Site Editor

Al B. Benson III, MD, FACP, FASCO

Advertisement
Advertisement

Can Machine-Learning Models Predict Response to Treatment in Hepatocellular Carcinoma?

By: Sarah Lynch
Posted: Monday, December 19, 2022

Researchers at the Yale School of Medicine have determined that creating machine learning (ML) models using patient data may improve the ability to predict response to treatment of hepatocellular carcinoma and tumor recurrence. In addition, these models may help to create more reliable criteria for organ donation and liver transplantation preparation, according to Julius Chapiro, MD, PhD, of the Yale School of Medicine, New Haven, Connecticut, and colleagues, who reported their findings in the American Journal of Roentgenology.

“The findings suggest that machine learning–based models can predict recurrence before therapy allocation in patients with early-stage hepatocellular carcinoma initially eligible for liver transplant,” stated Dr. Chapiro in a press release from the American Roentgen Ray Society.

A total of 120 patients with hepatocellular carcinoma, with a median age of 60, were selected for the study. They had all been treated with transplantation, resection, or thermal ablation, and all had pretreatment MRI and imaging surveillance after treatment. These imaging data were then used (along with the patients’ lab data) to create three machine learning models: a clinical model, an imaging model, and a combination of both. These models were projected from 1 to 6 years after the patient’s initial treatment. All models were evaluated with a Kaplan-Meier analysis to determine their clinical relevance.

Tumor recurrence occurred in 36.7% of the studied patients during follow-up. The machine learning–based models worked well in predicting the risk of the cancer recurrence, according to the investigators. The predictive performance was significantly higher for the imaging model than for the clinical model. Of note, no significant improvement was visible when using both the clinical and the imaging data combined. Optimizing the predictive performance of these models may encourage the development of new, more dependable liver transplantation criteria.

Disclosure: The study authors reported no conflicts of interest.


By continuing to browse this site you permit us and our partners to place identification cookies on your browser and agree to our use of cookies to identify you for marketing. Read our Privacy Policy to learn more.