Site Editor

Leo I. Gordon, MD, FACP


Using an AI Algorithm to Measure Prognostic Biomarkers in DLBCL

By: Jenna Carter, PhD
Posted: Friday, February 17, 2023

Whole-body 18F-FDG PET/CT is the standard of care for staging and assessing responses in patients with diffuse large B-cell lymphoma (DLBCL). Moreover, the disease dissemination reflected by the largest distance between two lesions in the baseline whole-body 18F-FDG PET/CT image (Dmax) was recently shown to be a complementary early prognostic factor, but the process requires expert adjustment and validation, which are often prone to observer variability. To refine this process, Kibrom B. Girum, PhD, of the University Paris-Saclay, and colleagues assessed whether total metabolically active tumor volume and Dmax biomarkers could be replaced by surrogate biomarkers and automatically calculated using an artificial intelligence (AI) algorithm. Their findings, published in The Journal of Nuclear Medicine, revealed that the progression-free survival of tumor volume biomarkers and surrogate biomarkers derived using AI were similar.

A total of 382 patients from the REMARC ( identifier NCT01122472) and LNH073B (NCT00498043) trials were evaluated in this study. Total metabolically active tumor volume and Dmax were measured from lymphoma lesions gathered via expert delineation of baseline whole-body 18F-FDG PET/CT images. An AI algorithm was trained on the REMARC maximum-intensity projection data to segment lymphoma regions, which were used to estimate surrogate total metabolic tumor volume and surrogate Dmax on both data sets. Comparisons in the ability of the original and surrogate biomarkers to stratify patient outcomes were then performed.

Overall findings revealed that surrogate total metabolically active tumor volume was highly correlated with the original tumor volume for REMARC and LNH073B data sets (Spearman r  =  .878 and .752, respectively), and so were surrogate Dmax and original Dmax (r  =  .709 and .714, respectively). The hazard ratios for progression-free survival for original volume versus maximum-intensity projection-based features derived using AI were also similar.

Based on these findings, study authors concluded that surrogate biomarkers calculated from only two PET maximum-intensity projection images may be valid prognostic biomarkers and may be automatically estimated using an AI algorithm.

Disclosure: For full disclosures of the study authors, visit

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.