Related Experiment Video
Updated: Aug 6, 2026
08:58
Label-Free Identification of Lymphocyte Subtypes Using Three-Dimensional Quantitative Phase Imaging and Machine Learning
Published on: November 19, 2018
Early Follicular Lymphoma Grading via PET-CT Fusion and Bayesian Deep Learning
Chunjun Qian1, Lulu He1, Qiuhui Jiang2
1Hertfordshire College, Changzhou Institute of Technology, Changzhou, 213032, Jiangsu, China.
Journal of Imaging Informatics in Medicine
|August 4, 2026
Summary
An AI framework uses PET-CT scans for non-invasive follicular lymphoma (FL) grading, improving accuracy and reducing observer variability. This physician-guided tool aids personalized treatment decisions by analyzing metabolic and structural imaging data.
Area of Science:
- Oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate follicular lymphoma (FL) grading is vital for personalized treatment.
- Current histopathology methods are invasive and prone to observer variability.
Purpose of the Study:
- To develop an AI framework for physician-guided, non-invasive FL grading using PET-CT imaging.
- To enhance diagnostic accuracy and reduce subjectivity in FL classification.
Main Methods:
- An enhanced dual-discriminator conditional GAN (DDCGAN) with similarity and chrominance constraints for image fusion.
- A Bayesian ResNet to model predictive uncertainty and differentiate between FL Grades I and II.
- Validation on a multi-center dataset of 837 patients with FL and diffuse large B-cell lymphoma (DLBCL).
Main Results:
- The AI framework achieved superior generalizability compared to single-modality and existing fusion models.
- Achieved an accuracy of 0.871, precision of 0.875, and macro-F1 of 0.816.
- Successfully resolved classification ambiguity between adjacent FL Grades I and II.
Conclusions:
- The developed AI framework provides a practical, non-invasive decision-support tool for FL grading.
- The task-oriented image fusion and uncertainty-aware approach support scalable clinical decision-making.
- This technology has the potential to improve patient management workflows in hospitals.