Understanding how tumours evolve while evading the immune system remains one of the most complex challenges in modern medicine. Most existing mathematical models rely on deterministic assumptions, using fixed values that rarely reflect real clinical scenarios. In practice, immune responses vary widely among patients, making such models insufficient for personalised predictions.
Introducing a New Framework for Biological Uncertainty
To address this limitation, researchers from the Escuela Superior Politécnica del Litoral (ESPOL) have developed an advanced computational modelling framework that integrates Type-3 Fuzzy Logic with neural networks. This approach enables the simulation of tumour–immune dynamics under conditions of uncertainty and chaotic behaviour. The study detailing this innovation has been published in Information Sciences.
Unlike traditional models that generate a single outcome, the proposed framework captures the inherent variability of biological systems, offering a more realistic representation of patient-specific immune responses.
Capturing Immune Response Delays and Tumour Dynamics
A central focus of the study is the delay, or latency, in cytotoxic T-cell activation, a critical factor in cancer progression. Even small variations in this immune response timing can determine whether a tumour is eliminated or progresses to an aggressive relapse.
Therefore, instead of computing a single trajectory, the model generates bands of uncertainty, visually representing multiple potential disease outcomes under the same treatment conditions. This capability allows clinicians to better appreciate the range of possible responses rather than relying on a single predicted path.
Interpretable Intelligence Beyond Black-Box AI
Importantly, the framework avoids the opacity of conventional “black-box” artificial intelligence systems. By using a logic-oriented and interpretable architecture, the model preserves chaotic structures and bifurcations, which often represent critical tipping points in a patient’s health status.
As a result, clinicians can understand not only the predicted outcome but also the underlying reasoning behind it. Moreover, the model outperformed conventional techniques such as Type-2 fuzzy systems and ANFIS, successfully maintaining the oscillatory behaviour of cancer dynamics even when working with incomplete or noisy data.
Visual Risk Maps for Precision Oncology
As reported by medicalxpress, one of the most impactful outcomes of this research is the development of visual clinical risk maps. Using intuitive linguistic rules—such as “If immune response delay is high and CD8+ cell levels are low, relapse risk is high”—the model classifies patients into safe or high-risk zones.
Consequently, this approach supports effective treatment stratification. Clinicians can identify patients who require immediate intervention to shorten immune activation time, while also recognising those likely to respond well to standard therapy.
Advancing Explainable AI in Clinical Decision-Making
By combining mathematical rigour with Explainable Artificial Intelligence (XAI), this modelling framework bridges the gap between complex computation and clinical usability. Ultimately, it enables more robust biomedical simulations that translate abstract data into clear, actionable insights.
In doing so, the research marks a significant step toward precision oncology, where personalised, transparent, and data-driven tools guide therapeutic decision-making.