Current Projects

Multiscale modeling of inherited and acquired cardiac arrhythmia

We employ advanced optics methods to stretch biological cells in a microfluidic assembly without any mechanical contact to record and characterize their cytoskeletal elasticity. An on-chip dual-beam optical tweezer setup is used to trap and stretch the cells flowing through a microfluidic channel. Our experiments showed distinct stretching curves for different cell types. Specifically, based on distinct cytoskeletal stretching properties, it is possible to identify breast cancer cells even before any morphological changes begin to manifest. For reliably identifying cancer cells, the optical microfluidic setup will be extended to accommodate Surface Enhanced Raman Spectroscopy (SERS) to characterize the cell-membrane-bound proteins in biological samples. The SERS substrate will be functionalized to identify early biomarkers of breast cancer such as human epidermal growth factor receptor 2 (HER2). This research has been funded by two NSF grants.

Machine learning and AI methods are being increasingly used to accurately learn complex data patterns and automate analytical model development. We use AI methods such as Artificial Neural Networks, and deep learning networks to facilitate accurate breast cancer tumor identification from histology images, perform underwater obstacle identification for underwater swarm robots and other image-based classification/prediction problems. A genetic algorithm-based hyperparameter optimization algorithm was developed to achieve optimal deep network for each dataset. The deep leaning-based algorithm is also being designed to identify disease progression in Atrial Fibrillation patients and their responsiveness to drug treatment. Deep neural networks are designed to automatically detect and classify faults/defects in semiconductor chips and printer circuit boards (funded by Virginia Microelectronics Consortium-VMEC).