JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Assessment of quantitative staging PET/computed tomography parameters using machine learning for early detection of progression in diffuse large B-cell lymphoma

  • 0Department of Nuclear Medicine.
Nuclear Medicine Communications +

|

June 30, 2025

|

June 30, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

Machine learning models using pretreatment 18-fluorodeoxyglucose PET/CT parameters can predict progression in diffuse large B-cell lymphoma (DLBCL). These quantitative metrics improve early risk stratification and guide treatment decisions for DLBCL patients.

Area Of Science

  • Nuclear medicine imaging
  • Machine learning in oncology
  • Hematologic malignancies

Background

  • Diffuse large B-cell lymphoma (DLBCL) is an aggressive non-Hodgkin lymphoma.
  • Early prediction of progression or relapse is crucial for timely treatment adjustments.
  • Standardized uptake value (SUV) thresholds in PET/CT are used for quantitative analysis.

Purpose Of The Study

  • To evaluate the predictive value of volumetric and dissemination parameters from pretreatment 18F-FDG PET/CT in DLBCL.
  • To apply machine learning algorithms for predicting disease progression/relapse.
  • To enhance early risk stratification and guide treatment decisions in DLBCL patients.

Main Methods

  • Retrospective review of 90 DLBCL patients treated with R-CHOP chemotherapy.
  • Extraction of quantitative PET parameters: total metabolic tumor volume (tMTV), total lesion glycolysis (tTLG), longest distance between tumor foci (Dmax), and metabolic bulk volume (MBV).
  • Development of machine learning models using PET parameters and clinical data to predict progression within 1 year.

Main Results

  • Sixteen out of 90 patients experienced progression within 1 year.
  • Significant differences in tMTV, tTLG, MBV, and Dmax were observed between progressive and non-progressive groups.
  • Machine learning models achieved AUCs of 0.701 (clinical data), 0.871 (PET parameters with random forest), and 0.838 (PET parameters and clinical data with Naive Bayes).

Conclusions

  • Quantitative parameters from staging 18F-FDG PET/CT, analyzed with machine learning, can effectively predict early progression in DLBCL.
  • This approach offers potential for improved early risk stratification.
  • The findings may aid in guiding personalized treatment strategies for DLBCL patients.

Keywords: