggested that incorporating such model into clinical practice could be beneficial for the management of RITH.

Hospital-wide, dynamic, individualized prediction of central line-associated bloodstream infections-development and temporal evaluation of six prediction models

BACKGROUND

Central line-associated bloodstream infections (CLABSI) are preventable hospital-acquired infections. Predicting CLABSI helps improve early intervention strategies and enhance patient safety.

AIM

To develop and temporally evaluate dynamic prediction models for continuous CLABSI risk monitoring.

METHODS

Data from hospitalized patients with central catheter(s) admitted to University Hospitals Leuven between 2014 and 2017 were used to develop five dynamic models (a landmark cause-specific model, two random forest models, and two XGBoost models) to predict 7-day CLABSI risk, accounting for competing events (death, discharge, and catheter removal). The models' predictions were then combined using a superlearner model. All models were temporally evaluated on data from the same hospital from 2018 to 2020 using performance metrics for discrimination, calibration, and clinical utility.

FINDINGS

Among 61629 catheter episodes in the training set, 1930 (3.1%) resulted in CLABSI, while in the test set of 44544 catheter episodes, 1059 (2.4%) experienced CLABSI. Among individual models, one XGBoost model achieved the highest AUROC of 0.748. Calibration was good for predicted risks up to 5%, while the cause-specific and XGBoost models overestimated higher predicted risks. The superlearner displayed a modest improvement in discrimination (AUROC up to 0.751) and better calibration than the cause-specific and XGBoost models, but worse than the random forest models. The models showed clinical utility to support standard care interventions (at risk thresholds between 0.5-4%), but not to support advanced interventions (at thresholds 15-25%).

CONCLUSION

Hospital-wide CLABSI prediction models offer clinical utility based on medium-risk thresholds. Clinical utility at present may be limited as the model performance deteriorated over time.

EHR-ML: A data-driven framework for designing machine learning applications with electronic health records

OBJECTIVE

The healthcare landscape is experiencing a transformation with the integration of Artificial Intelligence (AI) into traditional analytic workflows. However, its integration faces challenges resulting in a crisis of generalisability. Key obstacles include; 1) Insufficient consideration of local contextual factors, such as institution-specific data formats, practices, and protocols, which can lead to variability in clinical practices across different institutions. 2) ad-hoc data preparation and design of machine learning strategies. 3) manual subjective adjustment of design parameters resulting in sub-optimal performance. 4) EHR specific challenges regarding data biases affecting the model outcomes and unique intermittent temporal nature of the data necessitating specialised handling 5) lack of cross-institutional data validations.

METHODS

To address these challenges, EHR-ML, provides an easy to use structured framework for designing optimum machine learning applications in a data-driven manner. The framework supports ingestion of local institutional electronic health records (EHRs) and process standardisation. The study design and parameter optimisation is done in a fully data-driven evidence-based approach. It seamlessly integrating with existing quality control tools. To handle the unique characteristics of the EHR data, it offers customisable ensemble models. It enables the acquisition of EHR data from diverse systems and harmonise them into common formats following international standards.

RESULTS

The effectiveness of the EHR-ML is demonstrated through a series of case studies. These studies highlight its capability to develop high-performance models in a fully automated manner, consistently surpassing the performance of traditional methodologies. Furthermore, they exhibited strong generalisability across diverse healthcare settings.

DISCUSSION AND CONCLUSION

EHR-ML enhances the clinical relevance and accuracy of predictive models by incorporating local context into machine learning applications. Additionally, by providing an user-friendly fully-automated framework, it facilitates rapid hypothesis testing aimed to generate localised biomedical knowledge.

Competing and Noncompeting Risk Models for Predicting Kidney Allograft Failure

Background

Prognostic models are becoming increasingly relevant in clinical trials as potential surrogate end points and for patient management as clinical decision support tools. However, the effect of competing risks on model performance remains poorly investigated. We aimed to carefully assess the performance of competing risk and noncompeting risk models in the context of kidney transplantation, where allograft failure and death with a functioning graft are two competing outcomes.

Methods

We included 11,046 kidney transplant recipients enrolled in ten countries. We developed prediction models for long-term kidney graft failure prediction, without accounting (i.e., censoring) and accounting for the competing risk of death with a functioning graft, using Cox, Fine–Gray, and cause-specific Cox regression models. To this aim, we followed a detailed and transparent analytical framework for competing and noncompeting risk modeling and carefully assessed the models' development, stability, discrimination, calibration, overall fit, clinical utility, and generalizability in external validation cohorts and subpopulations. More than 15 metrics were used to provide an exhaustive assessment of model performance.

Results

Among 11,046 recipients in the derivation and validation cohorts, 1497 (14%) lost their graft and 1003 (9%) died with a functioning graft after a median follow-up postrisk evaluation of 4.7 years (interquartile range, 2.7–7.0). The cumulative incidence of graft loss was similarly estimated by Kaplan–Meier and Aalen–Johansen methods (17% versus 16% in the derivation cohort). Cox and competing risk models showed similar and stable risk estimates for predicting long-term graft failure (average mean absolute prediction error of 0.0140, 0.0138, and 0.0135 for Cox, Fine–Gray, and cause-specific Cox models, respectively). Discrimination and overall fit were comparable in the validation cohorts, with concordance index ranging from 0.76 to 0.87. Across various subpopulations and clinical scenarios, the models performed well and similarly, although in some high-risk groups (such as donors older than 65 years), the findings suggest a trend toward moderately improved calibration when using a competing risk approach.

Conclusions

Competing and noncompeting risk models performed similarly in predicting long-term kidney graft failure.

Generative AI Models in Time-Varying Biomedical Data: Scoping Review

BACKGROUND

Trajectory modeling is a long-standing challenge in the application of computational methods to health care. In the age of big data, traditional statistical and machine learning methods do not achieve satisfactory results as they often fail to capture the complex underlying distributions of multimodal health data and long-term dependencies throughout medical histories. Recent advances in generative artificial intelligence (AI) have provided powerful tools to represent complex distributions and patterns with minimal underlying assumptions, with major impact in fields such as finance and environmental sciences, prompting researchers to apply these methods for disease modeling in health care.

OBJECTIVE

While AI methods have proven powerful, their application in clinical practice remains limited due to their highly complex nature. The proliferation of AI algorithms also poses a significant challenge for nondevelopers to track and incorporate these advances into clinical research and application. In this paper, we introduce basic concepts in generative AI and discuss current algorithms and how they can be applied to health care for practitioners with little background in computer science.

METHODS

We surveyed peer-reviewed papers on generative AI models with specific applications to time-series health data. Our search included single- and multimodal generative AI models that operated over structured and unstructured data, physiological waveforms, medical imaging, and multi-omics data. We introduce current generative AI methods, review their applications, and discuss their limitations and future directions in each data modality.

RESULTS

We followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines and reviewed 155 articles on generative AI applications to time-series health care data across modalities. Furthermore, we offer a systematic framework for clinicians to easily identify suitable AI methods for their data and task at hand.

CONCLUSIONS

We reviewed and critiqued existing applications of generative AI to time-series health data with the aim of bridging the gap between computational methods and clinical application. We also identified the shortcomings of existing approaches and highlighted recent advances in generative AI that represent promising directions for health care modeling.

Survival and data-driven phenotypes in head and neck cancer

Head and neck cancer (HNC) is the seventh most common cancer worldwide, with a 5-year overall survival of 50-60%. Despite known survival factors, unexpected deaths underscore the need to refine pre-treatment survival predictions. This study aimed to identify novel data-driven HNC patient phenotypes, associated features and predict overall survival using deep survival clustering model VaDeSC. We used a retrospective cohort of 1341 HNC patients from Helsinki University Hospital, utilizing their pre-treatment clinical and demographic data from electronic health records. We identified six previously unrecognized HNC phenotypes with distinct survival patterns, highlighting the significance of BMI, sleep apnoea, TNM stage, tumour site, treatment intention, gender, and age-related treatment modalities. VaDeSC demonstrated strong predictive accuracy, clustering performance and generalizability, achieving C-index of 0.895 ± 0.015 on training and 0.782 ± 0.013 on test data. The phenotype purity was 0.102 ± 0.11 on training-validation data and 0.143 ± 0.006 on test data. Our findings demonstrated that clustering pre-treatment clinical survival data offers intuitively interpretable results with accurate individualised predictions. This approach holds significant potential for novel phenotype discovery across various diseases and endpoints.

A comparative study of methods for dynamic survival analysis

Introduction

Dynamic survival analysis has become an effective approach for predicting time-to-event outcomes based on longitudinal data in neurology, cognitive health, and other health-related domains. With advancements in machine learning, several new methods have been introduced, often using a two-stage approach: first extracting features from longitudinal trajectories and then using these to predict survival probabilities.

Methods

This work compares several combinations of longitudinal and survival models, assessing their predictive performance across different training strategies. Using synthetic and real-world cognitive health data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), we explore the strengths and limitations of each model.

