Related Experiment Video
Updated: Jun 10, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.8K
Avoidable biopsies? Validating artificial intelligence-based decision support software in indeterminate thyroid
Christopher J Carnabatu1, David T Fetzer2, Alexander Tessnow3
1Division of Endocrine Surgery, UT Southwestern Medical Center, Dallas, TX. Electronic address: https://twitter.com/CarnabatuMD.
Surgery
|October 13, 2024
Summary
This study validated an artificial intelligence (AI) system for thyroid nodule risk stratification. The AI improved malignancy classification accuracy, potentially reducing unnecessary biopsies for indeterminate thyroid nodules.
Area of Science:
- Medical imaging
- Artificial intelligence in medicine
- Oncology
Background:
- Artificial intelligence (AI) systems are increasingly approved for thyroid nodule risk stratification using sonographic data.
- Validating AI tools like Koios DS is crucial for improving diagnostic accuracy.
Purpose of the Study:
- To validate the Koios DS artificial intelligence (AI) system for enhanced risk stratification of indeterminate thyroid nodules.
- To assess the AI's impact on malignancy classification and biopsy decisions.
Main Methods:
- Retrospective analysis of 28 indeterminate thyroid nodules with molecular testing and surgical pathology.
- Evaluation of nodules using Koios DS, comparing AI-derived and radiologist-derived Thyroid Imaging Reporting and Data System (TI-RADS) levels.
- Assessment of malignancy classification performance and biopsy threshold adjustments using the AI-adapter.
Main Results:
- The AI system showed fair agreement (Cohen's κ=0.25) with radiologist-derived TI-RADS levels.
- The AI-adapter improved malignancy classification, increasing positive predictive value (PPV) from 33.3% to 54.5% while minimally impacting negative predictive value (NPV).
- Utilizing the AI-adapter could have deferred 14 biopsies, with 13 of those being benign.
Conclusions:
- Koios DS demonstrated consistent agreement with radiologist assessments for indeterminate thyroid nodules.
- The AI-adapter enhances malignancy risk stratification, improving PPV with minimal reduction in NPV.
- AI-assisted risk stratification may improve patient counseling and reduce unnecessary biopsies for indeterminate thyroid nodules.
