Survival analysis is a critical framework for modeling event time data, with applications in medicine, reliability, and economics. While machine learning has advanced survival models, challenges persist in balancing predictive accuracy with interpretability, especially for censored data and time-dependent predictions. Concept-based machine learning (CBL), which uses high-level concepts to improve performance and interpretability in classification and regression, has yet to be applied to survival analysis. This paper proposes a novel integration of CBL with survival analysis, enabling the prediction of survival functions while addressing censored data. The proposed method clusters data embeddings and computes probabilistic concept representations to enhance survival predictions. This approach improves both accuracy and interpretability, offering a powerful tool for analyzing time-to-event data in critical domains.