Abstract:Time-to-event analysis provides insights into clinical prognosis and treatment recommendations. However, this task is more challenging than standard regression problems due to the presence of censored observations. Additionally, the lack of confidence assessment, model robustness, and prediction calibration raises concerns about the reliability of predictions. To address these challenges, we propose an evidential regression model specifically designed for time-to-event prediction. The proposed model quantifies both epistemic and aleatory uncertainties using Gaussian Random Fuzzy Numbers and belief functions, providing clinicians with uncertainty-aware survival time predictions. The model is trained by minimizing a generalized negative log-likelihood function accounting for data censoring. Experimental evaluations using simulated datasets with different data distributions and censoring conditions, as well as real-world datasets across diverse clinical applications, demonstrate that our model delivers both accurate and reliable performance, outperforming state-of-the-art methods. These results highlight the potential of our approach for enhancing clinical decision-making in survival analysis.

What problem does this paper attempt to address?

This paper attempts to solve several key problems in time - to - event prediction, which are of great significance in clinical prognosis and treatment recommendations. Specifically, the paper mainly focuses on the following aspects: 1. **Existence of censored data**: One of the main challenges in time - to - event analysis is the existence of censored observations. Censoring means that the event time of some samples is unknown or not fully known, which makes it difficult to estimate the survival probability. 2. **Lack of confidence assessment, model robustness and prediction calibration**: Existing methods often lack confidence assessment when providing predictions, resulting in insufficient reliability and robustness of the model. In addition, the calibration problem of prediction results also affects the credibility of the model in practical applications. 3. **Quantification of uncertainty**: When dealing with censored data, how to effectively quantify the two main types of uncertainty (i.e., epistemic uncertainty and aleatory uncertainty) is an important issue. These uncertainties will affect the accuracy and reliability of prediction. To address the above challenges, the author proposes an evidence - based time - to - event prediction model with the following features: - **Using Gaussian Random Fuzzy Numbers (GRFNs) and belief functions to quantify epistemic uncertainty and aleatory uncertainty**. This method can not only provide survival time prediction with uncertainty, but also provide more reliable decision - support for clinicians. - **Trained by minimizing the generalized negative log - likelihood function**, which takes into account the censoring of data, thereby improving the adaptability and accuracy of the model. - **Experimental evaluation** shows that the model performs well on both simulated datasets and real - world datasets, outperforming the existing state - of - the - art methods. The experiments cover the performance under different data distributions and censoring conditions, demonstrating the superiority and wide applicability of this method. In conclusion, this paper aims to solve the problems of censored data processing, prediction calibration and uncertainty quantification in existing methods by introducing a new evidence - based time - to - event prediction model, thereby improving the support ability for clinical decision - making.

Evidential time-to-event prediction with calibrated uncertainty quantification

An evidential time-to-event prediction model based on Gaussian random fuzzy numbers

Toward Conditional Distribution Calibration in Survival Prediction

Uncertainty quantification for intervals

Predicting disease risks by matching quantiles estimation for censored data.

A Threshold-free Prospective Prediction Accuracy Measure for Censored Time to Event Data

Uncertainty-Aware Optimal Treatment Selection for Clinical Time Series

Uncertainty Quantification and Interpretability for Clinical Trial Approval Prediction

Assessing quantile prediction with censored quantile regression models

Enhancing Uncertainty Quantification in Drug Discovery with Censored Regression Labels

Uncertainty Quantification on Clinical Trial Outcome Prediction

Transformation-Invariant Learning of Optimal Individualized Decision Rules with Time-to-Event Outcomes

[In memory of A. M. Dogliotti].

EsurvFusion: An evidential multimodal survival fusion model based on Gaussian random fuzzy numbers

Uncertainty Regularized Evidential Regression

A personalized Uncertainty Quantification framework for patient survival models: estimating individual uncertainty of patients with metastatic brain tumors in the absence of ground truth

Decision-Focused Uncertainty Quantification

Risk Projection for Time-to-Event Outcome Leveraging Summary Statistics With Source Individual-Level Data

Estimation and Confidence Bands of A Conditional Survival Function with Censoring Indicators Missing at Random

Inference for High-Dimensional Censored Quantile Regression

Uncertainty Quantification of Surrogate Models using Conformal Prediction