Researchers have created an advanced artificial intelligence system that can distinguish between cancer cells and normal cells, as well as identify early signs of viral infections within cells. These discoveries are significant for enhancing diagnostic methods and developing new ways to monitor diseases. The AI can recognize changes within cells as tiny as 20nm, which is 5,000 times smaller than the diameter of a human hair. Such minute alterations are beyond the detection capabilities of traditional methods when observed by human eyes.
A team from the Centre for Genomic Regulation (CRG), the University of the Basque Country (UPV/EHU), Donostia International Physics Center (DIPC), and the Fundación Biofisica Bizkaia (FBB) has developed an innovative AI that differentiates between cancer cells and healthy cells, and also identifies very early stages of viral infections inside cells. Their research, published today in Nature Machine Intelligence, opens the door to better diagnostic tools and new disease monitoring techniques.
This tool, named AINU (which stands for AI of the NUcleus), examines high-resolution cell images. These images come from a specialized microscopy technique called STORM, which captures much finer details than regular microscopes. The resulting images showcase structures at an incredibly detailed nanoscale level.
A nanometre (nm) represents one-billionth of a meter, while a human hair measures about 100,000nm in width. The AI can identify rearrangements inside cells down to 20nm, which is 5,000 times smaller than a human hair’s width. These subtle variations would typically evade detection by human analysts using traditional methods.
“The resolution of these images is so advanced that our AI can accurately recognize specific patterns and differences, including changes in DNA organization within cells. This ability may allow us to detect alterations very soon after they happen. We envision that this type of insight could provide doctors with critical time to monitor diseases, tailor treatments, and enhance patient care,” explains ICREA Research Professor Pia Cosma, one of the co-authors of the study and a researcher at the Centre for Genomic Regulation in Barcelona.
‘Molecular-level facial recognition’
AINU operates as a convolutional neural network, a specific kind of AI that specializes in analyzing visual data, such as images. Convolutional neural networks power technologies like facial recognition for unlocking smartphones, as well as enabling self-driving cars to navigate by identifying various objects on the road.
In the medical field, such networks are employed to examine medical images, including mammograms and CT scans, detecting potential signs of cancer that might not be visible to human observers. They can also assist physicians in identifying anomalies in MRI or X-ray images, leading to quicker and more accurate diagnoses.
AINU is capable of detecting and analyzing minute structures within cells at the molecular level. The researchers trained the AI using high-resolution images of the nuclei of various cell types in different conditions, allowing it to learn to identify specific patterns by examining how the components within the nucleus are arranged in three-dimensional space.
For instance, cancer cells show notable structural changes in their nuclei compared to normal cells, such as differences in DNA organization and enzyme distribution. After being trained, AINU was able to scrutinize new images of cell nuclei and classify them as either cancerous or normal based solely on these distinguishing features.
The AI’s nanoscale imaging ability enabled it to notice changes in a cell’s nucleus just one hour after it was infected by herpes simplex virus type-1. It detected the virus by observing minor differences in DNA packing, indicative of the virus’s impact on the cell’s nuclear structure.
“Our approach allows for the identification of virus-infected cells shortly after infection onset. Currently, spotting an infection relies on visible symptoms or evident alterations in the body, which can take time. However, with AINU, we can immediately observe subtle changes in the cell nucleus,” states Ignacio Arganda-Carreras, co-author of the study and Ikerbasque Research Associate at UPV/EHU, affiliated with FBB-Biofisika Institute and DIPC in San Sebastián/Donostia.
“Researchers can utilize this technology to analyze how viruses influence cells almost immediately after they enter the body, potentially leading to improved treatments and vaccines. In clinical environments, AINU could provide rapid diagnosis of infections through a simple blood or tissue sample, resulting in a faster and more accurate process,” adds Limei Zhong, co-first author of the study and a researcher at the Guangdong Provincial People’s Hospital (GDPH) in Guangzhou, China.
Preparing for clinical application
Before this technology can be effectively tested or used in clinical settings, researchers need to address some key challenges. For example, STORM imaging can only be conducted with specialized equipment usually found in research labs. Setting up and maintaining such imaging systems requires significant investment in both technology and expertise.
Additionally, STORM imaging generally assesses only a handful of cells at a time. For effective diagnostics, particularly in clinical settings where rapid results are vital, it would be necessary to capture a much greater number of cells in a single image.
“There are numerous rapid advancements in the realm of STORM imaging that suggest microscopes may soon become available in less specialized laboratories, and ultimately in clinics as well. The challenges of accessibility and throughput may be solvable much sooner than we initially anticipated, and we aspire to conduct preclinical experiments in the near future,” Dr. Cosma remarks.
While clinical applications may still be a few years away, AINU is expected to expedite scientific research in the short term. The researchers discovered that this technology could identify stem cells with remarkable precision. Stem cells possess the unique ability to transform into any type of cell within the body, a characteristic known as pluripotency. This quality makes stem cells a focus of study for their potential in repairing or replacing damaged tissue.
AINU streamlines the process of recognizing pluripotent cells, enhancing the safety and efficacy of stem cell therapies. “Current methods of detecting high-quality stem cells require animal testing. Our AI model, however, only needs samples treated with specific markers that accentuate essential nuclear features. This not only makes the process easier and quicker but also contributes to reducing animal use in scientific research,” explains Davide Carnevali, the first author of the study and a researcher at the CRG.
