Related Experiment Videos
Contrast-enhanced ultrasound-based energy response phenotyping in thyroid nodule ablation: identifying heterogeneous
Xinjia Liu1, Xuejing Zhang1, Huifang Liu2
1Department of Medical Ultrasonics, The Central Hospital of Wuhan, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China.
Background:
Thermal ablation of benign thyroid nodules shows substantial inter-patient heterogeneity, yet no pre-procedural tool exists to stratify treatment response. This study aims to identify energy response phenotypes in thyroid nodule thermal ablation using latent class analysis (LCA) and develop a pre-ablation predictive model.
Methods:
This retrospective study analyzed 157 patients undergoing thermal ablation for benign thyroid nodules. LCA integrated pre-ablation contrast-enhanced ultrasound (CEUS) parameters [peak intensity (PI), time to peak (TTP), area under the curve (AUC)], ablation energy, and volume reduction rates (VRRs) at 3, 6, and 12 months to identify distinct phenotypes. A multinomial logistic regression model using baseline CEUS and clinical variables predicted phenotype membership, validated in a held-out testing set (n=47, 30%).
Results:
Three energy response phenotypes were identified with adequate classification certainty (entropy =0.833): class 1 (delayed-response, n=56, 35.7%) exhibited the highest PI and prolonged TTP (both P<0.001) with gradual improvement (VRR 89.86% at 12 months); class 2 (energy-resistant, n=33, 21.0%) demonstrated reduced PI, rapid washout, required 7.6-fold higher energy (P<0.001), yet achieved poorest outcomes (12-month VRR 86.56%, P<0.001); class 3 (optimal-response, n=68, 43.3%) showed balanced perfusion achieving the highest VRR (97.65% at 12 months, P<0.001). Pre-ablation TTP emerged as the dominant predictor (100% relative importance). The discriminant model achieved good discrimination for class 1 (AUC =0.878) and class 2 (AUC =0.916), but limited discrimination for class 3 (AUC =0.726). Ten-fold cross-validation showed moderate stability (accuracy 59.73%±15.93%). Sensitivity analyses excluding energy from LCA confirmed phenotype stability (κ=0.82).
Conclusions:
CEUS-based phenotyping successfully stratifies thyroid nodule ablation candidates, enabling personalized treatment planning. Pre-ablation prediction identifies energy-resistant patients requiring alternative therapeutic strategies.