In the context of biomedical engineering education, integrating artificial intelligence with mathematical modelling offers a promising path for enhancing students’ understanding of complex biological phenomena. This contribution presents an educational practice designed to introduce undergraduate students to Physics-Informed Neural Networks (PINNs), with a focus on their applications in biomedical modelling. The activity was part of a computer lab session within a continuum biomechanics course (2nd year of Biomedical Engineering degree) and consisted of two main phases.

In the first phase, students were guided through an existing PINN implementation in MATLAB that solved the Fisher equation—a partial differential equation commonly used to model tumour growth dynamics. This example enabled students to explore how deep learning frameworks can incorporate physical laws, and to understand the encoding of initial and boundary conditions, loss functions, and neural network architectures tailored to biomedical problems.

The second phase involved a simplified yet conceptually rich extension: adapting the original PINN code to solve the ordinary differential equation of a damped harmonic oscillator. By reducing the spatial complexity and focusing only on the temporal evolution of the system, students were able to reinforce their understanding of the PINN methodology in a more tractable setting.

The activity was well-received and proved effective in fostering interdisciplinary learning. Students not only improved their programming and modelling skills, but also developed an appreciation for the integration of data-driven and physics-based approaches in biomedical applications. This hands-on experience demonstrates the pedagogical value of PINNs as a bridge between traditional numerical methods and modern machine learning tools, highlighting their potential in the training of future biomedical engineers.

Keywords: Physics-Informed Neural Networks (PINNs), Biomedical Engineering, MATLAB programming.