Results

Among the considered survival models, the Random Survival Forest consistently delivered strong results across different datasets, longitudinal models, and training strategies. On the ADNI dataset the best performing method was Random Survival Forest with the last visit benchmark and super landmarking with an average tdAUC of 0.96 and brier score of 0.07. Several other methods, including Cox Proportional Hazards and the Recurrent Neural Network, achieve similar scores. While the tested longitudinal models often struggled to outperform simple benchmarks, neural network models showed some improvement in simulated scenarios with sufficiently informative longitudinal trajectories.

Discussion

Our findings underscore the importance of aligning model selection and training strategies with the specific characteristics of the data and the target application, providing valuable insights that can inform future developments in dynamic survival analysis.

Survival path model outperforms conventional static machine learning models in long-term dynamic prognosis prediction for patients with intermediate stage hepatocellular carcinoma

Motivation

Patients with intermediate stage hepatocellular carcinoma (HCC) require repeated disease monitoring, prognosis assessment, and treatment planning. A novel machine learning model called survival path mapping (SP) model was developed, while its performance as compared with conventional machine learning models remains unknown. Between January 2007 and December 2018, the time-series data of 2644 intermediate stage HCC patients from four medical centers in China were reviewed and included. Static machine learning models by Gaussian Naive Bayes (GNB), support vector machine (SVM), and random forest (RF) for the prediction of survivorship were built based on data at initial admission. Longitudinal data divided into different time slices were utilized for the construction of the SP model. The time-dependent c-index was compared between models.

Results

The training set, internal testing set, and external testing set consisted of 1560, 670, and 414 HCC patients, respectively. The survival path model had superior or non-inferior performance in prognosis prediction compared to GNB and RF models since the 12th month after initial diagnosis in the training set and the external testing set. The survival path model had higher time-dependent c-index over all conventional ML models since the 6th month in the external testing cohort. In conclusion, the survival path model had superior performance in long-term dynamic prognosis prediction compared to conventional static machine learning models for intermediate stage HCC.

Availability and implementation

The parameters of models are provided in the manuscript.

Concordance-based Predictive Uncertainty (CPU)-Index: Proof-of-concept with application towards improved specificity of lung cancers on low dose screening CT

In this paper, we introduce a novel concordance-based predictive uncertainty (CPU)-Index, which integrates insights from subgroup analysis and personalized AI time-to-event models. Through its application in refining lung cancer screening (LCS) predictions generated by an individualized AI time-to-event model trained with fused data of low dose CT (LDCT) radiomics with patient demographics, we demonstrate its effectiveness, resulting in improved risk assessment compared to the Lung CT Screening Reporting & Data System (Lung-RADS). Subgroup-based Lung-RADS faces challenges in representing individual variations and relies on a limited set of predefined characteristics, resulting in variable predictions. Conversely, personalized AI time-to-event models are hindered by transparency issues and biases from censored data. By measuring the prediction consistency between subgroup analysis and AI time-to-event models, the CPU-Index framework offers a nuanced evaluation of the bias-variance trade-off and improves the transparency and reliability of predictions. Consistency was estimated by the concordance index of subgroup analysis-based similarity rank and model prediction similarity rank. Subgroup analysis-based similarity loss was defined as the sum-of-the-difference between Lung-RADS and feature-level 0-1 loss. Model prediction similarity loss was defined as squared loss. To test our approach, we identified 3,326 patients who underwent LDCT for LCS from 1/1/2015 to 6/30/2020 with confirmation of lung cancer on pathology within one year. For each LDCT image, the lesion associated with a Lung-RADS score was detected using a pretrained deep learning model from Medical Open Network for AI (MONAI), from which radiomic features were extracted. Radiomics were optimally fused with patient demographics via a positional encoding scheme and used to train a neural multi-task logistic regression time-to-event model that predicts malignancy. Performance was maximized when radiomics features were fused with positionally encoded demographic features. In this configuration, our algorithm raised the AUC from 0.81 ± 0.04 to 0.89 ± 0.02. Compared to standard Lung-RADS, our approach reduced the False-Positive-Rate from 0.41 ± 0.02 to 0.30 ± 0.12 while maintaining the same False-Negative-Rate. Our methodology enhances lung cancer risk assessment by estimating prediction uncertainty and adjusting accordingly. Furthermore, the optimal integration of radiomics and patient demographics improved overall diagnostic performance, indicating their complementary nature.

Contrastive learning with transformer for adverse endpoint prediction in patients on DAPT post-coronary stent implantation

Background

Effective management of dual antiplatelet therapy (DAPT) following drug-eluting stent (DES) implantation is crucial for preventing adverse events. Traditional prognostic tools, such as rule-based methods or Cox regression, despite their widespread use and ease, tend to yield moderate predictive accuracy within predetermined timeframes. This study introduces a new contrastive learning-based approach to enhance prediction efficacy over multiple time intervals.

Methods

We utilized retrospective, real-world data from the OneFlorida + Clinical Research Consortium. Our study focused on two primary endpoints: ischemic and bleeding events, with prediction windows of 1, 2, 3, 6, and 12 months post-DES implantation. Our approach first utilized an auto-encoder to compress patient features into a more manageable, condensed representation. Following this, we integrated a Transformer architecture with multi-head attention mechanisms to focus on and amplify the most salient features, optimizing the representation for better predictive accuracy. Then, we applied contrastive learning to enable the model to further refine its predictive capabilities by maximizing intra-class similarities and distinguishing inter-class differences. Meanwhile, the model was holistically optimized using multiple loss functions, to ensure the predicted results closely align with the ground-truth values from various perspectives. We benchmarked model performance against three cutting-edge deep learning-based survival models, i.e., DeepSurv, DeepHit, and SurvTrace.

Results

The final cohort comprised 19,713 adult patients who underwent DES implantation with more than 1 month of records after coronary stenting. Our approach demonstrated superior predictive performance for both ischemic and bleeding events across prediction windows of 1, 2, 3, 6, and 12 months, with time-dependent concordance (Ctd) index values ranging from 0.88 to 0.80 and 0.82 to 0.77, respectively. It consistently outperformed the baseline models, including DeepSurv, DeepHit, and SurvTrace, with statistically significant improvement in the Ctd-index values for most evaluated scenarios.

Conclusion

The robust performance of our contrastive learning-based model underscores its potential to enhance DAPT management significantly. By delivering precise predictive insights at multiple time points, our method meets the current need for adaptive, personalized therapeutic strategies in cardiology, thereby offering substantial value in improving patient outcomes.

tdCoxSNN: Time-dependent Cox survival neural network for continuous-time dynamic prediction

The aim of dynamic prediction is to provide individualized risk predictions over time, which are updated as new data become available. In pursuit of constructing a dynamic prediction model for a progressive eye disorder, age-related macular degeneration (AMD), we propose a time-dependent Cox survival neural network (tdCoxSNN) to predict its progression using longitudinal fundus images. tdCoxSNN builds upon the time-dependent Cox model by utilizing a neural network to capture the nonlinear effect of time-dependent covariates on the survival outcome. Moreover, by concurrently integrating a convolutional neural network with the survival network, tdCoxSNN can directly take longitudinal images as input. We evaluate and compare our proposed method with joint modelling and landmarking approaches through extensive simulations. We applied the proposed approach to two real datasets. One is a large AMD study, the Age-Related Eye Disease Study, in which more than 50,000 fundus images were captured over a period of 12 years for more than 4,000 participants. Another is a public dataset of the primary biliary cirrhosis disease, where multiple laboratory tests were longitudinally collected to predict the time-to-liver transplant. Our approach demonstrates commendable predictive performance in both simulation studies and the analysis of the two real datasets.

Machine learning-based infection diagnostic and prognostic models in post-acute care settings: a systematic review

OBJECTIVES

This study aims to (1) review machine learning (ML)-based models for early infection diagnostic and prognosis prediction in post-acute care (PAC) settings, (2) identify key risk predictors influencing infection-related outcomes, and (3) examine the quality and limitations of these models.

MATERIALS AND METHODS

PubMed, Web of Science, Scopus, IEEE Xplore, CINAHL, and ACM digital library were searched in February 2024. Eligible studies leveraged PAC data to develop and evaluate ML models for infection-related risks. Data extraction followed the CHARMS checklist. Quality appraisal followed the PROBAST tool. Data synthesis was guided by the socio-ecological conceptual framework.

RESULTS

Thirteen studies were included, mainly focusing on respiratory infections and nursing homes. Most used regression models with structured electronic health record data. Since 2020, there has been a shift toward advanced ML algorithms and multimodal data, biosensors, and clinical notes being significant sources of unstructured data. Despite these advances, there is insufficient evidence to support performance improvements over traditional models. Individual-level risk predictors, like impaired cognition, declined function, and tachycardia, were commonly used, while contextual-level predictors were barely utilized, consequently limiting model fairness. Major sources of bias included lack of external validation, inadequate model calibration, and insufficient consideration of data complexity.

