Robust and Generalizable AI Models for Label-free Cellular Organelle Identification

Light microscopy is an essential research tool for characterizing cell's internal organs, known as organelles. Unfortunately, it is often challenging to experimentally label these structures for visualization without substantially disturbing the cells. Recent studies have shown that artificial intelligence (AI) can virtually label organelles in microscope images. Despite the promising potential, AI techniques have not been widely used due to the complicated AI training processes and requirement of large training data. To overcome such obstacles, this project will develop two groundbreaking AI image translation features. First, the research team will implement a transfer learning technique that allows the AI model to apply its previous learning experiences to new tasks, reducing the need of training images. Second, the research team will develop an adaptation mechanism to ensure accurate and consistent predictions across different imaging conditions, enabling model transfer and sharing between different laboratories. This new bioinfrastructure will provide scientists with a valuable tool for visualizing organelles and important biological processes within living cells. The project will support education and diversity through the recruitment of underrepresented researchers. Light microscopy is an essential research tool for characterizing cell's internal organs, known as organelles. Unfortunately, it is often challenging to experimentally label these structures for visualization without substantially disturbing the cells. Recent studies have shown that artificial intelligence (AI) can virtually label organelles in microscope images. Despite the promising potential, AI techniques have not been widely used due to the complicated AI training processes and requirement of large training data. To overcome such obstacles, this project will develop two groundbreaking AI image translation features. First, the research team will implement a transfer learning technique that allows the AI model to apply its previous learning experiences to new tasks, reducing the need of training images. Second, the research team will develop an adaptation mechanism to ensure accurate and consistent predictions across different imaging conditions, enabling model transfer and sharing between different laboratories. This new bioinfrastructure will provide scientists with a valuable tool for visualizing organelles and important biological processes within living cells. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.