Some machines can now use artificial intelligence to collect data and adapt to changing environments without human input. Autonomous adaptive data acquisition (AADA) can automate existing data collection techniques, enable new experiments, and shed new insights on collected data.

Holman et al. explore AADA applications in high-dimensional, hyperspectral imaging and propose an accessible, effective, and computationally efficient AADA solution to help scientists achieve non-invasive, label-free, time-resolved hyperspectral imaging of dynamic biological systems.

The paper highlights revolutionary advances in the implementation of autonomous systems in experimental data acquisition, focusing on AADA’s different forms and how to apply AADA to spatiochemically image complex experimental systems. The authors also propose a model for future implementation, with AADA as a modular building block within a much larger intelligent infrastructure.

“The published cases that we cover in our review describe the challenges of practical implementation for AADA, which relate to instrument noise, dimensionality of the parameter space, and the researcher’s definition of the experimental goal,” author Elizabeth Holman said. “We present each case as a combination of basic conceptual parts to clarify our perspective of viewing AADA implementation as a modular solution tailored to one’s experimental needs rather than as a universal solution for AADA needs in every field.”

The team hopes readers will consider how autonomous adaptive approaches can help with data collection, analysis, and interpretation, and communicate and collaborate across fields to develop AADA in a generalizable way that enables the improvement of many disparate studies.

Source: “Towards implementing autonomous adaptive data acquisition for scanning hyperspectral imaging of biological systems,” by Elizabeth A. Holman, Harinarayan Krishnan, Derek R. Holman, Hoi-Ying N. Holman, and Paul W. Sternberg, Applied Physics Reviews (2023). The article can be accessed at