DISCUSSION AND CONCLUSION

Despite the growth of advanced modeling approaches in infection-related models in PAC settings, evidence supporting their superiority remains limited. Future research should leverage a socio-ecological lens for predictor selection and model construction, exploring optimal data modalities and ML model usage in PAC, while ensuring rigorous methodologies and fairness considerations.

Enhanced neurological anomaly detection in MRI images using deep convolutional neural networks

Introduction

Neurodegenerative diseases, including Parkinson's, Alzheimer's, and epilepsy, pose significant diagnostic and treatment challenges due to their complexity and the gradual degeneration of central nervous system structures. This study introduces a deep learning framework designed to automate neuro-diagnostics, addressing the limitations of current manual interpretation methods, which are often time-consuming and prone to variability.

Methods

We propose a specialized deep convolutional neural network (DCNN) framework aimed at detecting and classifying neurological anomalies in MRI data. Our approach incorporates key preprocessing techniques, such as reducing noise and normalizing image intensity in MRI scans, alongside an optimized model architecture. The model employs Rectified Linear Unit (ReLU) activation functions, the Adam optimizer, and a random search strategy to fine-tune hyper-parameters like learning rate, batch size, and the number of neurons in fully connected layers. To ensure reliability and broad applicability, cross-fold validation was used.

Results and discussion

Our DCNN achieved a remarkable classification accuracy of 98.44%, surpassing well-known models such as ResNet-50 and AlexNet when evaluated on a comprehensive MRI dataset. Moreover, performance metrics such as precision, recall, and F1-score were calculated separately, confirming the robustness and efficiency of our model across various evaluation criteria. Statistical analyses, including ANOVA and t-tests, further validated the significance of the performance improvements observed with our proposed method. This model represents an important step toward creating a fully automated system for diagnosing and planning treatment for neurological diseases. The high accuracy of our framework highlights its potential to improve diagnostic workflows by enabling precise detection, tracking disease progression, and supporting personalized treatment strategies. While the results are promising, further research is necessary to assess how the model performs across different clinical scenarios. Future studies could focus on integrating additional data types, such as longitudinal imaging and multimodal techniques, to further enhance diagnostic accuracy and clinical utility. These findings mark a significant advancement in applying deep learning to neuro-diagnostics, with promising implications for improving patient outcomes and clinical practices.

Dynamic relations between longitudinal morphological, behavioral, and emotional indicators and cognitive impairment: evidence from the Chinese Longitudinal Healthy Longevity Survey

BACKGROUND

We aimed to assess the effects of body mass index (BMI), activities of daily living (ADL), and subjective well-being (SWB) on cognitive impairment and propose dynamic risk prediction models for aging cognitive decline.

METHODS

We leveraged the Chinese Longitudinal Healthy Longevity Survey from 1998 to 2018. Cognitive status was measured using the Chinese Mini-Mental State Examination. We employed repeated measures correlation to assess associations, linear mixed-effect models to characterize the longitudinal changes, and Cox proportional hazard regression to model survival time. Dynamic predictive models were established based on the Bayesian joint model and deep learning approach named dynamic-DeepHit. Marginal structural Cox models were adopted to control for time-varying confounding factors and assess effect sizes.

RESULTS

ADL, SWB, and BMI showed protective effects on cognitive impairment after controlling observed confounding factors, with respective direct hazard ratios of 0.756 (0.741, 0.771), 0.912 (0.902, 0.921), and 0.919 (0.909, 0.929). Dynamic risk predictive models manifested high accuracy (best AUC = 0.89). ADL was endowed with the best predictive capability, although the combination of BMI, ADL, and SWB showed the most remarkable performance.

CONCLUSIONS

BMI, ADL, and SWB are protective factors for cognitive impairment. A dynamic prediction model using these indicators can efficiently identify vulnerable individuals with high accuracy.

The Digital Transformation in Health: How AI Can Improve the Performance of Health Systems

Mobile health has the potential to revolutionize health care delivery and patient engagement. In this work, we discuss how integrating Artificial Intelligence into digital health applications focused on supply chain operation, patient management, and capacity building, among other use cases, can improve the health system and public health performance. We present the Causal Foundry Artificial Intelligence and Reinforcement Learning platform, which allows the delivery of adaptive interventions whose impact can be optimized through experimentation and real-time monitoring. The system can integrate multiple data sources and digital health applications. The flexibility of this platform to connect to various mobile health applications and digital devices, and to send personalized recommendations based on past data and predictions, can significantly improve the impact of digital tools on health system outcomes. The potential for resource-poor settings, where the impact of this approach on health outcomes could be decisive, is discussed. This framework is similarly applicable to improving efficiency in health systems where scarcity is not an issue.

The Deep Promotion Time Cure Model

We propose a novel method for predicting time-to-event data in the presence of cure fractions based on flexible survival models integrated into a deep neural network (DNN) framework. Our approach allows for nonlinear relationships and high-dimensional interactions between covariates and survival and is suitable for large-scale applications. To ensure the identifiability of the overall predictor formed of an additive decomposition of interpretable linear and nonlinear effects and potential higher-dimensional interactions captured through a DNN, we employ an orthogonalization layer. We demonstrate the usefulness and computational efficiency of our method via simulations and apply it to a large portfolio of U.S. mortgage loans. Here, we find not only a better predictive performance of our framework but also a more realistic picture of covariate effects.

DySurv: dynamic deep learning model for survival analysis with conditional variational inference

OBJECTIVE

Machine learning applications for longitudinal electronic health records often forecast the risk of events at fixed time points, whereas survival analysis achieves dynamic risk prediction by estimating time-to-event distributions. Here, we propose a novel conditional variational autoencoder-based method, DySurv, which uses a combination of static and longitudinal measurements from electronic health records to estimate the individual risk of death dynamically.

MATERIALS AND METHODS

DySurv directly estimates the cumulative risk incidence function without making any parametric assumptions on the underlying stochastic process of the time-to-event. We evaluate DySurv on 6 time-to-event benchmark datasets in healthcare, as well as 2 real-world intensive care unit (ICU) electronic health records (EHR) datasets extracted from the eICU Collaborative Research (eICU) and the Medical Information Mart for Intensive Care database (MIMIC-IV).

RESULTS

DySurv outperforms other existing statistical and deep learning approaches to time-to-event analysis across concordance and other metrics. It achieves time-dependent concordance of over 60% in the eICU case. It is also over 12% more accurate and 22% more sensitive than in-use ICU scores like Acute Physiology and Chronic Health Evaluation (APACHE) and Sequential Organ Failure Assessment (SOFA) scores. The predictive capacity of DySurv is consistent and the survival estimates remain disentangled across different datasets.

DISCUSSION

Our interdisciplinary framework successfully incorporates deep learning, survival analysis, and intensive care to create a novel method for time-to-event prediction from longitudinal health records. We test our method on several held-out test sets from a variety of healthcare datasets and compare it to existing in-use clinical risk scoring benchmarks.

CONCLUSION

While our method leverages non-parametric extensions to deep learning-guided estimations of the survival distribution, further deep learning paradigms could be explored.

Prediction of cardiovascular markers and diseases using retinal fundus images and deep learning: a systematic scoping review

Rapid development in deep learning for image analysis inspired studies to focus on predicting cardiovascular risk using retinal fundus images. This scoping review aimed to identify and describe studies using retinal fundus images and deep learning to predict cardiovascular risk markers and diseases. We searched MEDLINE and Embase on 17 November 2023. Abstracts and relevant full-text articles were independently screened by two reviewers. We included studies that used deep learning for the analysis of retinal fundus images to predict cardiovascular risk markers or cardiovascular diseases (CVDs) and excluded studies only using predefined characteristics of retinal fundus images. Study characteristics were presented using descriptive statistics. We included 24 articles published between 2018 and 2023. Among these, 23 (96%) were cross-sectional studies and eight (33%) were follow-up studies with clinical CVD outcomes. Seven studies included a combination of both designs. Most studies (96%) used convolutional neural networks to process images. We found nine (38%) studies that incorporated clinical risk factors in the prediction and four (17%) that compared the results to commonly used clinical risk scores in a prospective setting. Three of these reported improved discriminative performance. External validation of models was rare (21%). There is increasing interest in using retinal fundus images in cardiovascular risk assessment with some studies demonstrating some improvements in prediction. However, more prospective studies, comparisons of results to clinical risk scores, and models augmented with traditional risk factors can strengthen further research in the field.

Deep representation learning from electronic medical records identifies distinct symptom based subtypes and progression patterns for COVID-19 prognosis

OBJECTIVE

Symptoms are significant kind of phenotypes for managing and controlling of the burst of acute infectious diseases, such as COVID-19. Although patterns of symptom clusters and time series have been considered the high potential prediction factors for the prognosis of patients, the elaborated subtypes and their progression patterns based on symptom phenotypes related to the prognosis of COVID-19 patients still need be detected. This study aims to investigate patient subtypes and their progression patterns with distinct features of outcome and prognosis.

