Multimodale Beeldvormingsintelligentie voor Combinatorische Screening

Developing new drugs is a slow, costly process, often taking 12 to 15 years and billions of dollars. This research aims to speed up drug discovery by improving how we test drug com pounds on cell models, e.g., organoids, spheroids, and other cell clusters. A major challenge in drug testing is combinatorial screening, which tests from thousands of drug compounds to millions of combinations at once. This is a major time-consuming process in drug discovery. For 10k–100k compounds, the data acquisition only can take days. This study introduces two advanced imaging techniques—Lens-Free Imaging (LFI) and Electrical Impedance Tomography (EIT)—which work together to produce real-time, large-scale images that reveal both the structure and function of samples. The novel LFI-EIT multimodal imaging platform has the potential to replace conventional brightfield-fluorescent microscopy in biomedical use. The main objectives are to make LFI faster for image reconstruction, combine it with EIT for more detailed data, and create an automated system for quicker processing. Key contributions include a new imaging framework for wide-field, real-time views, deep learning models for more accurate image reconstruction, and a system that speeds up analysis. This research aims to make drug testing faster and more informative, ultimately helping to accelerate drug development, especially for diseases that vary among patients. Besides this, we also proposed a sparse neural network for improving EIT reconstruction. As a result, the image capture region (i.e., field of view) can be expanded 20× larger than conventional light microscopes, with sub-pixel resolution. Furthermore, it has the potential to support large-scale microfluidics in such a large region. By integrating these systems, the overall drug screening is expected to be approximately 70-90% shorter. Lastly, the EIT add-on can help the system monitor samples' inner physiological processes faster and more accurately than using toxic fluorescent biomarkers.