(315b) Rapid Sensing of Heat-Stress Markers in Organisms: CNN Based Learning Using Micrographs of Chicken Red Blood Cells Dispersed in Liquid Crystals

An imbalance between bodily heat production and heat loss leads to heat stress in organisms. Heat stress is detrimental to all agricultural systems but is a particularly important stressor for the poultry industry; poultry entails fast growth and high yield, resulting in greater metabolic activity and higher body heat production, along with diminished well-being of the animal. When stressed, cells overexpress heat shock proteins (such as HSP70, a well-established intracellular stress indicator) and may undergo changes in their mechanical properties. Liquid crystals (LCs) respond to, and can be used to rapidly characterize, changes in mechanical properties of red blood cells; this was the driving principle for this research. Expression of HSP70 yielded a different LC response pattern as well as different cellular straining when compared to cells not expressing HSP70 when seen under an optical microscope. For rapid detection of such distinct cellular straining and LC patterns - scenarios where human judgement could be prohibitively difficult or slow - convolution neural network (CNN) based machine-learning (ML) models were trained on hundreds of such micrographs. Trained models exhibited remarkable accuracy of up to 99% on unseen microscope samples. Additionally, mechanical properties of both chicken and human red blood cells were altered by crosslinking them with glutaraldehyde and their micrographs were used to successfully test the CNN based classification methodology prior to the heat-stress experiments.