Machine learning identifies quality of stored blood

An international research collaboration describe means to automatically assess the quality of stored red blood cells: “We developed a strategy to avoid human subjectivity by assessing the quality of red blood cells using imaging flow cytometry and deep learning. We successfully automated traditional expert assessment by training a computer with example images of healthy and unhealthy morphologies. However, we noticed that experts disagree on ∼18% of cells, so instead of relying on experts’ visual assessment, we taught a deep-learning network the degradation phenotypes objectively from images of red blood cells sampled over time. Although training with diverse samples is needed to create and validate a clinical-grade model, doing so would eliminate subjective assessment and facilitate research.”     MORE

Image Credit:  Minh Doan, Joseph Sebastian, Tracey Turner, Jason Acker, Michael Kolios, Anne Carpenter/Harvard, Broad Institute