Deep Survival Model Research Articles

Learning to Train and to Explain a Deep Survival Model with Large-Scale Ovarian Cancer Transcriptomic Data

Background/Objectives: Ovarian cancer is a complex disease with poor outcomes that affects women worldwide. The lack of successful therapeutic options for this malignancy has led to the need to identify novel biomarkers for patient stratification. Here, we aim to develop the outcome predictors based on the gene expression data as they may serve to identify categories of patients who are more likely to respond to certain therapies. Methods: We used The Cancer Genome Atlas (TCGA) ovarian cancer transcriptomic data from 372 patients and approximately 16,600 genes to train and evaluate the deep learning survival models. In addition, we collected an in-house validation dataset of 12 patients to assess the performance of the trained survival models for their direct use in clinical practice. Despite deceptive generalization capabilities, we demonstrated how our model can be interpreted to uncover biological processes associated with survival. We calculated the contributions of the input genes to the output of the best trained model and derived the corresponding molecular pathways. Results: These pathways allowed us to stratify the TCGA patients into high-risk and low-risk groups (p-value 0.025). We validated the stratification ability of the identified pathways on the in-house dataset consisting of 12 patients (p-value 0.229) and on the external clinical and molecular dataset consisting of 274 patients (p-value 0.006). Conclusions: The deep learning-based models for survival prediction with RNA-seq data could be used to detect and interpret the gene-sets associated with survival in ovarian cancer patients and open a new avenue for future research.

Deep survival analysis for interpretable time-varying prediction of preeclampsia risk

ObjectiveSurvival analysis is widely utilized in healthcare to predict the timing of disease onset. Traditional methods of survival analysis are usually based on Cox Proportional Hazards model and assume proportional risk for all subjects. However, this assumption is rarely true for most diseases, as the underlying factors have complex, non-linear, and time-varying relationships. This concern is especially relevant for pregnancy, where the risk for pregnancy-related complications, such as preeclampsia, varies across gestation. Recently, deep learning survival models have shown promise in addressing the limitations of classical models, as the novel models allow for non-proportional risk handling, capturing nonlinear relationships, and navigating complex temporal dynamics. MethodsWe present a methodology to model the temporal risk of preeclampsia during pregnancy and investigate the associated clinical risk factors. We utilized a retrospective dataset including 66,425 pregnant individuals who delivered in two tertiary care centers from 2015 to 2023. We modeled the preeclampsia risk by modifying DeepHit, a deep survival model, which leverages neural network architecture to capture time-varying relationships between covariates in pregnancy. We applied time series k-means clustering to DeepHit’s normalized output and investigated interpretability using Shapley values. ResultsWe demonstrate that DeepHit can effectively handle high-dimensional data and evolving risk hazards over time with performance similar to the Cox Proportional Hazards model, achieving an area under the curve (AUC) of 0.78 for both models. The deep survival model outperformed traditional methodology by identifying time-varied risk trajectories for preeclampsia, providing insights for early and individualized intervention. K-means clustering resulted in patients delineating into low-risk, early-onset, and late-onset preeclampsia groups—notably, each of those has distinct risk factors. ConclusionThis work demonstrates a novel application of deep survival analysis in time-varying prediction of preeclampsia risk. Our results highlight the advantage of deep survival models compared to Cox Proportional Hazards models in providing personalized risk trajectory and demonstrating the potential of deep survival models to generate interpretable and meaningful clinical applications in medicine.

Deep Survival Models Can Improve Long-Term Mortality Risk Estimates from Chest Radiographs

Deep learning has recently demonstrated the ability to predict long-term patient risk and its stratification when trained on imaging data such as chest radiographs. However, existing methods formulate estimating patient risk as a binary classification, typically ignoring or limiting the use of temporal information, and not accounting for the loss of patient follow-up, which reduces the fidelity of estimation and limits the prediction to a certain time horizon. In this paper, we demonstrate that deep survival and time-to-event prediction models can outperform binary classifiers at predicting mortality and risk of adverse health events. In our study, deep survival models were trained to predict risk scores from chest radiographs and patient demographic information in the Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer screening trial (25,433 patient data points used in this paper) for 2-, 5-, and 10-year time horizons. Binary classification models that predict mortality at these time horizons were built as baselines. Compared to the considered alternative, deep survival models improve the Brier score (5-year: 0.0455 [95% CI, 0.0427–0.0482] vs. 0.0555 [95% CI, (0.0535–0.0575)], p < 0.05) and expected calibration error (ECE) (5-year: 0.0110 [95% CI, 0.0080–0.0141] vs. 0.0747 [95% CI, 0.0718–0.0776], p < 0.05) for those fixed time horizons and are able to generate predictions for any time horizon, without the need to retrain the models. Our study suggests that deep survival analysis tools can outperform binary classification in terms of both discriminative performance and calibration, offering a potentially plausible solution for forecasting risk in clinical practice.

