Early Follicular Lymphoma Grading via PET-CT Fusion and Bayesian Deep Learning.
Artificial intelligence combining PET-CT scans can accurately grade lymphoma without biopsy, helping doctors plan treatment without invasive surgery.
A dual-discriminator conditional GAN fusing PET-CT images with Bayesian ResNet achieves 87.1% accuracy and 0.816 macro-F1 in non-invasive follicular lymphoma grading across 837 patients at four centers, including FL Grade I/II differentiation and DLBCL distinction. Bayesian uncertainty quantification resolves ambiguity between adjacent grades, providing a non-invasive biopsy-sparing decision-support tool for clinical FL grading workflows.
What the study was
- Study design
- original_research
- Population
- Multi-center dataset of FL and DLBCL patients (4 Chinese hospitals)
- Sample size
- 837
- Category
- AI / Hematologic Malignancies
- Maturity
- Exploratory
- Journal
- J Imaging Inform Med
Why it surfaced
Multi-center n=837 PET-CT fusion + Bayesian DL framework for FL grading; accuracy 0.871; dual-relevance to hematologic malignancies and AI diagnostics; biopsy-free FL grading has direct clinical workflow utility; uncertainty quantification is a practical clinical addition.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.