Auditable agentic AI for evidence-grounded thyroid ultrasound diagnosis and reporting

2026-08-12

AI Summary

This work addresses the limited clinical auditability of existing AI systems for thyroid ultrasound, which typically handle lesion localization, risk stratification, and report generation in isolation. The authors propose ThyroidXAgent—the first agent-based architecture enabling physician-in-the-loop correction and full diagnostic traceability—by orchestrating specialized diagnostic modules and constructing a structured evidence chain for end-to-end multitask diagnosis. Leveraging the multicenter OpenThyroidDB dataset, the system integrates segmentation, benign-malignant classification, lymph node metastasis prediction, and report generation via an evidence-driven multimodal language model assembly strategy. A novel metric, ThyClinScore, is introduced to assess clinical semantic consistency. Evaluated on 28,458 cases, the system achieves a Dice score of 87.21% for nodule segmentation and an AUROC of 0.9466 for malignancy classification, significantly improving diagnostic accuracy and report consistency while reducing segmentation and reporting time by 35.9% and 27.4%, respectively.

Abstract

Thyroid ultrasound diagnosis requires coordinated lesion localization, measurement, risk stratification and reporting, yet most AI systems address these tasks in isolation and provide limited support for clinical review. We present ThyroidXAgent, a clinician-interactive agentic AI system that coordinates specialized diagnostic tools and stores their outputs as an auditable case-level evidence record. The system was developed using OpenThyroidDB, a multicentre, multitask resource integrating approximately 0.3 million ultrasound images and 24,000 paired reports, and was evaluated on 28,458 non-overlapping test cases, including 8,721 cases from 35 centres in the private NHC-MISD-TUS cohort. Across heterogeneous datasets, ThyroidXAgent achieved a mean Dice score of 87.21 percent for nodule segmentation and a mean AUROC of 0.9466 for benign-malignant classification. The same workflow supported lymph-node metastasis prediction and follicular versus papillary thyroid carcinoma classification, with AUROCs of 0.864 and 0.805, respectively. For report generation, evidence-grounded assembly outperformed multimodal language-model baselines across three cohorts. ThyClinScore, a lesion-level clinical semantic metric introduced here, showed the strongest correlation with a location-aware language-model judge. ThyroidXAgent improved physician classification accuracy, increased report diagnostic consistency from 70.3 percent to 86.2 percent, and reduced segmentation and reporting time by 35.9 percent and 27.4 percent, respectively. These findings support auditable, clinician-correctable agentic AI for thyroid ultrasound diagnosis and reporting.

Problem

• thyroid ultrasound diagnosis
• auditable AI
• clinical reporting
• evidence-grounded diagnosis
• agentic AI

Innovation

• auditable AI
• agentic AI
• evidence-grounded reporting
• multitask thyroid ultrasound diagnosis
• clinician-interactive system