Researchers and engineers at UCI recently created a machine learning-assisted biochip that can both examine and differentiate between cancers and healthy tissues at the single cell level. The data produced by the device can be used to study tumor heterogeneity, which can help reduce resistance to cancer therapies.
Single-cell analysis is critical in understanding cancer because intercellular homogeneity within the same tumor and intratumor homogeneity within various tumors are the leading causes of resistance to cancer therapies. The instruments and techniques frequently used to perform single-cell analysis are large, expensive, require human specialists to operate, and take a long time to prepare. The team at UCI detailed their solution to this problem in a paper describing a new machine learning-assisted nanoparticle-printed biochip for single-cell analysis and precise characterization of a variety of cancer cells.
"The World Health Organization says that nearly 60 percent of deaths from breast cancer happen because of a lack of early detection programs in countries with meager resources," said senior author Rahim Esfandyarpour; "Our work has potential applications in single-cell studies, in tumor heterogeneity studies and, perhaps, in point-of-care cancer diagnostics -- especially in developing nations where cost, constrained infrastructure and limited access to medical technologies are of the utmost importance."
The biochip works by monitoring differences in the electrical properties of diseased and healthy cells of samples which travel through microfluidic channels. In order to make this device easy to manufacture in diverse settings, many of the materials used are both reusable and inexpensive. Additionally, the researchers devised a way to prototype the main parts of the biochip in about 20 minutes with an inkjet printer. Machine learning is used to analyze and make predictions on the large amount of pixel data which is outputted by the device. By finding patterns in the images that humans cannot see, the program is able to correctly identify aggressive and non-aggressive disease 96% of the time. According to Howard Petty, PhD, a professor of ophthalmology and visual sciences, and of microbiology and immunology at Michigan Medicine, a human looking at these images would achieve about 70% accuracy on the same task.
The machine learning system can also replace a team of highly skilled analysts to study the output data, reducing the cost and time it takes to predict precise outcomes. N-feature classifiers are used by the team at UCI to create a system capable of differentiating a variety of cell types in a label-free manner. The system can also classify cancer subtype cells.
Tools for data analytics and machine learning are playing a growing role in improving cancer treatment and diagnosis. With the global healthcare analytics market currently valued at $14 billion, and expected to grow to 50.4 billion by 2025, much of this funding and attention will probably be pointed at solving cancer. This project was supported by startup funding from UCI’s Henry Samueli School of Engineering.