Optimisation of Acoustic Manipulation for Single Cell Manipulation

Sammanfattning

In microfluidic systems, resonant acoustic fields enable controlled particle manipulation, a phenomena known as acoustophoresis. These contactless and low-cost processes have successfully been applied to manipulate bulk particle populations, however robust single-cell manipulation for single-cell analysis remains a challenge. This work presents an experimental investigation into optimising the controllability of an acoustofluidic manipulation platform through comparison of different algorithmic control approaches under varying experimental parameters. The study identified voltage as a critical parameter governing a trade-off between manipulation speed and controllability, with an optimal operating regime required to achieve reliable single-cell control. Different control strategies were evaluated using polystyrene particles and living cells. While the state-of-the-art epsilon-greedy algorithm showed limited performance, a novel adaptive approach based on model predictive control achieved a 100% success rate after optimisation, enabling highly controllable single-cell manipulation. In addition, the results highlight the importance of temporally structured actuation for stable particle trajectories and demonstrate robust performance of the proposed approach, even for more complex manipulation tasks. These findings establish the feasibility of employing local optimisation, driven by reinforcement learning, for acoustofluidic single-cell handling and represent a promising step towards advanced single-cell analysis and other biomedical applications. Additionally, as a first step toward future controllability approaches, a field‑characterisation procedure was developed and performed to analyse the modal responses of acoustofluidic devices. The method employed a prototype acoustofluidic platform where numerical simulations were combined with experiments encompassing 1) viewing spatial distributions of tracer particles and 2) laser Doppler vibrometry. The resulting insights allowed characterisation the acoustic behaviour, intend to allow further decision informed development of the platform.