Artificial intelligence detects the presence of viruses

January 9, 2019, UCLA Engineering Institute for Technology Advancement
Deep learning-based sensing of viruses using holography. Credit: UCLA Engineering Institute for Technology Advancement

Many biosensing applications rely on characterization of specific analytes such as proteins, viruses and bacteria, among many other targets, which can be accomplished by using micro- or nano-scale particles. In such biosensors, these particles are coated with a surface chemistry that makes them stick to the target analyte forming clusters in response. The higher the target analyte concentration is, the larger the number of clusters gets. Therefore, monitoring and characterizing these particle clusters can tell us if the target analyte is present in a sample and in what concentration. Current methods to perform such an analysis are limited in that they are either capable of only a coarse readout or rely on expensive and bulky microscopes, which limit their applicability to address different biosensing needs, especially in resource limited environments.

To overcome the shortcomings of the existing solutions, UCLA researchers have developed a rapid and automated biosensing method based on holography coupled with deep learning – currently, one of the most promising and successfully used methods in artificial intelligence, AI. In this system, all the particle clusters and individual micro- in a sample are first imaged in 3-D as holograms, all at the same time, and over a very large sample area of more than 20 mm2, more than ten-fold larger than the imaging area of a standard optical microscope. Next, a trained deep neural network processes these holograms and rapidly reconstructs them into images of clusters similar to those that could be obtained with a standard scanning microscope, but doing this much faster and for a significantly larger sample volume. During this process, all the particle clusters at the micro-scale (revealing the presence of the target analyte) are automatically counted with a sensitivity similar to a laboratory-grade microscope.

As a proof of concept, UCLA researchers successfully demonstrated the application of this deep learning-based biosensing approach to detect (HSV) and achieved a detection limit of ~ 5 viruses per micro-liter, providing a clinically relevant level of sensitivity for HSV detection. HSV is one of the most widespread viral infections that is estimated to have affected more than 50% of the adults in the US.

This work was published as a cover article in ACS Photonics, a journal of the American Chemical Society. The research was led by Dr. Aydogan Ozcan, an associate director of the California NanoSystems Institute (CNSI) and the Chancellor's Professor of electrical and at the UCLA Henry Samueli School of Engineering and Applied Science, along with Yichen Wu, a graduate student, and Aniruddha Ray, a postdoctoral scholar, at the UCLA electrical and computer engineering department.

"Our work demonstrates an automated, inexpensive platform for rapid read-out and quantification of a wide variety of particle clustering-based biosensors. This unique capability enabled by will help democratize instrumentation, making them suitable for wide-scale use even in developing countries," said Ozcan.

Explore further: Deep learning democratizes nano-scale imaging

More information: Yichen Wu et al. Deep Learning Enables High-Throughput Analysis of Particle-Aggregation-Based Biosensors Imaged Using Holography, ACS Photonics (2018). DOI: 10.1021/acsphotonics.8b01479

Related Stories

Deep learning democratizes nano-scale imaging

December 18, 2018

Many problems in physical and biological sciences as well as engineering rely on our ability to monitor objects or processes at nano-scale, and fluorescence microscopy has been used for decades as one of our most useful information ...

Deep learning reconstructs holograms

October 16, 2017

Deep learning has been experiencing a true renaissance especially over the last decade, and it uses multi-layered artificial neural networks for automated analysis of data. Deep learning is one of the most exciting forms ...

Recommended for you

New thermoelectric material delivers record performance

January 17, 2019

Taking advantage of recent advances in using theoretical calculations to predict the properties of new materials, researchers reported Thursday the discovery of a new class of half-Heusler thermoelectric compounds, including ...

Zirconium isotope a master at neutron capture

January 17, 2019

The probability that a nucleus will absorb a neutron is important to many areas of nuclear science, including the production of elements in the cosmos, reactor performance, nuclear medicine and defense applications.

Mechanism helps explain the ear's exquisite sensitivity

January 16, 2019

The human ear, like those of other mammals, is so extraordinarily sensitive that it can detect sound-wave-induced vibrations of the eardrum that move by less than the width of an atom. Now, researchers at MIT have discovered ...


Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.