Related Experiment Video
Updated: Jun 10, 2025
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Evaluating the clinical performance of an updated microRNA classifier in indeterminate and RAS-mutated thyroid nodule
Abhinay Tumati1, Teagan E Marshall2, Benjamin Greenspun1
1Department of Surgery, Weill Cornell Medicine, New York, NY.
Background:
Integrating microRNA markers with next-generation sequencing panels may enhance risk assessment of cytologically indeterminate thyroid nodules. The ThyGeNEXT-ThyraMIRv1 multiplatform test version 1 demonstrated limited utility in risk-stratifying RAS-mutated indeterminate thyroid nodules. We sought to validate the updated ThyraMIRv2 platform in clinical practice.
Methods:
ThyGeNEXT/ThyraMIRv2, a 3-tiered microRNA classifier, were evaluated using a previously studied multi-institutional cohort of Bethesda III/IV nodules, with positive results having risk of malignancy ≥10%. In addition, ThyraMIRv2's clinical utility in RAS-mutated indeterminate thyroid nodules was assessed.
Results:
In 366 indeterminate thyroid nodules, ThyraMIRv2 platform yielded a 30.3% positive-call rate. ThyraMIRv2 platform + nodules had greater operative rates (63.9% vs 36.1%, P < .0001) and cancer/noninvasive follicular thyroid neoplasm with papillary-like nuclear features diagnosis (65.9% vs 25.0%, P < .0001) than ThyraMIRv2 platform nodules. Compared with multiplatform test version 1, ThyraMIRv2 platform's diagnostic testing parameters did not improve significantly. Among 68 RAS-mutated nodules, ThyraMIRv2 classified 36.8%, 55.9%, and 7.4% as positive, moderate, and negative, respectively. All moderate nodules had risk of malignancy ≥10% and were combined with the positive cohort. No significant differences existed in operative rate (81.0% vs 60.0%, P = .272) or cancer/noninvasive follicular thyroid neoplasm with papillary-like nuclear features diagnosis (47.6% vs 40.0%, P > .999) between RAS-mutated positive/moderate and negative groups. For RAS-mutated nodules, ThyraMIRv2 demonstrated improved sensitivity (93.8% vs 64.7, P = .003) and decreased specificity (4.5% vs 34.8%, P = .008) compared with ThyGeNEXT-ThyraMIRv1 multiplatform test version 1, with comparable negative predictive value (33.3% vs 40.0%, P = .731) and positive predictive value (58.8% vs 59.5%, P = .864).
Conclusion:
ThyraMIRv2 platform does not improve indeterminate thyroid nodule malignancy stratification compared to ThyGeNEXT-ThyraMIRv1 multiplatform test version 1. ThyraMIRv2 improves malignant RAS-mutated nodule detection but increases false positives. Future studies encompassing a larger cohort of RAS-mutated with surgical pathology results are warranted to better characterize the performance parameters of this classifier.