Predicting Adverse Outcomes Following Catheter Ablation Treatment for Atrial Flutter/Fibrillation

To develop prognostic survival models for predicting adverse outcomes after catheter ablation treatment for non-valvular atrial fibrillation (AF) and/or atrial flutter (AFL). We used a linked dataset including hospital administrative data, prescription medicine claims, emergency department presentations, and death registrations of patients in New South Wales, Australia. The cohort included patients who received catheter ablation for AF and/or AFL. Traditional and deep survival models were trained to predict major bleeding events and a composite of heart failure, stroke, cardiac arrest, and death. Out of a total of 3,285 patients in the cohort, 177 (5.3%) experienced the composite outcome-heart failure, stroke, cardiac arrest, death-and 167 (5.1%) experienced major bleeding events after catheter ablation treatment. Models predicting the composite outcome had high-risk discrimination accuracy, with the best model having a concordance index >0.79 at the evaluated time horizons. Models for predicting major bleeding events had poor risk discrimination performance, with all models having a concordance index <0.66. The most impactful features for the models predicting higher risk were comorbidities indicative of poor health, older age, and therapies commonly used in sicker patients to treat heart failure and AF and AFL. Diagnosis and medication history did not contain sufficient information for precise risk prediction of experiencing major bleeding events. Predicting the composite outcome yielded promising results, but future research is needed to validate the usefulness of these models in clinical practice. Machine learning models for predicting the composite outcome have the potential to enable clinicians to identify and manage high-risk patients following catheter ablation for AF and AFL proactively.

Causal debiasing for unknown bias in histopathology-A colon cancer use case.

Advancement of AI has opened new possibility for accurate diagnosis and prognosis using digital histopathology slides which not only saves hours of expert effort but also makes the estimation more standardized and accurate. However, preserving the AI model performance on the external sites is an extremely challenging problem in the histopathology domain which is primarily due to the difference in data acquisition and/or sampling bias. Although, AI models can also learn spurious correlation, they provide unequal performance across validation population. While it is crucial to detect and remove the bias from the AI model before the clinical application, the cause of the bias is often unknown. We proposed a Causal Survival model that can reduce the effect of unknown bias by leveraging the causal reasoning framework. We use the model to predict recurrence-free survival for the colorectal cancer patients using quantitative histopathology features from seven geographically distributed sites and achieve equalized performance compared to the baseline traditional Cox Proportional Hazards and DeepSurvival model. Through ablation study, we demonstrated benefit of novel addition of latent probability adjustment and auxiliary losses. Although detection of cause of unknown bias is unsolved, we proposed a causal debiasing solution to reduce the bias and improve the AI model generalizibility on the histopathology domain across sites. Open-source codebase for the model training can be accessed from https://github.com/ramon349/fair_survival.git.

Imaging genetics based dementia risk prediction using deep survival neural network

AbstractBackgroundDementia has a large impact on people’s lives and is one of the leading causes of death. Therefore, accurate risk assessment is essential for improving early diagnostics and intervention. Previous studies have proposed various models for dementia risk prediction, based on imaging biomarkers and genetic factors. However, most of them have linear, unimodality or low‐dimensional nature due to model limitations and computational challenges of integrating such heterogeneous and high‐dimensional domains as images and genotype data. Recent development of Artificial Intelligence, specifically neural networks, enabled such integration, yet focused on classification problems, not risk prediction. We present a deep survival neural network to meet the purpose.MethodData were obtained from the Rotterdam Study. Individuals with complete data and available brain MRI before dementia diagnosis were included. MRI scans were pre‐processed using voxel‐based morphometry according to Good et al. into gray matter images. We included ApoE‐ε4 and 76 top‐risk SNPs based on most recently published GWAS (Bellenguez et al.). We developed the model combining Convolutional Neural Network and Survival Analysis, with the maximum‐likelihood estimator and C‐index of cox proportional hazard (Cox) model as the loss function and metric. To evaluate model performances, we used Cox model as comparisons: one baseline model with gender, age, genetic inputs and one extended model with additional extracted brain measurements. We performed 10 cross‐validation by stratified sampling to reveal the true differences.ResultTable1 shows little performance differences among the models, as expected age is already the strongest predictor. However, we see that raw gray matter and genetics inputs contribute to prediction additionally in all models. Furthermore, using the method from Tsang et al. we show strong interactions among image, age and genetic features (Fig.1). Fig.2 visualizes the most important brain regions for prediction on 3D image using GRAD‐CAM.ConclusionWe developed a deep survival model for dementia risk prediction which do not outperform classical model with brain measurements at present. However, our model is able to integrate high‐dimensional MRI and genetic data, detect and visualize the importance and interactions among input data, which may give additional insight for dementia prediction and etiologic research.