METHODS

This study included a total of 14,139 longitudinal electronic medical records (EMRs) obtained from four hospitals in Hubei Province, China, involving 2,683 individuals in the early stage of COVID-19 pandemic. A deep representation learning model was developed to help acquire the symptom profiles of patients. K-means clustering algorithm is used to divide them into distinct subtypes. Subsequently, symptom progression patterns were identified by considering the subtypes associated with patients upon admission and discharge. Furthermore, we used Fisher's test to identify significant clinical entities for each subtype.

RESULTS

Three distinct patient subtypes exhibiting specific symptoms and prognosis have been identified. Particularly, Subtype 0 includes 44.2% of the whole and is characterized by poor appetite, fatigue and sleep disorders; Subtype 1 includes 25.6% cases and is characterized by confusion, cough with bloody sputum, encopresis and urinary incontinence; Subtype 2 includes 30.2% cases and is characterized by dry cough and rhinorrhea. These three subtypes demonstrate significant disparities in prognosis, with the mortality rates of 4.72%, 8.59%, and 0.25% respectively. Furthermore, symptom cluster progression patterns showed that patients with Subtype 0 who manifest dark yellow urine, chest pain, etc. in the admission stage exhibit an elevated risk of transforming into the more severe subtypes with poor outcome, whereas those presenting with nausea and vomiting tend to incline towards entering the milder subtype.

CONCLUSION

This study has proposed a clinical meaningful approach by utilizing the deep representation learning and real-world EMR data containing symptom phenotypes to identify the COVID-19 subtypes and their progression patterns. The results would be potentially useful to help improve the precise stratification and management of acute infectious diseases.

Machine Learning for Dynamic Prognostication of Patients With Hepatocellular Carcinoma Using Time-Series Data: Survival Path Versus Dynamic-DeepHit HCC Model

Objectives

Patients with intermediate or advanced hepatocellular carcinoma (HCC) require repeated disease monitoring, prognosis assessment and treatment planning. In 2018, a novel machine learning methodology "survival path" (SP) was developed to facilitate dynamic prognosis prediction and treatment planning. One year after, a deep learning approach called Dynamic Deephit was developed. The performance of the two state-of-art models in dynamic prognostication have not been compared.

Methods

We trained and tested the SP and Dynamic DeepHit models in a large cohort of 2511 HCC patients using time-series data. The time-series data were converted into data of time slices, with an interval of three months. The time-dependent c-index for OS at given prediction time (t = 1, 6, 12, 18 months) and evaluation time (∆t = 3, 6, 9, 12, 18, 24, 36, 48 months) were compared.

Results

The comparison between SP model and Dynamic DeepHit-HCC model showed the latter had significant better performance at the time of initial admission. The time-dependent c-index of Dynamic DeepHit-HCC model gradually decreased with the extension of time (from 0.756 to 0.639 in the training set; from 0.787 to 0.661 in internal testing set; from 0.725 to 0.668 in multicenter testing set); while the time-dependent c-index of SP model displayed an increased trend (from 0.665 to 0.748 in the training set; from 0.608 to 0.743 in internal testing set; from 0.643 to 0.720 in multicenter testing set). When the prediction time comes to 6 months or later since initial treatment, the survival path model outperformed the dynamic DeepHit model at late evaluation times (∆t > 12 months).

Conclusions

This research highlighted the unique strengths of both models. The SP model had advantage in long term prediction while the Dynamic DeepHit-HCC model had advantages in prediction at near time points. Fine selection of models is needed in dealing with different scenarios.

Machine learning and statistical models to predict all-cause mortality in type 2 diabetes: Results from the UK Biobank study

AIMS

This study aims to compare the performance of contemporary machine learning models with statistical models in predicting all-cause mortality in patients with type 2 diabetes mellitus and to develop a user-friendly mortality risk prediction tool.

METHODS

A prospective cohort study was conducted including 22,579 people with diabetes from the UK Biobank. Models evaluated include Cox proportional hazards, random survival forests (RSF), gradient boosting (GB) survival, DeepSurv, and DeepHit.

RESULTS

Over a median follow-up period of 9 years, 2,665 patients died. Machine learning models outperformed the Cox model in the validation dataset, with C-index values of 0.72-0.73 vs. 0.71 for Cox (p < 0.01). Deep learning models, particularly DeepHit, demonstrated superior calibration and achieved lower Brier scores (0.09 vs. 0.10 for Cox, p < 0.05). An online prediction tool based on the DeepHit was developed for patient care: http://123.57.42.89:6006/.

CONCLUSIONS

Machine learning models performed better than statistical models, highlighting the potential of machine learning techniques for predicting all-cause mortality risk and facilitating personalized healthcare management for individuals with diabetes.

Harnessing the power of longitudinal medical imaging for eye disease prognosis using Transformer-based sequence modeling

Deep learning has enabled breakthroughs in automated diagnosis from medical imaging, with many successful applications in ophthalmology. However, standard medical image classification approaches only assess disease presence at the time of acquisition, neglecting the common clinical setting of longitudinal imaging. For slow, progressive eye diseases like age-related macular degeneration (AMD) and primary open-angle glaucoma (POAG), patients undergo repeated imaging over time to track disease progression and forecasting the future risk of developing a disease is critical to properly plan treatment. Our proposed Longitudinal Transformer for Survival Analysis (LTSA) enables dynamic disease prognosis from longitudinal medical imaging, modeling the time to disease from sequences of fundus photography images captured over long, irregular time periods. Using longitudinal imaging data from the Age-Related Eye Disease Study (AREDS) and Ocular Hypertension Treatment Study (OHTS), LTSA significantly outperformed a single-image baseline in 19/20 head-to-head comparisons on late AMD prognosis and 18/20 comparisons on POAG prognosis. A temporal attention analysis also suggested that, while the most recent image is typically the most influential, prior imaging still provides additional prognostic value.

Hybridizing mechanistic modeling and deep learning for personalized survival prediction after immune checkpoint inhibitor immunotherapy

We present a study where predictive mechanistic modeling is combined with deep learning methods to predict individual patient survival probabilities under immune checkpoint inhibitor (ICI) immunotherapy. This hybrid approach enables prediction based on both measures that are calculable from mechanistic models of key mechanisms underlying ICI therapy that may not be directly measurable in the clinic and easily measurable quantities or patient characteristics that are not always readily incorporated into predictive mechanistic models. A deep learning time-to-event predictive model trained on a hybrid mechanistic + clinical data set from 93 patients achieved higher per-patient predictive accuracy based on event-time concordance, Brier score, and negative binomial log-likelihood-based criteria than when trained on only mechanistic model-derived values or only clinical data. Feature importance analysis revealed that both clinical and model-derived parameters play prominent roles in increasing prediction accuracy, further supporting the advantage of our hybrid approach.

Joint AI-driven event prediction and longitudinal modeling in newly diagnosed and relapsed multiple myeloma

Multiple myeloma management requires a balance between maximizing survival, minimizing adverse events to therapy, and monitoring disease progression. While previous work has proposed data-driven models for individual tasks, these approaches fail to provide a holistic view of a patient's disease state, limiting their utility to assist physician decision-making. To address this limitation, we developed a transformer-based machine learning model that jointly (1) predicts progression-free survival (PFS), overall survival (OS), and adverse events (AE), (2) forecasts key disease biomarkers, and (3) assesses the effect of different treatment strategies, e.g., ixazomib, lenalidomide, dexamethasone (IRd) vs lenalidomide, dexamethasone (Rd). Using TOURMALINE trial data, we trained and internally validated our model on newly diagnosed myeloma patients (N = 703) and externally validated it on relapsed and refractory myeloma patients (N = 720). Our model achieved superior performance to a risk model based on the multiple myeloma international staging system (ISS) (p < 0.001, Bonferroni corrected) and comparable performance to survival models trained separately on each task, but unable to forecast biomarkers. Our approach outperformed state-of-the-art deep learning models, tailored towards forecasting, on predicting key disease biomarkers (p < 0.001, Bonferroni corrected). Finally, leveraging our model's capacity to estimate individual-level treatment effects, we found that patients with IgA kappa myeloma appear to benefit the most from IRd. Our study suggests that a holistic assessment of a patient's myeloma course is possible, potentially serving as the foundation for a personalized decision support system.

Deep Learning-Based Dynamic Risk Prediction of Venous Thromboembolism for Patients With Ovarian Cancer in Real-World Settings From Electronic Health Records

PURPOSE

Patients with epithelial ovarian cancer (EOC) have an elevated risk for venous thromboembolism (VTE). To assess the risk of VTE, models were developed by statistical or machine learning algorithms. However, few models have accommodated deep learning (DL) algorithms in realistic clinical settings. We aimed to develop a predictive DL model, exploiting rich information from electronic health records (EHRs), including dynamic clinical features and the presence of competing risks.

METHODS

We extracted EHRs of 1,268 patients diagnosed with EOC from January 2007 through December 2017 at the National Cancer Center, Korea. DL survival networks using fully connected layers, temporal attention, and recurrent neural networks were adopted and compared with multi-perceptron-based classification models. Prediction accuracy was independently validated in the data set of 423 patients newly diagnosed with EOC from January 2018 to December 2019. Personalized risk plots displaying the individual interval risk were developed.

