AI Algorithm ‘RED’ Dramatically Accelerates Cancer Detection in Liquid Biopsies
A groundbreaking new artificial intelligence algorithm is poised to revolutionize cancer detection, offering the potential for faster diagnoses, improved monitoring of treatment effectiveness, and ultimately, better patient outcomes. The technology, dubbed RED (Rare Event Detection), can identify minuscule traces of cancer cells in blood samples – known as a liquid biopsy – in as little as 10 minutes, a process that currently demands hours of painstaking review by highly trained specialists.
The challenge with liquid biopsies lies in the sheer volume of cells present on a single slide. Millions of cells on a single slide.
Researchers at the USC Viterbi School of Engineering and the USC Dornsife College of Letters, Arts and Sciences have overcome this hurdle with RED, a deep learning algorithm detailed in a recent paper in Precision Oncology. The research team, led by Javier Murgoitio-Esandi, Assad Oberai, and Peter Kuhn, developed a system that doesn’t rely on pre-defined characteristics of cancer cells. Instead, RED identifies unusual patterns and prioritizes findings based on their rarity.
“Machines do not need to curate information likewise humans do,” explained a senior researcher involved in the project. This allows RED to function independently, unlike many existing computational tools that require human intervention. The algorithm operates on the principle of identifying “outliers,” much like the classic Sesame Street game where players identify the item that doesn’t belong. As one researcher put it, RED can effectively “seperate outliers from non-outliers” within the vast cellular landscape.
The growth builds upon previous work related to breast cancer, but its potential extends far beyond a single cancer type.The team tested RED’s capabilities in two key ways: analyzing blood samples from patients with advanced breast cancer and adding cancer cells to healthy blood samples to assess the algorithm’s detection rate. The results where striking.
RED demonstrated a remarkable ability to identify added cancer cells, achieving:
- 99% detection of added epithelial cancer cells
- 97% detection of added endothelial cells
crucially, the algorithm reduced the amount of data requiring review by an astonishing 1,000 times. According to researchers, this not only accelerates the diagnostic process but also minimizes the potential for human bias. “We are able to find more signal than the old approach. We were able to find twice as many interesting cells compared to the old approach,” a lead scientist stated.
The implications of this technology are far-reaching, with potential applications in understanding outcomes for cancers like pancreatic cancer and multiple myeloma, in addition to breast cancer. The team hopes RED will ultimately contribute to answering three critical questions in a patient’s cancer journey: Do I have cancer? Is my cancer gone or returning? And what is the best next treatment?
“Each of those parts of the patient journey,we want to support with data from the blood,” explained a senior researcher. This work represents a notable leap forward in the submission of artificial intelligence to healthcare, paving the way for earlier and more accurate cancer diagnoses.
The research, funded by the Ming Hsieh Institute for Research on Engineering-Medicine for Cancer Research, exemplifies the power of convergent research, bringing together expertise from diverse fields to address a complex medical challenge.As one researcher emphasized, “This is one of the really great examples where modern AI is really changing the way we do healthcare research.” The team’s next steps involve continuing to refine the algorithm and expanding its application to a wider range of cancers, with the ultimate goal of radically improving cancer detection in patients.
Reference: Murgoitio-Esandi J,Tessone D,Naghdloo A,et al. Unsupervised detection of rare events in liquid biopsy assays. npj Precis onc.2025;9(1):1-12. doi: 10.1038/s41698-025-01015-3