SwarmDeepSurv: swarm intelligence advances deep survival network for prognostic radiomics signatures in four solid cancers

Survival models exist to study relationships between biomarkers and treatment effects. Deep learning-powered survival models supersede the classical Cox proportional hazards (CoxPH) model, but substantial performance drops were observed on high-dimensional features because of irrelevant/redundant information. To fill this gap, we proposed SwarmDeepSurv by integrating swarm intelligence algorithms with the deep survival model. Furthermore, four objective functions were designed to optimize prognostic prediction while regularizing selected feature numbers. When testing on multicenter sets (n= 1,058) of four different cancer types, SwarmDeepSurv was less prone to overfitting and achieved optimal patient risk stratification compared with popular survival modeling algorithms. Strikingly, SwarmDeepSurv selected different features compared with classical feature selection algorithms, including the least absolute shrinkage and selection operator (LASSO), with nearly no feature overlapping across these models. Taken together, SwarmDeepSurv offers an alternative approach to model relationships between radiomics features and survival endpoints, which can further extend to study other input data types including genomics.

Leveraging deep survival models to predict quality of care risk in diverse hospital readmissions

Hospital readmissions rate is reportedly high and has caused huge financial burden on health care systems in many countries. It is viewed as an important indicator of health care providers’ quality of care. We examine the use of machine learning-based survival analysis to assess quality of care risk in hospital readmissions. This study applies various survival models to explore the risk of hospital readmissions given patient demographics and their respective hospital discharges extracted from a health care claims dataset. We explore advanced feature representation techniques such as BioBERT and Node2Vec to encode high-dimensional diagnosis code features. To our knowledge, this study is the first to apply deep-learning based survival-analysis models for predicting hospital readmission risk agnostic of specific medical conditions and a fixed window for readmission. We found that modeling the time from discharge date to readmission date as a Weibull distribution as in the SparseDeepWeiSurv model yields the best discriminative power and calibration. In addition, embedding representations of the diagnosis codes do not contribute to improvement in model performance. We find dependency of each model’s performance on the time point at which it is evaluated. This time dependency of the models’ performance on the health care claims data may necessitate a different choice of model in quality of care issue detection at different points in time. We show the effectiveness of deep-learning based survival-analysis models in estimating the quality of care risk in hospital readmissions.

Mitigating Membership Inference in Deep Survival Analyses with Differential Privacy.

Deep neural networks have been increasingly integrated in healthcare applications to enable accurate predicative analyses. Sharing trained deep models not only facilitates knowledge integration in collaborative research efforts but also enables equitable access to computational intelligence. However, recent studies have shown that an adversary may leverage a shared model to learn the participation of a target individual in the training set. In this work, we investigate privacy-protecting model sharing for survival studies. Specifically, we pose three research questions. (1) Do deep survival models leak membership information? (2) How effective is differential privacy in defending against membership inference in deep survival analyses? (3) Are there other effects of differential privacy on deep survival analyses? Our study assesses the membership leakage in emerging deep survival models and develops differentially private training procedures to provide rigorous privacy protection. The experimental results show that deep survival models leak membership information and our approach effectively reduces membership inference risks. The results also show that differential privacy introduces a limited performance loss, and may improve the model robustness in the presence of noisy data, compared to non-private models.

Deep survival modeling of longitudinal retinal OCT volumes for predicting the onset of atrophy in patients with intermediate AMD.

