Researchers have developed an artificial intelligence algorithm that can automate the detection of cancer cells in blood samples, reducing a process that typically takes hours to approximately 10 minutes. This technological advancement could lead to faster diagnoses, better monitoring of cancer recurrence, and more informed treatment decisions. The new tool, developed by a team at the University of Southern California, addresses the significant challenge of finding rare cancer cells among millions of healthy blood cells, a task often likened to finding a “needle in a haystack.”
The method, known as a liquid biopsy, works by identifying cancer cells that have broken away from tumors and are circulating in the bloodstream. While the concept is not new, existing techniques for finding these cells are typically slow, expensive, and require highly trained specialists to manually analyze slides. The newly developed algorithm, named RED (Rare Event Detection), streamlines this process, offering a more efficient and potentially more accessible way to detect and monitor the disease. The research, a collaboration between the USC Viterbi School of Engineering and the USC Dornsife College of Letters, Arts and Sciences, was detailed in the journal Precision Oncology.
The Challenge of Finding Circulating Tumor Cells
When a tumor metastasizes, it sheds cells into the circulatory system. These circulating tumor cells, or CTCs, hold valuable information for oncologists, but they are incredibly rare. A blood sample might contain just a handful of these cells among billions of red and white blood cells, making their detection a significant technical hurdle. For years, the gold standard has involved a painstaking manual review of blood sample images by pathologists and technicians, a process that is both time-consuming and subject to human error.
This manual process has been a major bottleneck in the widespread clinical adoption of liquid biopsies for cancer management. While the potential of CTC analysis is enormous—offering a non-invasive way to monitor disease progression and treatment efficacy—the logistical challenges have limited its use. The need for a faster, more reliable, and automated system has been a long-standing goal in the field of oncology, driving researchers to explore new technological solutions.
A Novel AI-Powered Solution
The RED algorithm represents a significant step toward overcoming these challenges. Developed by a USC team including doctoral candidate Javier Murgoitio-Esandi, aerospace and mechanical engineering professor Assad Oberai, and professor of biological sciences Peter Kuhn, the AI was trained to distinguish cancer cells from normal blood cells with high accuracy. The system analyzes high-resolution images of blood samples and flags suspicious cells, presenting a curated selection to pathologists for final review. This dramatically reduces the manual labor involved and accelerates the entire process.
The AI’s development was fueled by a deep learning approach, leveraging a vast dataset of human-annotated images related to breast cancer. This dataset was compiled over more than a decade of research led by Kuhn, whose work in the field has been partly inspired by his own mother’s cancer diagnosis. By training on this extensive library of images, the RED algorithm learned the subtle morphological differences that distinguish cancerous cells from their healthy counterparts, enabling it to perform its task with remarkable speed and precision.
Rigorous Testing and Validation
To validate the effectiveness of the RED algorithm, the research team conducted two primary types of tests. First, they used the AI to analyze blood samples from patients with advanced breast cancer, a scenario where the presence of CTCs was already known. In a second set of experiments, the researchers spiked healthy blood samples with a known number of cancer cells to see if the algorithm could successfully identify them.
The results were impressive. The algorithm demonstrated a high degree of sensitivity, successfully identifying 99% of added epithelial cancer cells and 97% of added endothelial cells. This level of accuracy, combined with the system’s speed, suggests that it could be a powerful tool in a clinical setting, providing doctors with crucial information for patient care in a fraction of the time required by current methods.
Broader Trends in AI-Based Cancer Screening
The development of the RED algorithm is part of a larger trend of using artificial intelligence to revolutionize cancer detection. Several other innovative methods are also showing promise in early-stage research and clinical trials. For instance, another approach uses a technique called nano-scale Differential Scanning Fluorimetry (nanoDSF) combined with machine learning to analyze plasma denaturation profiles in blood samples. A proof-of-concept study using this method was able to distinguish between glioma patients and healthy individuals with 92% accuracy.
In another development, scientists in China have created a test that can diagnose pancreatic, gastric, and colorectal cancer from a tiny spot of dried blood in just a few minutes. This method uses machine learning to analyze metabolic changes in the blood and has shown a sensitivity of 82–100% in initial studies. Meanwhile, the UK’s National Health Service is preparing for clinical trials of the miONCO-Dx test, which uses AI to analyze microRNA fragments in the blood to detect up to 12 common cancers with over 99% accuracy. This test requires only a small blood sample and can also help pinpoint the location of the cancer.
Future Implications for Oncology
The integration of AI into cancer diagnostics holds the potential to transform patient care. By making liquid biopsies faster, more accurate, and more accessible, technologies like the RED algorithm could enable more frequent and less invasive monitoring of cancer patients. This could allow oncologists to track the effectiveness of treatments in near real-time and quickly detect any signs of recurrence, leading to more personalized and effective therapeutic strategies. Furthermore, the ability to rapidly screen for cancer could lead to earlier diagnoses, which is one of the most critical factors in improving patient survival rates.
While these new technologies are still in various stages of development and require further validation in large-scale clinical trials, they represent a significant step forward in the fight against cancer. The convergence of advanced imaging, molecular biology, and artificial intelligence is opening up new frontiers in oncology, offering hope for a future where cancer can be detected earlier, monitored more effectively, and treated with greater precision. The work being done by researchers at USC and other institutions around the world is paving the way for a new era of cancer diagnostics, one that is faster, smarter, and ultimately, more life-saving.