RESULTS

DL-based survival networks achieved a superior area under the receiver operating characteristic curve (AUROC) between 0.95 and 0.98 while the AUROC of classification models was between 0.85 and 0.90. As clinical information benefits the prediction accuracy, the proposed dynamic survival network outperformed other survival networks for the test and validation data set with the highest time-dependent concordance index (0.974, 0.975) and lowest Brier score (0.051, 0.049) at 6 months after a cancer diagnosis. Our visualization showed that the interval risk fluctuating along with the changes in longitudinal clinical features.

CONCLUSION

Adaption of dynamic patient clinical features and accounting for competing risks from EHRs into the DL algorithms demonstrated VTE risk prediction with high accuracy. Our results show that this novel dynamic survival network can provide personalized risk prediction with the potential to assist risk-based clinical intervention to prevent VTE among patients with EOC.

Predicting Drugs Suspected of Causing Adverse Drug Reactions Using Graph Features and Attention Mechanisms

BACKGROUND

Adverse drug reactions (ADRs) refer to an unintended harmful reaction that occurs after the administration of a medication for therapeutic purposes, which is unrelated to the intended pharmacological action of the drug. In the United States, ADRs account for 6% of all hospital admissions annually. The cost of ADR-related illnesses in 2016 was estimated at USD 528.4 billion. Increasing the awareness of ADRs is an effective measure to prevent them. Assessing suspected drugs in adverse events helps to enhance the awareness of ADRs.

METHODS

In this study, a suspect drug assisted judgment model (SDAJM) is designed to identify suspected drugs in adverse events. This framework utilizes the graph isomorphism network (GIN) and an attention mechanism to extract features based on patients' demographic information, drug information, and ADR information.

RESULTS

By comparing it with other models, the results of various tests show that this model performs well in predicting the suspected drugs in adverse reaction events. ADR signal detection was conducted on a group of cardiovascular system drugs, and case analyses were performed on two classic drugs, Mexiletine and Captopril, as well as on two classic antithyroid drugs. The results indicate that the model can accomplish the task of predicting drug ADRs. Validation using benchmark datasets from ten drug discovery domains shows that the model is applicable to classification tasks on the Tox21 and SIDER datasets.

CONCLUSIONS

This study applies deep learning methods to construct the SDAJM model for three purposes: (1) identifying drugs suspected to cause adverse drug events (ADEs), (2) predicting the ADRs of drugs, and (3) other drug discovery tasks. The results indicate that this method can offer new directions for research in the field of ADRs.

The Dynamic Nature of the Socioeconomic Determinants of Cardiovascular Health: A Narrative Review

Despite decades of social epidemiologic research, health inequities remain pervasive and ubiquitous in Canada and elsewhere. One reason may be our use of socioeconomic measurement, which has often relied on single point-in-time exposures. To explore the extent to which researchers have incorporated dynamic socioeconomic measurement into cardiovascular health outcome evaluations, we performed a narrative review. We estimated the prevalence of socioeconomic longitudinal cardiovascular research studies that identified socioeconomic exposures at 2 or more points in time between the years of 2019 and 2023. We defined cardiovascular outcome studies as those that examined coronary artery disease, myocardial infarction, acute coronary syndrome, stroke, heart failure, cardiac arrhythmias, cardiac death, cardiometabolic factors, transient ischemic attacks, peripheral artery disease, or hypertension. Socioeconomic exposures included individual income, neighbourhood income, intergenerational social mobility, education, occupation, insurance status, and economic security. Seven percent of socioeconomic cardiovascular outcome studies have measured socioeconomic status at 2 or more points in time throughout the follow-up period, hypothesized mechanisms by which dynamic socioeconomic measures affected outcome focused on social mobility, accumulation, and critical period theories. Insights, implications, and future directions are discussed, in which we highlight ways in which postal code data can be better used methodologically as a dynamic socioeconomic measure. Future research must incorporate dynamic socioeconomic measurement to better inform root causes, interventions, and health-system designs if health equity is to be improved.

Enhanced osteoporotic fracture prediction in postmenopausal women using Bayesian optimization of machine learning models with genetic risk score

This study aimed to enhance the fracture risk prediction accuracy in major osteoporotic fractures (MOFs) and hip fractures (HFs) by integrating genetic profiles, machine learning (ML) techniques, and Bayesian optimization. The genetic risk score (GRS), derived from 1,103 risk single nucleotide polymorphisms (SNPs) from genome-wide association studies (GWAS), was formulated for 25,772 postmenopausal women from the Women's Health Initiative dataset. We developed four ML models: Support Vector Machine (SVM), Random Forest, XGBoost, and Artificial Neural Network (ANN) for binary fracture outcome and 10-year fracture risk prediction. GRS and FRAX clinical risk factors (CRFs) were used as predictors. Death as a competing risk was accounted for in ML models for time-to-fracture data. ML models were subsequently fine-tuned through Bayesian optimization, which displayed marked superiority over traditional grid search. Evaluation of the models' performance considered an array of metrics such as accuracy, weighted F1 Score, the area under the precision-recall curve (PRAUC), and the area under the receiver operating characteristic curve (AUC) for binary fracture predictions, and the C-index, Brier score, and dynamic mean AUC over a 10-year follow-up period for fracture risk predictions. We found that GRS-integrated XGBoost with Bayesian optimization is the most effective model, with an accuracy of 91.2% (95% CI: 90.4-92.0%) and an AUC of 0.739 (95% CI: 0.731-0.746) in MOF binary predictions. For 10-year fracture risk modeling, the XGBoost model attained a C-index of 0.795 (95% CI: 0.783-0.806) and a mean dynamic AUC of 0.799 (95% CI: 0.788-0.809). Compared to FRAX, the XGBoost model exhibited a categorical net reclassification improvement (NRI) of 22.6% (P = .004). A sensitivity analysis, which included BMD but lacked GRS, reaffirmed these findings. Furthermore, portability tests in diverse non-European groups, including Asians and African Americans, underscored the model's robustness and adaptability. This study accentuates the potential of combining genetic insights and optimized ML in strengthening fracture predictions, heralding new preventive strategies for postmenopausal women.

Exploring factors influencing time from dispatch to unit availability according to the transport decision in the pre-hospital setting: an exploratory study

BACKGROUND

Efficient resource distribution is important. Despite extensive research on response timings within ambulance services, nuances of time from unit dispatch to becoming available still need to be explored. This study aimed to identify the determinants of the duration between ambulance dispatch and readiness to respond to the next case according to the patients' transport decisions.

METHODS

Time from ambulance dispatch to availability (TDA) analysis according to the patients' transport decision (Transport versus Non-Transport) was conducted using R-Studio™ for a data set of 93,712 emergency calls managed by a Middle Eastern ambulance service from January to May 2023. Log-transformed Hazard Ratios (HR) were examined across diverse parameters. A Cox regression model was utilised to determine the influence of variables on TDA. Kaplan-Meier curves discerned potential variances in the time elapsed for both cohorts based on demographics and clinical indicators. A competing risk analysis assessed the probabilities of distinct outcomes occurring.

RESULTS

The median duration of elapsed TDA was 173 min for the transported patients and 73 min for those not transported. The HR unveiled Significant associations in various demographic variables. The Kaplan-Meier curves revealed variances in TDA across different nationalities and age categories. In the competing risk analysis, the 'Not Transported' group demonstrated a higher incidence of prolonged TDA than the 'Transported' group at specified time points.

CONCLUSIONS

Exploring TDA offers a novel perspective on ambulance services' efficiency. Though promising, the findings necessitate further exploration across diverse settings, ensuring broader applicability. Future research should consider a comprehensive range of variables to fully harness the utility of this period as a metric for healthcare excellence.

Survival Analysis of High-Dimensional Data With Graph Convolutional Networks and Geometric Graphs

This article proposes a survival model based on graph convolutional networks (GCNs) with geometric graphs directly constructed from high-dimensional features. First, we clarify that the graphs used in GCNs play an important role in processing the relational information of samples, and the graphs that align well with the underlying data structure could be beneficial for survival analysis. Second, we show that sparse geometric graphs derived from high-dimensional data are more favorable compared with dense graphs when used in GCNs for survival analysis. Third, from this insight, we propose a model for survival analysis based on GCNs. By using multiple sparse geometric graphs and a proposed sequential forward floating selection algorithm, the new model is able to simultaneously perform survival analysis and unveil the local neighborhoods of samples. The experimental results on real-world datasets show that the proposed survival analysis approach based on GCNs outperforms a variety of existing methods and indicate that geometric graphs can aid survival analysis of high-dimensional data.

