AI Learns to Detect Cancer Molecules Using ‘Face Recognition’ Technology

Imagine if the technology already used to recognize faces in surveillance cameras or unlock smartphones were used to detect diseases early. That’s what a group of Spanish researchers have just imagined, creating an artificial intelligence (AI) tool called AINU (AI Cores), capable of recognizing specific patterns and changes in the shape of DNA molecules characteristic of cancer and viral infections.

The work is featured in Tuesday’s issue of the journal Nature Machine IntelligenceIt involves scientists from the Centre for Genomic Regulation (CRG), the University of the Basque Country (UPV/EHU), the International Physics Centre of Donostia (DIPC) and the Foundation for Biophysics of Vizcaya (FBB, located at the Institute of Biophysics). The instrument scans high-resolution images of cells obtained using a special microscopy technique called STORM, which creates an image that covers much more detail than conventional microscopes can see.

“We developed an artificial intelligence algorithm that, combined with the use of high-resolution images, allowed us to identify some chromatin changes in the cell nucleus,” explains Pia Cosma, co-author of the study and researcher at the Center for Genomic Regulation (CRG) in Barcelona, ​​to elDiario.

These high-resolution images reveal structures at nanometer (nm) resolution and allow the instrument to detect rearrangements within cells as small as 20 nm, 5,000 times smaller than the width of a human hair, changes that are too small and undetectable for human observers to detect using traditional methods.

“The resolution of these images is high enough that our AI can recognize specific patterns and differences with remarkable accuracy, helping to detect changes very soon after they occur,” explains Cosma. Cancer cells have distinctive changes in their nuclear structure compared to normal cells, such as changes in the way their DNA is organized or the distribution of enzymes within the nucleus. Once trained, AINU was able to analyze new images of cell nuclei and classify them as cancerous or normal based on these characteristics alone.

Look for cancer, look for Wally

AINU is a convolutional neural network, a type of artificial intelligence designed specifically to analyze visual data such as images. In medicine, convolutional neural networks are used to analyze medical images such as mammograms or CT scans and identify signs of cancer that the human eye might miss. They can also help detect abnormalities in MRI or X-ray images, helping to make a diagnosis faster and more accurately.

In this case, the system is a kind of Where is Wally? in which the machine is not looking for a character in a striped shirt, but for abnormal cells that could cause pathology. However, there are some small differences in facial recognition. “To identify a person’s face, you have to use a lot of images to train the algorithm,” says the researcher. “In our case, we used few images because the resolution is very high, but when we encounter two cells that look exactly the same by any other method, this algorithm can tell them apart.”

For this reason, he says, the authors believe that one day this kind of information could allow doctors to gain time to monitor the disease, personalize treatment and improve patient outcomes.” The main advantage of the system is that a prioriOnce a cancer type is identified, the algorithm can continue to detect it in other patients, regardless of their specific mutation. “It will depend on the type of cancer,” says Cosma. “If one patient has one mutation and another has another, with the algorithm we will detect both.”

An hour after infection

The system has also shown its usefulness in viral infections. Using this approach, the AI ​​was able to detect changes in the cell nucleus just an hour after infection with the herpes simplex virus type 1. The model can detect the presence of the virus by detecting small differences in DNA density that occur when the virus begins to change the structure of the cell nucleus.

Our method can detect virus-infected cells very soon after infection begins.

Ignacio Arganda-Carreras
— Co-author of the study and Ikerbaska research fellow at UPV/EHU

“Our method allows us to detect cells infected with the virus very soon after the infection begins,” he says. Ignacio Arganda-Carrerasco-author of the study and Ikerbaska research fellow at UPV/EHU. “Usually, it takes time for doctors to detect an infection because they rely on visible symptoms or more serious changes in the body. But with AINU, we can immediately see small changes in the cell nucleus.”

“This technology can be used to see how viruses affect cells almost immediately after entering the body, which could help develop more effective treatments and vaccines,” adds Limei Zhong, co-author of the study and a researcher at the Guangdong Provincial People’s Hospital (GDPH) in Guangzhou, China. “In hospitals and clinics, AINU can be used to diagnose infections from a simple blood or tissue sample, making the process faster and more accurate.”

How to get tested at the clinic

The study’s authors caution that they still have important limitations to overcome before the technology is ready for testing or clinical implementation. For example, STORM images can only be acquired using specialized equipment typically found only in biomedical research labs. Installing and maintaining the imaging systems required for AI requires significant investment in both equipment and technical skills.

“The availability and performance limitations are more solvable problems than we thought, and we hope to conduct preclinical experiments soon.

Pia Cosma
— Co-author of the study and research fellow at the Center for Genomic Regulation (CRG)

Another limitation is that STORM imaging analyzes multiple cells at once. For diagnostic purposes, especially in clinical settings where speed and efficiency are critical, doctors will need to capture many more cells in a single image to be able to detect or track disease.

But, according to Cosma, these microscopes could soon be available in smaller or less specialized labs. “Simpler microscopes are being developed, and the hope is that very soon they will be available in any lab in any hospital,” he says. “The ideal application would be to analyze a liquid tumor, so that we can detect its early presence by analyzing the blood, but we have to be very careful because we are not there yet; this is just the first proof-of-concept work.”

Identify stem cells

Finally, the study authors saw that this technology could also identify stem cells with very high accuracy. These cells have the potential to become any type of cell in the body, and are being studied for their ability to repair or replace damaged tissue. Developing this potential could help make treatments safer and more effective.

“Current methods for detecting high-quality stem cells rely on animal testing,” says Davide Carnevali, first author of the study and a researcher at CRG. “However, all our AI model needs to work is a sample stained with specific markers that highlight key nuclear features. “In addition to being simpler and faster, it could speed up stem cell research while helping to reduce the use of animals in science.”

Belief in mistakes: The original headline (“Artificial intelligence learns to detect cancer early using ‘facial recognition'”) was changed because it could be misconstrued to mean that cancer could be detected from photographs of patients’ faces.