Researchers develop deep learning models to classify cancer cells by type

A cancer cell that initiates metastasis, or spreads disease from its primary site, is different from a cancer cell that stays in the original tumor. Distinguishing the cell types that initiate metastasis helps determine the severity of the cancer and helps doctors make treatment decisions.

of APL machine learning, by AIP Publishing, Texas Tech researchers have developed a deep learning model that classifies cancer cells by type. The tool requires a simple microscope and a small amount of computational power, and produces results that are as good or better than more sophisticated and complex techniques.

Cancer cells are highly heterogeneous and recent studies suggest that specific cell subpopulations, rather than the whole, are responsible for cancer metastasis. Identifying cancer cell subpopulations is an important step in determining disease severity. ”

Wei Lee, Author

Current methods of classifying cancer cells involve sophisticated instrumentation, time-consuming biological techniques, or chemical labeling.

“The problem with these complex and time-consuming techniques is that they require resources and effort that can be devoted to investigating different areas of cancer prevention and recovery,” said author Karl Gardner. I’m here.

Magnetic nanoparticles have been used in some studies to track cancer cells, but labeling these can affect the downstream analysis and measurement integrity of the cells.

“Our classification procedure does not consist of additional chemicals or biological solutions when taking pictures of the cells,” says Gardner. It is a method to specify by “free”. ”

The team’s neural networks are also easy to use, efficient, and automated. After feeding the images, the tool converts the data into probabilities. A result of less than 0.5 classifies the cancer as one cell type, while a number greater than 0.5 designates another cell type.

The tool was trained to optimize prediction accuracy using a series of images of two cancer cell lines. He reached an accuracy of over 94% across the datasets used in the study.

Currently, the training data only considers single cancer cells. However, studies have shown that circulating tumor cell clusters are more involved in the spread of cancer. The authors aim to extend and generalize the model to include both single cells and clusters.