In patients with age-related macular degeneration (AMD), the risk of progression to late stages is highly heterogeneous, and the prognostic imaging biomarkers remain unclear. We propose a deep survival model to predict the progression towards the late atrophic stage of AMD. The model combines the advantages of survival modelling, accounting for time-to-event and censoring, and the advantages of deep learning, generating prediction from raw 3D OCT scans, without the need for extracting a predefined set of quantitative biomarkers. We demonstrate, in an extensive set of evaluations, based on two large longitudinal datasets with 231 eyes from 121 patients for internal evaluation, and 280 eyes from 140 patients for the external evaluation, that this model improves the risk estimation performance over standard deep learning classification models.

Reverse survival model (RSM): a pipeline for explaining predictions of deep survival models

Reverse survival model (RSM): a pipeline for explaining predictions of deep survival models

DeepMTS: Deep Multi-Task Learning for Survival Prediction in Patients With Advanced Nasopharyngeal Carcinoma Using Pretreatment PET/CT.

Nasopharyngeal Carcinoma (NPC) is a malignant epithelial cancer arising from the nasopharynx. Survival prediction is a major concern for NPC patients, as it provides early prognostic information to plan treatments. Recently, deep survival models based on deep learning have demonstrated the potential to outperform traditional radiomics-based survival prediction models. Deep survival models usually use image patches covering the whole target regions (e.g., nasopharynx for NPC) or containing only segmented tumor regions as the input. However, the models using the whole target regions will also include non-relevant background information, while the models using segmented tumor regions will disregard potentially prognostic information existing out of primary tumors (e.g., local lymph node metastasis and adjacent tissue invasion). In this study, we propose a 3D end-to-end Deep Multi-Task Survival model (DeepMTS) for joint survival prediction and tumor segmentation in advanced NPC from pretreatment PET/CT. Our novelty is the introduction of a hard-sharing segmentation backbone to guide the extraction of local features related to the primary tumors, which reduces the interference from non-relevant background information. In addition, we also introduce a cascaded survival network to capture the prognostic information existing out of primary tumors and further leverage the global tumor information (e.g., tumor size, shape, and locations) derived from the segmentation backbone. Our experiments with two clinical datasets demonstrate that our DeepMTS can consistently outperform traditional radiomics-based survival prediction models and existing deep survival models.

Automatic Tumor Grading on Colorectal Cancer Whole-Slide Images: Semi-Quantitative Gland Formation Percentage and New Indicator Exploration.

Tumor grading is an essential factor for cancer staging and survival prognostication. The widely used the WHO grading system defines the histological grade of CRC adenocarcinoma based on the density of glandular formation on whole-slide images (WSIs). We developed a fully automated approach for stratifying colorectal cancer (CRC) patients’ risk of mortality directly from histology WSI relating to gland formation. A tissue classifier was trained to categorize regions on WSI as glands, stroma, immune cells, background, and other tissues. A gland formation classifier was trained on expert annotations to categorize regions as different degrees of tumor gland formation versus normal tissues. The glandular formation density can thus be estimated using the aforementioned tissue categorization and gland formation information. This estimation was called a semi-quantitative gland formation ratio (SGFR), which was used as a prognostic factor in survival analysis. We evaluated gland formation percentage and validated it by comparing it against the WHO cutoff point. Survival data and gland formation maps were then used to train a spatial pyramid pooling survival network (SPPSN) as a deep survival model. We compared the survival prediction performance of estimated gland formation percentage and the SPPSN deep survival grade and found that the deep survival grade had improved discrimination. A univariable Cox model for survival yielded moderate discrimination with SGFR (c-index 0.62) and deep survival grade (c-index 0.64) in an independent institutional test set. Deep survival grade also showed better discrimination performance in multivariable Cox regression. The deep survival grade significantly increased the c-index of the baseline Cox model in both validation set and external test set, but the inclusion of SGFR can only improve the Cox model less in external test and is unable to improve the Cox model in the validation set.

