The objective of WP1 is to develop artificial intelligence methods and algorithms, control and robotic solutions designed to enable SPM systems to operate autonomously with minimal human intervention for life sciences and medical applications. Developments performed so far include an AFM software tool add-on to integrate autonomous machine learning-based segmentation algorithms applicable to optical microscopy images of cells, to autonomously define regions of interest to be scanned by the AFM system. Moreover, a multimodal simulation tool has been developed to generate synthetic topographic and mechanical images to be used in the training of neural networks to autonomously segment multimodal AFM images and define automatically the regions of interest. Finally, progress has been made towards developing a neural network algorithm able to classify the quality of force curves for using in the automatic definition of optimal acquisition-setting parameters in the imaging of living cells. On the other hand, a segmentation algorithm has been developed to identify isolated DNA and protein molecules in AFM topographic images. Moreover, developments into enhanced nanomechanical and nanoelectrical imaging of biomolecules have been reported, as well as control routines for fast and smart molecular imaging.
The objective of WP2 is to develop artificial intelligence methods and algorithms for the stream and batch processing of SPM data to produce topographic, nanomechanical, and nanoelectrical images of cells and biomolecules, and to develop AI algorithms for biomedical applications (protein DNA-binding, lipid nanoparticle characterization, lipid nanoparticle-cell tracking, cancer diagnostics) based on multiparametric SPM4.0 data. Development performed so far includes AI-enhanced data processing methods for the analysis of force curve maps to produce nanomechanical and nanoelectrical maps and to analyze topographic images to segment automatically DNA-protein complexes. Developments towards AI-assisted identification of soft nanoparticles in AFM topographic images have been reported, as well as building data sets of normal and cancerous cells to be used in the training of neural networks to classify cancerous and normal cells. Finally, an algorithm to generate synthetic force curves is under development for use in training of advanced neural networks applied to force curve maps.
The objective of WP3 is to integrate and interface the AI algorithms and robotic control developed in WP1 and WP2 to build autonomous multiparametric functional SPM systems powered by AI (SPM4.0) for applications in cell and molecular biology and medicine. To develop a user-friendly software toolbox for batch post-processing of large sets of multiparametric SPM4.0 images for biomedical applications. WP3 has just started its activities, and the main result refers to the development of an automated software tool for batch processing of force curve maps acquired on cells.
The objective of WP4 is to develop applications of the novel SPM4.0s and software toolboxes powered by artificial intelligence to open problems in Life Sciences and Medicine (DNA-protein binding site determination, high-throughput lipid nanoparticle multiparametric characterization, label-free nanoparticle tracking in cells, digital cancer diagnostics, multiscale elastography). The activities on this WP have centered for the moment on the development of sample preparation protocols to be used in the different applications.
