Machine Learning Helps Boost Accuracy for Liver Cancer Screening
If you’re using alpha fetal protein (AFP) to diagnose liver cancer, you know results can be inexact. You could miss a diagnosis until the later stages of disease, when it’s most difficult to treat.
Researchers, funded in part by NCI’s Division of Cancer Biology, are using a machine learning model to help boost AFP’s diagnostic power. By examining the fusion genes underlying this protein, their model was able to predict genes associated with a high- versus a low-risk of Hepatocellular Carcinoma (HCC).
Read the full report, “Serum Fusion Transcripts to Assess the Risk of Hepatocellular Carcinoma and the Impact of Cancer Treatment through Machine Learning,” in the American Journal of Pathology.
Using this protein-and-fusion gene combo, the researchers had a much higher accuracy (up to 95%) when screening for HCC in their study subjects. They also were able to uncover cancer in people with “normal” serum AFP levels.
Corresponding author, Dr. Jian-Hua Luo, of the University of Pittsburgh School of Medicine, noted that their approach not only helped diagnose cancer, but also helped track residual cancer after treatment. “With our approach, we could see how well treatment is working and predict if cancer would reoccur,” he said.
He added, “Because it requires only a simple blood test, this approach could give primary care physicians, surgeons, and oncologists an efficient decision-making tool to identify early-stage liver cancer, potentially reducing the mortality of the disease.”