Ultrasound Large-scale Model in Thyroid Field: A Multi-center Study on Medical-Engineering Integration

A Multicenter Study on Multimodal Diagnosis and Process Optimization of Thyroid Diseases Based on Ultrasonic Intelligent Agents

Status

Active, not recruiting

Phases

Unknown

Study type

Observational

Source

ClinicalTrials.gov

Registry ID

NCT07296575

Enrollment

2000

Registered

2025-12-22

Start date

2025-10-01

Completion date

2027-12-31

Last updated

2025-12-22

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Thyroid, Thyroid Abnormalities, Thyroid Cancer

Keywords

thyroid, Artificial Intelligence, Ultrasound

Brief summary

The goal of this observational study is to evaluate the diagnostic accuracy and clinical workflow integration of an ultrasound intelligent agent (UIA) for thyroid disease management in a real-world multicenter setting. The primary research question is: Can the UIA improve diagnostic consistency and efficiency for thyroid nodules (TI-RADS 1-5), Hashimoto's thyroiditis, and cervical lymph node metastasis compared to traditional ultrasound interpretation? Participants will include adults (18-80 years) undergoing thyroid ultrasound at 16 participating hospitals across China. Key inclusion criteria cover patients with suspected thyroid disorders requiring imaging, while exclusion criteria address poor image quality or concurrent clinical trials. Over 2,000 cases (50% thyroid nodules, 30% diffuse lesions, 12.5% non-nodular abnormalities, 7.5% special populations) will be prospectively enrolled. Data collection integrates static/dynamic ultrasound images, laboratory results, and AI-generated reports. Primary endpoints include model performance metrics (AUC, sensitivity/specificity, TI-RADS Kappa ≥0.8), workflow efficiency (report generation time ≤5 minutes), and pediatric/pregnancy-specific reference standards. Secondary analyses will assess inter-rater reliability (Cohen's Kappa) and longitudinal outcomes via 6-12-month follow-up. This study aims to establish evidence-based guidelines for AI-augmented thyroid diagnosis, particularly in underserved regions, while addressing gaps in current AI validation frameworks related to multi-modality data fusion and special population adaptability.

Interventions

DIAGNOSTIC_TESTNo intervention

no Intervention

Study design

Observational model

OTHER

Time perspective

CROSS_SECTIONAL

Eligibility

Sex/Gender

ALL

Healthy volunteers

Yes

Inclusion criteria

* Ages 18-80, clinically suspected thyroid disease (e.g., enlargement, nodules, pain) requiring ultrasound examination; * Diffuse lesions must demonstrate both ultrasound characteristics and laboratory evidence; * Dynamic video must fully cover the maximum diameter of the nodule without significant probe movement; * Voluntary signed informed consent.

Exclusion criteria

* Poor image quality (severe gas interference, artifacts obscuring structural visualization); * Inability to cooperate with examination (consciousness impairment, extreme non-compliance); * Prior participation in other thyroid ultrasound-related clinical trials; postoperative thyroid recurrence; * Thyroid malformation/ectopia affecting visualization.

Design outcomes

Primary

Measure	Time frame	Description
Predictive Performance of Large Models in Ultrasound Thyroid Applications for Thyroid Diseases	Within 12 months of enrollment for each patient at the time of study completion.	Diagnosing thyroid diseases using a large model in the field of ultrasound thyroid imaging, with histopathological examination results of thyroid lesions as the gold standard, to evaluate the model's sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC) for diagnosing thyroid diseases.

Countries

China

Outcome results

None listed

Source: ClinicalTrials.gov · Data processed: Feb 4, 2026