Analyzing topics in social media for improving digital twinning based product development

Digital twinning enables manufacturers to create digital representations of physical entities, thus implementing virtual simulations for product development. Previous efforts of digital twinning neglect the decisive consumer feedback in product development stages, failing to cover the gap between physical and digital spaces. This work mines real-world consumer feedbacks through social media topics, which is significant to product development. We specifically analyze the prevalent time of a product topic, giving an insight into both consumer attention and the widely-discussed time of a product. The primary body of current studies regards the prevalent time prediction as an accompanying task or assumes the existence of a preset distribution. Therefore, these proposed solutions are either biased in focused objectives and underlying patterns or weak in the capability of generalization towards diverse topics. To this end, this work combines deep learning and survival analysis to predict the prevalent time of topics. We propose a specialized deep survival model which consists of two modules. The first module enriches input covariates by incorporating latent features of the time-varying text, and the second module fully captures the temporal pattern of a rumor by a recurrent network structure. Moreover, a specific loss function different from regular survival models is proposed to achieve a more reasonable prediction. Extensive experiments on real-world datasets demonstrate that our model significantly outperforms the state-of-the-art methods.

Comparison of Performance of Deep Survival and Cox Proportional Hazard Models: an Application on the Lung Cancer Dataset

The goal of this study is to compare the performance of the deep survival model and the Cox regression model in an open-access Lung cancer dataset consisting of survivors and dead patients. In the study, it is applied to an open access dataset named "Lung Cancer Data" to compare the performances of the CPH and deepsurv models. The performance of the models is evaluated by C-index, AUC, and Brier score. The concordance index of the deep survival model is 0.64296, the Brier score was 0.128921, and the AUC was 0.6835. With the Cox regression model, the concordance index is calculated as 0.61445, brier score 0.1667, and AUC 0.5832. According to the Concordance index, brier score, and AUC criteria, the deep survival model performed better than the cox regression model. DeepSurv's forecasting, modeling, and predictive capabilities pave the path for future deep neural network and survival analysis research. DeepSurv has the potential to supplement traditional survival analysis methods and become the standard method for medical doctors to examine and offer individualized treatment alternatives with more research.

Attention-based deep survival model for time series data

In the era of internet of things and Industry 4.0, smart products and manufacturing systems emit signals tracking their operating condition in real-time. Survival analysis shows its strength in modeling such signals to determine the condition of in-service equipment and products to yield critical operational decisions, i.e., maintenance and repair. One appealing aspect of survival analysis is the possibility to include subjects in the model which did not have their failure yet or when the exact failure time is unknown.NN-based survival models, i.e., deep survival models, show superior performance in modeling the non-linear relationship between the reliability function and covariates. We propose a novel deep survival model, seq2surv, to incorporate the seq2seq structure and attention mechanism to enhance the ability to analyze a sequence of signals in the survival analysis. Similar to the seq2seq model which shows superior performance in machine translation, we designed the seq2surv model to translate from a sequence of signals to a sequence of survival probabilities and to update the reliability predictions along with real-time monitoring. Our results show that the seq2surv model outperforms existing deep survival approaches in terms of higher prediction accuracy and lower errors in the survival function estimation on both simulated and real-world datasets.

A deep learning survival model for triple-negative breast cancer patients.

e12594 Background: Triple-negative breast cancer (TNBC) is the subtype of breast cancer with the worst prognosis. There is no reliable model for survival prediction of TNBC patients. The traditional Cox regression analysis with poor prediction power cannot satisfy the clinical needs. The purpose was to establish a deep learning model and develop a new prognostic system for TNBC patients. Methods: This study collected data of TNBC patients from the Surveillance, Epidemiology, and End Results (SEER) program between 2010 and 2016. 70% were used to develop the deep learning model, 15% were used as the validation set, and 15% as the independent testing set. Then the concordance-index (c-index) and Brier score (IBS) were calculated and compared with the Cox regression analysis and random forest. Finally, according to the classification of the deep survival model, an individualized prognosis system was established. Results: A total of 37,818 patients were enrolled in this study. In the validation set, the c-index of the deep learning was 0.799, which was better than the traditional Cox regression model (0.774) and random forest (0.763). The independent testing set further proved the robustness of the deep survival model (c-index 0.788). The new prognosis system based on the deep survival model reached an area under the curve (AUC) of 0.805, which was better than the Tumor, Node, Metastases (TNM) staging system (0.771). Conclusions: Deep learning model had better prediction power than the Cox regression analysis and the random forest. The established prognosis system can better predict prognosis and aid individual risk stratification for TNBC patients patients.

