Abstract:Survival analysis is a widely known method for predicting the likelihood of an event over time. The challenge of dealing with censored samples still remains. Traditional methods, such as the Cox Proportional Hazards (CPH) model, hinge on the limitations due to the strong assumptions of proportional hazards and the predetermined relationships between covariates. The rise of models based on deep neural networks (DNNs) has demonstrated enhanced effectiveness in survival analysis. This research introduces the Implicit Continuous-Time Survival Function (ICTSurF), built on a continuous-time survival model, and constructs survival distribution through implicit representation. As a result, our method is capable of accepting inputs in continuous-time space and producing survival probabilities in continuous-time space, independent of neural network architecture. Comparative assessments with existing methods underscore the high competitiveness of our proposed approach. Our implementation of ICTSurF is available at <a class="link-external link-https" href="https://github.com/44REAM/ICTSurF" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

This paper attempts to solve several key problems in survival analysis: 1. **Improving time precision**: Existing survival analysis models based on deep neural networks (DNNs) usually require specific architectures or strict discretization schemes, which limit the time precision of input and output values. The method proposed in this paper aims to overcome this limitation through the Continuous - Time Survival Function (CTSF), allowing the model to accept inputs and generate survival probabilities in a continuous time space. 2. **Enhancing flexibility**: Traditional survival analysis methods such as the Cox Proportional Hazards model (CPH) rely on fixed time points for discretization processing, which limits the flexibility of the model for data. The Implicit Continuous - Time Survival Function (ICTSurF) proposed in this paper can flexibly handle data at different time points without the need for predefined specific time points. 3. **Improving performance**: Although existing discrete - time survival models can overcome the CPH assumption, they still rely on predefined time intervals when calculating survival probabilities. By introducing an implicit representation method, ICTSurF can estimate survival probabilities more accurately without relying on the neural network architecture and shows better performance with fewer discretization points. ### Specific problem descriptions - **Time precision problem**: Due to discretization processing, traditional methods cannot accurately capture slight changes in time, resulting in less accurate prediction results. - **Flexibility problem**: Existing methods usually require predefined time intervals and cannot flexibly handle data of different samples at different time points. - **Flexibility problem**: Existing methods may not perform as expected when dealing with complex data, especially in the case of a small number of quantization time points. ### Solutions This paper proposes a new DNN - based survival analysis model - ICTSurF, and its main contributions include: 1. **Introducing the ICTSurF model**: This model can calculate the hazard rate at any given time point without requiring fixed, predefined time points as input data, allowing the model to receive accurate inputs. 2. **Learning nonlinear interactions**: Through DNN, ICTSurF can learn the nonlinear relationships between time and other covariates without relying on strong parametric assumptions. 3. **Comparative experiments**: Experiments are carried out using real - world and simulated datasets to verify the advantages of ICTSurF over existing models. 4. **Exploring discretization schemes**: Two discretization schemes (individual discretization and unified discretization) are studied to explore their impact on model performance, and it is found that the flexible discretization scheme is superior to the rigid discretization scheme. ### Summary The main objective of this paper is to solve the time precision, flexibility, and performance problems existing in current survival analysis methods by introducing the ICTSurF model, thereby providing a more efficient and flexible new tool for the field of survival analysis.

ICTSurF: Implicit Continuous-Time Survival Functions with Neural Networks

Learning Survival Distribution with Implicit Survival Function

Copula-based Semiparametric Nonnormal Transformed Linear Model for Survival Data with Dependent Censoring

SurvCORN: Survival Analysis with Conditional Ordinal Ranking Neural Network

Adaptive Transformer Modelling of Density Function for Nonparametric Survival Analysis

Case-Base Neural Networks: survival analysis with time-varying, higher-order interactions

ResDeepSurv: A Survival Model for Deep Neural Networks Based on Residual Blocks and Self-attention Mechanism

DNNSurv: Deep Neural Networks for Survival Analysis Using Pseudo Values

Deep Neural Networks for Survival Analysis Using Pseudo Values

deepAFT: A nonlinear accelerated failure time model with artificial neural network

Deep Recurrent Survival Analysis

Spatio-temporally smoothed deep survival neural network

Neural interval-censored survival regression with feature selection

Deep Generative Survival Analysis: Nonparametric Estimation of Conditional Survival Function

Copula-Based Deep Survival Models for Dependent Censoring

Pseudo value-based Deep Neural Networks for Multi-state Survival Analysis

HACSurv: A Hierarchical Copula-based Approach for Survival Analysis with Dependent Competing Risks

A Simulation Study to Compare the Predictive Performance of Survival Neural Networks with Cox Models for Clinical Trial Data

Siamese Survival Analysis with Competing Risks

Neural Topic Models with Survival Supervision: Jointly Predicting Time-to-Event Outcomes and Learning How Clinical Features Relate

FastSurvival: Hidden Computational Blessings in Training Cox Proportional Hazards Models