Developing deep learning-based strategies to predict the risk of hepatocellular carcinoma among patients with nonalcoholic fatty liver disease from electronic health records

OBJECTIVE

The accuracy of deep learning models for many disease prediction problems is affected by time-varying covariates, rare incidence, covariate imbalance and delayed diagnosis when using structured electronic health records data. The situation is further exasperated when predicting the risk of one disease on condition of another disease, such as the hepatocellular carcinoma risk among patients with nonalcoholic fatty liver disease due to slow, chronic progression, the scarce of data with both disease conditions and the sex bias of the diseases. The goal of this study is to investigate the extent to which the aforementioned issues influence deep learning performance, and then devised strategies to tackle these challenges. These strategies were applied to improve hepatocellular carcinoma risk prediction among patients with nonalcoholic fatty liver disease.

METHODS

We evaluated two representative deep learning models in the task of predicting the occurrence of hepatocellular carcinoma in a cohort of patients with nonalcoholic fatty liver disease (n = 220,838) from a national EHR database. The disease prediction task was carefully formulated as a classification problem while taking censorship and the length of follow-up into consideration.

RESULTS

We developed a novel backward masking scheme to deal with the issue of delayed diagnosis which is very common in EHR data analysis and evaluate how the length of longitudinal information after the index date affects disease prediction. We observed that modeling time-varying covariates improved the performance of the algorithms and transfer learning mitigated reduced performance caused by the lack of data. In addition, covariate imbalance, such as sex bias in data impaired performance. Deep learning models trained on one sex and evaluated in the other sex showed reduced performance, indicating the importance of assessing covariate imbalance while preparing data for model training.

CONCLUSIONS

The strategies developed in this work can significantly improve the performance of hepatocellular carcinoma risk prediction among patients with nonalcoholic fatty liver disease. Furthermore, our novel strategies can be generalized to apply to other disease risk predictions using structured electronic health records, especially for disease risks on condition of another disease.

Optimizing dynamic predictions from joint models using super learning

Joint models for longitudinal and time-to-event data are often employed to calculate dynamic individualized predictions used in numerous applications of precision medicine. Two components of joint models that influence the accuracy of these predictions are the shape of the longitudinal trajectories and the functional form linking the longitudinal outcome history to the hazard of the event. Finding a single well-specified model that produces accurate predictions for all subjects and follow-up times can be challenging, especially when considering multiple longitudinal outcomes. In this work, we use the concept of super learning and avoid selecting a single model. In particular, we specify a weighted combination of the dynamic predictions calculated from a library of joint models with different specifications. The weights are selected to optimize a predictive accuracy metric using V-fold cross-validation. We use as predictive accuracy measures the expected quadratic prediction error and the expected predictive cross-entropy. In a simulation study, we found that the super learning approach produces results very similar to the Oracle model, which was the model with the best performance in the test datasets. All proposed methodology is implemented in the freely available R package JMbayes2.

Hybridizing mechanistic mathematical modeling with deep learning methods to predict individual cancer patient survival after immune checkpoint inhibitor therapy

We present a study where predictive mechanistic modeling is used in combination with deep learning methods to predict individual patient survival probabilities under immune checkpoint inhibitor (ICI) therapy. This hybrid approach enables prediction based on both measures that are calculable from mechanistic models (but may not be directly measurable in the clinic) and easily measurable quantities or characteristics (that are not always readily incorporated into predictive mechanistic models). The mechanistic model we have applied here can predict tumor response from CT or MRI imaging based on key mechanisms underlying checkpoint inhibitor therapy, and in the present work, its parameters were combined with readily-available clinical measures from 93 patients into a hybrid training set for a deep learning time-to-event predictive model. Analysis revealed that training an artificial neural network with both mechanistic modeling-derived and clinical measures achieved higher per-patient predictive accuracy based on event-time concordance, Brier score, and negative binomial log-likelihood-based criteria than when only mechanistic model-derived values or only clinical data were used. Feature importance analysis revealed that both clinical and model-derived parameters play prominent roles in neural network decision making, and in increasing prediction accuracy, further supporting the advantage of our hybrid approach. We anticipate that many existing mechanistic models may be hybridized with deep learning methods in a similar manner to improve predictive accuracy through addition of additional data that may not be readily implemented in mechanistic descriptions.

Added prognostic value of 3D deep learning-derived features from preoperative MRI for adult-type diffuse gliomas

BACKGROUND

To investigate the prognostic value of spatial features from whole-brain MRI using a three-dimensional (3D) convolutional neural network for adult-type diffuse gliomas.

METHODS

In a retrospective, multicenter study, 1925 diffuse glioma patients were enrolled from 5 datasets: SNUH (n = 708), UPenn (n = 425), UCSF (n = 500), TCGA (n = 160), and Severance (n = 132). The SNUH and Severance datasets served as external test sets. Precontrast and postcontrast 3D T1-weighted, T2-weighted, and T2-FLAIR images were processed as multichannel 3D images. A 3D-adapted SE-ResNeXt model was trained to predict overall survival. The prognostic value of the deep learning-based prognostic index (DPI), a spatial feature-derived quantitative score, and established prognostic markers were evaluated using Cox regression. Model evaluation was performed using the concordance index (C-index) and Brier score.

RESULTS

The MRI-only median DPI survival prediction model achieved C-indices of 0.709 and 0.677 (BS = 0.142 and 0.215) and survival differences (P < 0.001 and P = 0.002; log-rank test) for the SNUH and Severance datasets, respectively. Multivariate Cox analysis revealed DPI as a significant prognostic factor, independent of clinical and molecular genetic variables: hazard ratio = 0.032 and 0.036 (P < 0.001 and P = 0.004) for the SNUH and Severance datasets, respectively. Multimodal prediction models achieved higher C-indices than models using only clinical and molecular genetic variables: 0.783 vs. 0.774, P = 0.001, SNUH; 0.766 vs. 0.748, P = 0.023, Severance.

CONCLUSIONS

The global morphologic feature derived from 3D CNN models using whole-brain MRI has independent prognostic value for diffuse gliomas. Combining clinical, molecular genetic, and imaging data yields the best performance.

Development and Application of a Novel Machine Learning Model Predicting Pancreatic Cancer-Specific Mortality

Precise prognostication is vital for guiding treatment decisions in people diagnosed with pancreatic cancer. Existing models depend on predetermined variables, constraining their effectiveness. Our objective was to explore a novel machine learning approach to enhance a prognostic model for predicting pancreatic cancer-specific mortality and, subsequently, to assess its performance against Cox regression models. Datasets were retrospectively collected and analyzed for 9,752 patients diagnosed with pancreatic cancer and with surgery performed. The primary outcomes were the mortality of patients with pancreatic carcinoma at one year, three years, and five years. Model discrimination was assessed using the concordance index (C-index), and calibration was assessed using Brier scores. The Survival Quilts model was compared with Cox regression models in clinical use, and decision curve analysis was done. The Survival Quilts model demonstrated robust discrimination for one-year (C-index 0.729), three-year (C-index 0.693), and five-year (C-index 0.672) pancreatic cancer-specific mortality. In comparison to Cox models, the Survival Quilts models exhibited a higher C-index up to 32 months but displayed inferior performance after 33 months. A subgroup analysis was conducted, revealing that within the subset of individuals without metastasis, the Survival Quilts models showcased a significant advantage over the Cox models. In the cohort with metastatic pancreatic cancer, Survival Quilts outperformed the Cox model before 24 months but exhibited a weaker performance after 25 months. This study has developed and validated a novel machine learning-based Survival Quilts model to predict pancreatic cancer-specific mortality that outperforms the Cox regression model.

Utility of multimodal longitudinal imaging data for dynamic prediction of cardiovascular and renal disease: the CARDIA study

Background

Medical examinations contain repeatedly measured data from multiple visits, including imaging variables collected from different modalities. However, the utility of such data for the prediction of time-to-event is unknown, and only a fraction of the data is typically used for risk prediction. We hypothesized that multimodal longitudinal imaging data could improve dynamic disease prognosis of cardiovascular and renal disease (CVRD).

Methods

In a multi-centered cohort of 5,114 CARDIA participants, we included 166 longitudinal imaging variables from five imaging modalities: Echocardiography (Echo), Cardiac and Abdominal Computed Tomography (CT), Dual-Energy x-ray Absorptiometry (DEXA), Brain Magnetic Resonance Imaging (MRI) collected from young adulthood to mid-life over 30 years (1985-2016) to perform dynamic survival analysis of CVRD events using machine learning dynamic survival analysis (Dynamic-DeepHit, LTRCforest, and Extended Cox for Time-varying Covariates). Risk probabilities were continuously updated as new data were collected. Model performance was assessed using integrated AUC and C-index and compared to traditional risk factors.

Results

Longitudinal imaging data, even when being irregularly collected with high missing rates, improved CVRD dynamic prediction (0.03 in integrated AUC, up to 0.05 in C-index compared to traditional risk factors; best model's C-index = 0.80-0.83 up to 20 years from baseline) from young adulthood followed up to midlife. Among imaging variables, Echo and CT variables contributed significantly to improved risk estimation. Echo measured in early adulthood predicted midlife CVRD risks almost as well as Echo measured 10-15 years later (0.01 C-index difference). The most recent CT exam provided the most accurate prediction for short-term risk estimation. Brain MRI markers provided additional information from cardiac Echo and CT variables that led to a slightly improved prediction.

