Imaging medulloblastoma tumors presents several challenges, including tissue autofluorescence, low signal-to-noise ratios, and the complexity of distinguishing immune cells within densely packed tumor regions. Current analysis methods rely largely on signal brightness and frequently confuse true biology with autofluorescence or spectral overlap. My work develops a morphology-aware machine learning model that uses visual structure and texture, rather than brightness alone, to distinguish real marker expression from these artifacts. By improving how cells are identified and classified, my approach seeks to produce more reliable cell typing.

Through this SuperUROP, I hope to gain hands-on experience applying machine learning and image processing techniques to biological data, which will complement my lab experience in experimental biology. I am eager to enhance my understanding of advanced statistical analyses and imaging pipelines, which will be crucial for my long-term interests in disease research and public health.