Segmentation of Neural Cells by Instance Using Deep Learning Methods

Neurological disorders, encompassing neurodegenerative diseases like Alzheimer's and brain tumors, pose a significant global health challenge, leading to high mortality and disability rates. Assessing the efficacy of treatments for these disorders remains challenging. Accurate instance segmentation, facilitated by computer vision, holds the potential for groundbreaking drug discoveries to address these disorders effectively. The precision of the current methods is restricted for neuronal cells, and the neuroblastoma cell line SH-SY5Y is a constant source of difficulties. The peculiar, asymmetric, and concave form of neuronal cells may cause this difficulty in segmenting them with traditional techniques. The paper applies many methodologies, including TL, Machine Learning, and Deep Learning techniques, to train and validate instance segmentation models targeting neuronal cells with excellent accuracy using phase contrast microscopy pictures. A successful outcome in this project holds the potential to advance neurobiology research by providing robust quantitative data, thereby facilitating easier measurement of the effects of diseases and treatments on neuronal cells.