Computational Data Science Approaches for Biomedical Images and Biological Data

Solving real-life problems in different application fields by developing models, algorithms, and software tools to discover, understand, and gain insight into scientific phenomena through analyzing data produced in experiments and simulations, especially biomedical images and biological data, characterized by complexity, heterogeneity, and massive size. Biomedical images visualize human organs with different scales down to the cell resolution. They are essential data for studying diseases, discovering new therapies, and improving human health care. The most used imaging tools rely on X-rays (CT scans), magnetism (MRI), sound (ultrasound), radiopharmaceuticals (SPECT, PET), or light (endoscopy, OCT, optical microscopy). Beyond their specific aim in research studies and diagnostic practices, computational data science approaches are needed to process and analyze huge amount of imaging data for solving problems such as denoising and deblurring, classification, detection, segmentation, lineage tracing, and tracking in different application fields. High throughput experiments produce massive and deeply informative amount of biological data, highly different in complexity, scale and format. The extraction of knowledge from these data in a comprehensive and holistic manner requires to adopt strategies aimed at integrating multi-modal and multi-source data. Omics data (genomics, proteomics, metabolomics, and transcriptomics) are considered by big data sciences, and their integration with multimodal imaging data has improved diagnosis and treatments in complex diseases, such as Alzheimer's and Parkinson's diseases and cancer, leading toward precision medicine. According to system biology, biological data can be organized in structures able to describe the role of each biological factor as part of a complex and highly interconnected system (organism, tissue, cell, disease). For the goal of precision medicine, these structures, alias networks, can refer even to a single patient or context. The principles and methods of graph theory, coupled with the machine and deep learning approaches, allow to read and interpret the richness of information contained in networks, as well as to capture the distances between networks describing different contexts. In the view of an open and successful science, sharing data and methods with the scientific community strongly contributes to the progression of knowledge and methodologies.