Super-resolution ultrasound radiomics for pre-FNA prediction of nondiagnostic (Bethesda I) thyroid nodules - Takeaways - MDSpire

Super-resolution ultrasound radiomics for pre-FNA prediction of nondiagnostic (Bethesda I) thyroid nodules

  • By

  • Shaozheng He

  • Guo-Rong Lyu

  • Mingli Cai

  • Jian Lin

  • Kunzhang Zeng

  • Junfa Sheng

  • May 1, 2026

  • 0 min

Share

  • 1

    A super-resolution radiomics framework using GAN significantly improves the prediction of nondiagnostic thyroid nodules prior to fine-needle aspiration.

  • 2

    The study involved 437 patients, with 338 in the development cohort and 99 in the validation cohort, utilizing two-dimensional B-mode ultrasound images.

  • 3

    The SR-RF model achieved an AUC of 0.7435 in the independent validation cohort, outperforming the NR-RF model's AUC of 0.596.

  • 4

    Qualitative error analysis identified false positives linked to cystic textures and false negatives associated with isoechoic solid nodules.

  • 5

    Integrating GAN-based SR radiomics with post hoc calibration enhances personalized risk assessments and reduces unnecessary repeat interventions.

Original Source(s)

Super-resolution ultrasound radiomics for pre-FNA prediction of nondiagnostic (Bethesda I) thyroid nodules

  • by Shaozheng He, Guo-Rong Lyu, Mingli Cai, Jian Lin, Kunzhang Zeng, Junfa Sheng

  • May 1, 2026

Related Content

Case Report: [225Ac]Ac-PSMA-617 therapy in a patient with metastatic castration-resistant prostate cancer (mCRPC) and intracranial meningioma

Differentiation of hepatocellular carcinoma from hepatic hemangioma using diffusion-derived vessel density map-based radiomics features

Glioblastoma, IDH-wildtype, with a novel MEF2D-NTRK1 gene fusion: a case report