In cancer patients, there can be a huge variation in the types of cancer cells from one patient to another, even within the same disease. The identification of the particular cell types present may be very useful in choosing which treatment would be most effective, but the methods of doing so are time consuming and often hampered by human errors and limits of human vision.
In a breakthrough that could signal a new era in cancer diagnosis and treatment, a team from Osaka University and colleagues have shown how these problems can be overcome through a system based on artificial intelligence that can identify different types of cancerous imaging, achieving greater precision than human judgment. This approach can have great benefits in the field of oncology.
The system is based on a convolutional neural network, a form of artificial intelligence modeled in the human visual system. In this study, reported in the journal Cancer ResearchThis system was used to distinguish cancerous cells from mice and humans, as well as equivalent cells that were also selected for radiation resistance.
"First we trained our system on 8,000 images of cells obtained from a phase contrast microscope," says corresponding author Hideshi Ishii. "We then tested his accuracy on another 2,000 images to see if he learned the characteristics that distinguish rat cancer cells from human, and cancer cells radiosensitive to radiosensitive."
By creating a two-dimensional chart of the findings obtained by the system, the results for each cell type were grouped, although they were clearly separated from the other cells. This showed that after training, the system could correctly identify cells based on their microscopic images alone.
"The automation and high accuracy with which this system can identify cells should be very useful in determining exactly which cells are present in a tumor or circulating in the body of cancer patients," said lead author Masayasu Toratani. "For example, knowing whether or not radioresistant cells are present is vital in deciding whether radiotherapy would be effective, and the same approach can be applied after treatment to see if it has the desired effect."
In the future, the team hopes to train the system on more types of cancer cells, with the ultimate goal of establishing a universal system that can automatically identify and distinguish all these cells.