Abstract
Cancer remains one of the most lethal diseases worldwide, which makes its early detection a critical public health priority. The development of early diagnostic systems enables timely identification of individuals with a high risk of developing cancer and supports informed clinical decision-making. In this study, existing approaches to early cancer diagnosis were systematically analyzed. Neural networks were selected as the methodological foundation for designing the expert system. A knowledge base for the expert system was constructed using domain expertise provided by oncology specialists. Key predictors for building a neural-network model were identified. A mathematical model of the system was implemented in the MATLAB environment using three training algorithms: the Levenberg–Marquardt algorithm, Bayesian regularization, and the conjugate-gradient method. A structural design of an expert-system-based early cancer diagnostic system is proposed. Future research will focus on developing a deep-learning model to achieve higher accuracy in forecasting cancer probability.