Improving PET/CT-Based Whole-Body Lesion Segmentation Using Prediction Uncertainty-Augmented Models
Researchers have developed a new framework to improve the segmentation of lesions in whole-body PET/CT scans for cancer staging. This approach integrates Bayesian ensembling to reduce variability and quantifies uncertainty to highlight areas of potential misclassification. The uncertainty-aware training enhances lesion detection, though it involves a trade-off with precision, and a case-adaptive routing strategy further refines performance. AI
IMPACT Enhances diagnostic accuracy in oncology by improving lesion detection and segmentation in medical imaging.