Integration of cloud services for storage and processing of cryomicroscopic images: practical experience using MINIO and CVAT

In modern biomedical science, the efficient processing of large volumes of visual data is critically important for analyzing cellular structures. This abstract describes practical experience integrating the MinIO and CVAT cloud services to automate the processes of storage, annotation, and analysis of cryo-microscopy images. The application of these tools enhances the accuracy of cell segmentation, ensures scalability, and improves the reproducibility of research. Cryo-microscopy is a powerful method for visualizing biological samples at the nanoscale. However, processing the resulting images requires significant computational resources and effective tools for data storage and analysis. Integrating cloud services, such as the MinIO for data storage and the CVAT for annotation, optimizes these processes. Cryo-microscopy images were collected from various sources and stored in the MinIO cloud storage, providing reliable and scalable access to the data. The CVAT tool was used for precise delineation of cellular structures. The annotation process involved manual or semi-automatic marking of regions of interest in the images, which is critically important for training artificial intelligence models. The annotated images were prepared for training deep learning models, such as the U-Net and Mask R-CNN, which have proven effective in image segmentation tasks. The models were trained on the annotated data using the TensorFlow and PyTorch libraries. After training, the models were applied to automatic segmentation of new cryo-microscopy images. The inference results were compared with manual annotations to assess the accuracy and reliability of the models. Integration with the Jupyter Notebook enabled researchers to interactively analyze inference results and generate analytical reports. Integrating the MinIO and CVAT cloud services into the cryo-microscopy image processing workflow significantly enhances the efficiency and accuracy of cellular structure analysis. The use of modern technologies and tools facilitates the process automation, ensures scalability and reproducibility of research, which is an important step in advancing biomedical research and improving diagnostics.