Conclusions

Longitudinal multimodal imaging data readily collected from follow-up exams can improve CVRD dynamic prediction. Echocardiography measured early can provide a good long-term risk estimation, while CT/calcium scoring variables carry atherosclerotic signatures that benefit more immediate risk assessment starting in middle-age.

Incorporating longitudinal history of risk factors into atherosclerotic cardiovascular disease risk prediction using deep learning

It is increasingly clear that longitudinal risk factor levels and trajectories are related to risk for atherosclerotic cardiovascular disease (ASCVD) above and beyond single measures. Currently used in clinical care, the Pooled Cohort Equations (PCE) are based on regression methods that predict ASCVD risk based on cross-sectional risk factor levels. Deep learning (DL) models have been developed to incorporate longitudinal data for risk prediction but its benefit for ASCVD risk prediction relative to the traditional Pooled Cohort Equations (PCE) remain unknown. Our study included 15,565 participants from four cardiovascular disease cohorts free of baseline ASCVD who were followed for adjudicated ASCVD. Ten-year ASCVD risk was calculated in the training set using our benchmark, the PCE, and a longitudinal DL model, Dynamic-DeepHit. Predictors included those incorporated in the PCE: sex, race, age, total cholesterol, high density lipid cholesterol, systolic and diastolic blood pressure, diabetes, hypertension treatment and smoking. The discrimination and calibration performance of the two models were evaluated in an overall hold-out testing dataset. Of the 15,565 participants in our dataset, 2170 (13.9%) developed ASCVD. The performance of the longitudinal DL model that incorporated 8 years of longitudinal risk factor data improved upon that of the PCE [AUROC: 0.815 (CI 0.782-0.844) vs 0.792 (CI 0.760-0.825)] and the net reclassification index was 0.385. The brier score for the DL model was 0.0514 compared with 0.0542 in the PCE. Incorporating longitudinal risk factors in ASCVD risk prediction using DL can improve model discrimination and calibration.

Deep learning-assisted survival prognosis in renal cancer: A CT scan-based personalized approach

This paper presents a deep learning (DL) approach for predicting survival probabilities of renal cancer patients based solely on preoperative CT imaging. The proposed approach consists of two networks: a classifier- and a survival- network. The classifier attempts to extract features from 3D CT scans to predict the ISUP grade of Renal cell carcinoma (RCC) tumors, as defined by the International Society of Urological Pathology (ISUP). Our classifier is a 3D convolutional neural network to avoid losing crucial information on the interconnection of slides in 3D images. We employ multiple procedures, including image augmentation, preprocessing, and concatenation, to improve the performance of the classifier. Given the strong correlation between ISUP grading and renal cancer prognosis in the clinical context, we use the ISUP grading features extracted by the classifier as the input to the survival network. By leveraging this clinical association and the classifier network, we are able to model our survival analysis using a simple DL-based network. We adopt a discrete LogisticHazard-based loss to extract intrinsic survival characteristics of RCC tumors from CT images. This allows us to build a completely parametric survival model that varies with patients' tumor characteristics and predicts non-proportional survival probability curves for different patients. Our results demonstrated that the proposed method could predict the future course of renal cancer with reasonable accuracy from the CT scans. The proposed method obtained an average concordance index of 0.72, an integrated Brier score of 0.15, and an area under the curve value of 0.71 on the test cohorts.

Dynamic prediction of malignant ventricular arrhythmias using neural networks in patients with an implantable cardioverter-defibrillator

BACKGROUND

Risk stratification for ventricular arrhythmias currently relies on static measurements that fail to adequately capture dynamic interactions between arrhythmic substrate and triggers over time. We trained and internally validated a dynamic machine learning (ML) model and neural network that extracted features from longitudinally collected electrocardiograms (ECG), and used these to predict the risk of malignant ventricular arrhythmias.

METHODS

A multicentre study in patients implanted with an implantable cardioverter-defibrillator (ICD) between 2007 and 2021 in two academic hospitals was performed. Variational autoencoders (VAEs), which combine neural networks with variational inference principles, and can learn patterns and structure in data without explicit labelling, were trained to encode the mean ECG waveforms from the limb leads into 16 variables. Supervised dynamic ML models using these latent ECG representations and clinical baseline information were trained to predict malignant ventricular arrhythmias treated by the ICD. Model performance was evaluated on a hold-out set, using time-dependent receiver operating characteristic (ROC) and calibration curves.

FINDINGS

2942 patients (61.7 ± 13.9 years, 25.5% female) were included, with a total of 32,129 ECG recordings during a mean follow-up of 43.9 ± 35.9 months. The mean time-varying area under the ROC curve for the dynamic model was 0.738 ± 0.07, compared to 0.639 ± 0.03 for a static (i.e. baseline-only model). Feature analyses indicated dynamic changes in latent ECG representations, particularly those affecting the T-wave morphology, were of highest importance for model predictions.

INTERPRETATION

Dynamic ML models and neural networks effectively leverage routinely collected longitudinal ECG recordings for personalised and updated predictions of malignant ventricular arrhythmias, outperforming static models.

FUNDING

This publication is part of the project DEEP RISK ICD (with project number 452019308) of the research programme Rubicon which is (partly) financed by the Dutch Research Council (NWO). This research is partly funded by the Amsterdam Cardiovascular Sciences (personal grant F.V.Y.T).

Survival Analysis for Multimode Ablation Using Self-Adapted Deep Learning Network Based on Multisource Features

Novel multimode thermal therapy by freezing before radio-frequency heating has achieved a desirable therapeutic effect in liver cancer. Compared with surgical resection, ablation treatment has a relatively high risk of tumor recurrence. To monitor tumor progression after ablation, we developed a novel survival analysis framework for survival prediction and efficacy assessment. We extracted preoperative and postoperative MRI radiomics features and vision transformer-based deep learning features. We also combined the immune features extracted from peripheral blood immune responses using flow cytometry and routine blood tests before and after treatment. We selected features using random survival forest and improved the deep Cox mixture (DCM) for survival analysis. To properly accommodate multitype input features, we proposed a self-adapted fully connected layer for locally and globally representing features. We evaluated the method using our clinical dataset. Of note, the immune features rank the highest feature importance and contribute significantly to the prediction accuracy. The results showed a promising C$^{\mathit{td}}$-index of 0.885 $\pm$ 0.040 and an integrated Brier score of 0.041 $\pm$ 0.014, which outperformed state-of-the-art method combinations of survival prediction. For each patient, individual survival probability was accurately predicted over time, which provided clinicians with trustable prognosis suggestions.

Sample size and predictive performance of machine learning methods with survival data: A simulation study

Prediction models are increasingly developed and used in diagnostic and prognostic studies, where the use of machine learning (ML) methods is becoming more and more popular over traditional regression techniques. For survival outcomes the Cox proportional hazards model is generally used and it has been proven to achieve good prediction performances with few strong covariates. The possibility to improve the model performance by including nonlinearities, covariate interactions and time-varying effects while controlling for overfitting must be carefully considered during the model building phase. On the other hand, ML techniques are able to learn complexities from data at the cost of hyper-parameter tuning and interpretability. One aspect of special interest is the sample size needed for developing a survival prediction model. While there is guidance when using traditional statistical models, the same does not apply when using ML techniques. This work develops a time-to-event simulation framework to evaluate performances of Cox regression compared, among others, to tuned random survival forest, gradient boosting, and neural networks at varying sample sizes. Simulations were based on replications of subjects from publicly available databases, where event times were simulated according to a Cox model with nonlinearities on continuous variables and time-varying effects and on the SEER registry data.

Using Bayesian Neural Networks to Select Features and Compute Credible Intervals for Personalized Survival Prediction

An Individual Survival Distribution (ISD) models a patient's personalized survival probability at all future time points. Previously, ISD models have been shown to produce accurate and personalized survival estimates (for example, time to relapse or to death) in several clinical applications. However, off-the-shelf neural-network-based ISD models are usually opaque models due to their limited support for meaningful feature selection and uncertainty estimation, which hinders their wide clinical adoption. Here, we introduce a Bayesian-neural-network-based ISD (BNN-ISD) model that produces accurate survival estimates but also quantifies the uncertainty in model's parameter estimation, which can be used to (1) rank the importance of the input features to support feature selection and (2) compute credible intervals around ISDs for clinicians to assess the model's confidence in its prediction. Our BNN-ISD model utilized sparsity-inducing priors to learn a sparse set of weights to enable feature selection. We provide empirical evidence, on 2 synthetic and 3 real-world clinical datasets, that BNN-ISD system can effectively select meaningful features and compute trustworthy credible intervals of the survival distribution for each patient. We observed that our approach accurately recovers feature importance in the synthetic datasets and selects meaningful features for the real-world clinical data as well, while also achieving state-of-the-art survival prediction performance. We also show that these credible regions can aid in clinical decision-making by providing a gauge of the uncertainty of the estimated ISD curves.

