Development of a Deep Learning algorithm for biomedical image segmentation: study of metastatic Extracellular vesicles

Computer vision (CV) tools have emerged as indispensable assets in the fields of medicine and biology, providing a broad spectrum of applications for both research and diagnostic purposes. These tools facilitate the precise localization and delineation of specific tissues and cellular structures within time-lapse sequences or static images. This process is known as “images segmentation”, the identification of objects of interest in a picture. Leveraging advanced artificial intelligence (AI) algorithms, image segmentation has become a cornerstone in biomedical research. Despite significant advances in the diagnosis and treatment of primary tumors, the prevention and management of metastasis remain critical and urgent areas of research into clinical oncology. This unmet need underscores the necessity of continued investigation and innovation in order to improve outcomes for patients affected by malignant tumors. Addressing this, the AI CUrES project aims to deepen our understanding of metastasis biology by predicting the role of circulating extracellular vesicles (EVs) in cancer metastasis with high resolution. Leveraging the diverse applications of image segmentation models in biological and medical research, this study aims to validate a tool for monitoring particles involved in the metastatic process of colorectal cancer through several key objectives. The first one is to establish a comprehensive segmentation benchmark using meticulously gathered in-house images of selected cells. Concurrently, multiple datasets will be created to measure and validate image characteristics, ensuring their suitability for future cell segmentation tasks. These datasets will be refined to enhance their utility for the final application of the chosen AI algorithm. The second key objective is to validate an effective pipeline for delineating the specific particles of our case study. Our video will feature images of cells, such as lung cells, and extracellular vesicles (EVs) captured in the laboratory following complex preparation and staining procedures. The ultimate goal is to utilize the developed framework, encompassing best practices for time-lapse acquisition and suitable deep learning algorithms, to study the behaviour of extracellular vesicles (EVs) around selected cells in targeted experiments.