Developing Deep Survival Model for Remaining Useful Life Estimation Based on Convolutional and Long Short-Term Memory Neural Networks

The application of mechanical equipment in manufacturing is becoming more and more complicated with technology development and adoption. In order to keep the high reliability and stability of the production line, reducing the downtime to repair and the frequency of routine maintenance is necessary. Since machine and components’ degradations are inevitable, accurately estimating the remaining useful life of them is crucial. We propose an integrated deep learning approach with convolutional neural networks and long short-term memory networks to learn the latent features and estimate remaining useful life value with deep survival model based on the discrete Weibull distribution. We conduct the turbofan engine degradation simulation dataset from Commercial Modular Aero-Propulsion System Simulation dataset provided by NASA to validate our approach. The improved results have proven that our proposed model can capture the degradation trend of a fault and has superior performance under complex conditions compared with existing state-of-the-art methods. Our study provides an efficient feature extraction scheme and offers a promising prediction approach to make better maintenance strategies.

The effect of elevated triglycerides and purified eicosapentaenoic acid ethyl ester (EPA) for cardiovascular risk reduction in the UK Biobank population

Abstract Elevated plasma triglycerides (TG) are associated with insulin resistance, metabolic syndrome and major adverse cardiovascular events (MACE). Elevated plasma TG (&amp;gt;150 and 200–499 mg/dL) were recently reported to be a predictor of heart failure in patients with coronary heart disease on statin therapy with a respective 19% and 24% risk increase compared to normal TG levels. REDUCE-IT was a major double-blind randomised controlled trial that tested a novel formulation of highly purified eicosapentaenoic acid ethyl ester (EPA) in 8179 patients with elevated TG on statin therapy. In the trial, daily 4g of EPA lowered MACE by an absolute rate of 4.8% compared to a placebo over a median 4.9-year follow-up. This reflected a 25% risk reduction in MACE and a number needed to treat of 21. EPA recently received significant public attention in late 2019 as the FDA has approved it for secondary prevention and high-risk primary prevention of cardiovascular disease (CVD) in patients with elevated TG that match the trial criteria in the USA. We aimed to investigate the risk reduction in MACE from using EPA in the UK population. We used the UK BioBank, a panomic resource following 500,000 participants over &amp;gt;10 years, with similar age and sex adjusted rates of CVD as the UK population. We first calculated the hazard ratios and Kaplan-Meier survival curves by deciles of increasing TG for incidence of combined CV outcomes: stroke, coronary heart disease, and atherosclerosis. Non-linear CoxPH and DeepSurvival models were trained using different variables from &amp;gt;200,000 UK BioBank participants' data. C-index with standard error and confidence interval estimated with bootstrap sampling ensured quality control. We then matched the UK Biobank population with the REDUCE-IT inclusion criteria and estimated the reduction of combined cardiovascular outcomes if EPA was approved in this population. Hazard ratios increased for TG levels to 5.44 between the 10th decile, and the 1st decile TG level, which was used as baseline. 3563 UK Biobank participants matched with the REDUCE-IT criteria. With the assumption that EPA would have the same effect on the UK Biobank population, we estimate that if the participants were taking EPA, only 29% of the risk group versus actual 37% would have suffered an outcome within the UK Biobank follow-up period. This means that 289 less individuals would have suffered an event instead of the recorded 1318, for a total of 1037. That means according to the number needed to treat analysis, 13 patients would need to be treated to prevent 1 patient from experiencing an event. Elevated TG increase risk for CV events in the UK Biobank population. Purified EPA might become an important tool in our arsenal for CVD secondary and primary prevention. Survival Curves &amp; CI of TG for CVD Funding Acknowledgement Type of funding source: None

Prediction of conversion to Alzheimer's disease using deep survival analysis of MRI images.

The prediction of the conversion of healthy individuals and those with mild cognitive impairment to the status of active Alzheimer’s disease is a challenging task. Recently, a survival analysis based upon deep learning was developed to enable predictions regarding the timing of an event in a dataset containing censored data. Here, we investigated whether a deep survival analysis could similarly predict the conversion to Alzheimer’s disease. We selected individuals with mild cognitive impairment and cognitively normal subjects and used the grey matter volumes of brain regions in these subjects as predictive features. We then compared the prediction performances of the traditional standard Cox proportional-hazard model, the DeepHit model and our deep survival model based on a Weibull distribution. Our model achieved a maximum concordance index of 0.835, which was higher than that yielded by the Cox model and comparable to that of the DeepHit model. To our best knowledge, this is the first report to describe the application of a deep survival model to brain magnetic resonance imaging data. Our results demonstrate that this type of analysis could successfully predict the time of an individual’s conversion to Alzheimer’s disease.