In this Proof-of-Concept project, we demonstrated the BED-TEM idea through a software package written in Python. The platform was designed to process microscopy data, identify relevant material changes and support the selection of experimental parameters. Its main achievements were as follows:
(i) A modular software framework was developed in Python, providing the basis for connecting TEM data input, image-processing routines, user-defined criteria and experimental-design functionality. The software demonstrates how specialist analysis methods can be incorporated into a practical interface without requiring users to operate individual command-line tools.
(ii) In situ experiments were implemented using the DENSsolutions Lightning Arctic System. Materials were exposed to controlled stimuli, and the resulting microscopy datasets were investigated to determine whether structural changes could be recognized with sufficient spatial and temporal resolution. These experiments supplied realistic data for assessing the proposed processing workflow.
(iii) Data-processing routines were implemented so that microscopy data could be analysed online during in situ experiments. This represents an important advance over conventional offline processing because experimental observations can, in principle, be used immediately to inform subsequent stimulus values. It establishes the foundation for a closed feedback loop between data acquisition, image analysis, Bayesian experimental design and in situ control.
(iv) The work also revealed areas requiring further development. In particular, robust recognition of phase information remains challenging, especially when image quality, noise or subtle structural differences affect the data. The direct control of the TEM and in situ hardware also requires closer integration, additional validation and improved automation.