Private Continuous Survival Analysis with Distributed Multi-Site Data

Effective disease surveillance systems require large-scale epidemiological data to improve health outcomes and quality of care for the general population. As data may be limited within a single site, multi-site data (e.g., from a number of local/regional health systems) need to be considered. Leveraging distributed data across multiple sites for epidemiological analysis poses significant challenges. Due to the sensitive nature of epidemiological data, it is imperative to design distributed solutions that provide strong privacy protections. Current privacy solutions often assume a central site, which is responsible for aggregating the distributed data and applying privacy protection before sharing the results (e.g., aggregation via secure primitives and differential privacy for sharing aggregate results). However, identifying such a central site may be difficult in practice and relying on a central site may introduce potential vulnerabilities (e.g., single point of failure). Furthermore, to support clinical interventions and inform policy decisions in a timely manner, epidemiological analysis need to reflect dynamic changes in the data. Yet, existing distributed privacy-protecting approaches were largely designed for static data (e.g., one-time data sharing) and cannot fulfill dynamic data requirements. In this work, we propose a privacy-protecting approach that supports the sharing of dynamic epidemiological analysis and provides strong privacy protection in a decentralized manner. We apply our solution in continuous survival analysis using the Kaplan-Meier estimation model while providing differential privacy protection. Our evaluations on a real dataset containing COVID-19 cases show that our method provides highly usable results.

Predictive function of tumor burden-incorporated machine-learning algorithms for overall survival and their value in guiding management decisions in patients with locally advanced nasopharyngeal carcinoma

Objective

Accurate prognostic predictions and personalized decision-making on induction chemotherapy (IC) for individuals with locally advanced nasopharyngeal carcinoma (LA-NPC) remain challenging. This research examined the predictive function of tumor burden-incorporated machine-learning algorithms for overall survival (OS) and their value in guiding treatment in patients with LA-NPC.

Methods

Individuals with LA-NPC were reviewed retrospectively. Tumor burden signature-based OS prediction models were established using a nomogram and two machine-learning methods, the interpretable eXtreme Gradient Boosting (XGBoost) risk prediction model, and DeepHit time-to-event neural network. The models' prediction performances were compared using the concordance index (C-index) and the area under the curve (AUC). The patients were divided into two cohorts based on the risk predictions of the most successful model. The efficacy of IC combined with concurrent chemoradiotherapy was compared to that of chemoradiotherapy alone.

Results

The 1 221 eligible individuals, assigned to the training (n = 813) or validation (n = 408) set, showed significant respective differences in the C-indices of the XGBoost, DeepHit, and nomogram models (0.849 and 0.768, 0.811 and 0.767, 0.730 and 0.705). The training and validation sets had larger AUCs in the XGBoost and DeepHit models than the nomogram model in predicting OS (0.881 and 0.760, 0.845 and 0.776, and 0.764 and 0.729, P < 0.001). IC presented survival benefits in the XGBoost-derived high-risk but not low-risk group.

Conclusion

This research used machine-learning algorithms to create and verify a comprehensive model integrating tumor burden with clinical variables to predict OS and determine which patients will most likely gain from IC. This model could be valuable for delivering patient counseling and conducting clinical evaluations.

Comparison of MRI radiomics-based machine learning survival models in predicting prognosis of glioblastoma multiforme

OBJECTIVE

To compare the performance of radiomics-based machine learning survival models in predicting the prognosis of glioblastoma multiforme (GBM) patients.

METHODS

131 GBM patients were included in our study. The traditional Cox proportional-hazards (CoxPH) model and four machine learning models (SurvivalTree, Random survival forest (RSF), DeepSurv, DeepHit) were constructed, and the performance of the five models was evaluated using the C-index.

RESULTS

After the screening, 1792 radiomics features were obtained. Seven radiomics features with the strongest relationship with prognosis were obtained following the application of the least absolute shrinkage and selection operator (LASSO) regression. The CoxPH model demonstrated that age (HR = 1.576, p = 0.037), Karnofsky performance status (KPS) score (HR = 1.890, p = 0.006), radiomics risk score (HR = 3.497, p = 0.001), and radiomics risk level (HR = 1.572, p = 0.043) were associated with poorer prognosis. The DeepSurv model performed the best among the five models, obtaining C-index of 0.882 and 0.732 for the training and test set, respectively. The performances of the other four models were lower: CoxPH (0.663 training set / 0.635 test set), SurvivalTree (0.702/0.655), RSF (0.735/0.667), DeepHit (0.608/0.560).

CONCLUSION

This study confirmed the superior performance of deep learning algorithms based on radiomics relative to the traditional method in predicting the overall survival of GBM patients; specifically, the DeepSurv model showed the best predictive ability.

Prognostic significance of cyclin D1 expression pattern in HPV-negative oral and oropharyngeal carcinoma: A deep-learning approach

BACKGROUND

We aimed to establish image recognition and survival prediction models using a novel scoring system of cyclin D1 expression pattern in patients with human papillomavirus-negative oral or oropharyngeal squamous cell carcinoma.

METHODS

The clinicopathological data of 610 patients with human papillomavirus-negative oral/oropharyngeal squamous cell carcinoma were analyzed retrospectively. Cox univariate and multivariate risk regression analyses were performed to compare cyclin D1 expression pattern scoring with the traditional scoring method-cyclin D1 expression level scoring-in relation to patients' overall and progression-free survival. An image recognition model employing the cyclin D1 expression pattern scoring system was established by YOLOv5 algorithms. From this model, two independent survival prediction models were established using the DeepHit and DeepSurv algorithms.

RESULTS

Cyclin D1 had three expression patterns in oral and oropharyngeal squamous cell carcinoma cancer nests. Superior to cyclin D1 expression level scoring, cyclin D1 expression pattern scoring was significantly correlated with the prognosis of patients with oral squamous cell carcinoma (p < 0.0001) and oropharyngeal squamous cell carcinoma (p < 0.05). Moreover, it was an independent prognostic risk factor in both oral squamous cell carcinoma (p < 0.0001) and oropharyngeal squamous cell carcinoma (p < 0.05). The cyclin D1 expression pattern-derived image recognition model showed an average test set accuracy of 78.48% ± 4.31%. In the overall survival prediction models, the average concordance indices of the test sets established by DeepSurv and DeepHit were 0.71 ± 0.02 and 0.70 ± 0.01, respectively.

CONCLUSION

Combined with the image recognition model of the cyclin D1 expression pattern, the survival prediction model had a relatively good prediction effect on the overall survival prognosis of patients with human papillomavirus-negative oral or oropharyngeal squamous cell carcinoma.

Bridging the Worlds of Pharmacometrics and Machine Learning

Precision medicine requires individualized modeling of disease and drug dynamics, with machine learning-based computational techniques gaining increasing popularity. The complexity of either field, however, makes current pharmacological problems opaque to machine learning practitioners, and state-of-the-art machine learning methods inaccessible to pharmacometricians. To help bridge the two worlds, we provide an introduction to current problems and techniques in pharmacometrics that ranges from pharmacokinetic and pharmacodynamic modeling to pharmacometric simulations, model-informed precision dosing, and systems pharmacology, and review some of the machine learning approaches to address them. We hope this would facilitate collaboration between experts, with complementary strengths of principled pharmacometric modeling and flexibility of machine learning leading to synergistic effects in pharmacological applications.

Incorporating longitudinal history of risk factors into atherosclerotic cardiovascular disease risk prediction using deep learning

Background

It is increasingly clear that longitudinal risk factor levels and trajectories are related to risk for atherosclerotic cardiovascular disease (ASCVD) above and beyond single measures. Currently used in clinical care, the Pooled Cohort Equations (PCE) are based on regression methods that predict ASCVD risk based on cross-sectional risk factor levels. Deep learning (DL) models have been developed to incorporate longitudinal data for risk prediction but its benefit for ASCVD risk prediction relative to the traditional Pooled Cohort Equations (PCE) remain unknown.

Objective

To develop a ASCVD risk prediction model that incorporates longitudinal risk factors using deep learning.

Methods

Our study included 15,565 participants from four cardiovascular disease cohorts free of baseline ASCVD who were followed for adjudicated ASCVD. Ten-year ASCVD risk was calculated in the training set using our benchmark, the PCE, and a longitudinal DL model, Dynamic-DeepHit. Predictors included those incorporated in the PCE: sex, race, age, total cholesterol, high density lipid cholesterol, systolic and diastolic blood pressure, diabetes, hypertension treatment and smoking. The discrimination and calibration performance of the two models were evaluated in an overall hold-out testing dataset.

Results

Of the 15,565 participants in our dataset, 2,170 (13.9%) developed ASCVD. The performance of the longitudinal DL model that incorporated 8 years of longitudinal risk factor data improved upon that of the PCE [AUROC: 0.815 (CI: 0.782-0.844) vs 0.792 (CI: 0.760-0.825)] and the net reclassification index was 0.385. The brier score for the DL model was 0.0514 compared with 0.0542 in the PCE.

Conclusion

Incorporating longitudinal risk factors in ASCVD risk prediction using DL can improve model discrimination and calibration.