This white paper presents a comparative study evaluating S-Detect™, Samsung Medison’s AI-powered tool, in the context of thyroid ultrasound for nodule classification. The study compares S-Detect™’s performance against experienced sonographers following the British Thyroid Association (BTA) guidelines, which are widely used in UK clinical practice.
S-Detect™: This software, developed by Samsung Medison, aims to classify nodules found by ultrasound to indicate the possibility of malignancy and to assist in patient management. This software is used in breast and thyroid imaging and shows high sensitivity.
Using 51 patient cases, both normal thyroid ultrasound (classified as BTA U1–U2) and abnormal thyroid ultrasound findings (U3–U5) were independently assessed by a sonographer and the S-Detect™ software. S-Detect™ demonstrated:
- 100% sensitivity, correctly identifying all nodules flagged as abnormal.
- 83.7% specificity, accurately recognizing normal thyroid patterns.
- 100% negative predictive value, ensuring no malignant nodules were missed.
- A Kappa coefficient of 0.287, indicating fair agreement with the clinical standard.
The S-Detect™ software shows “fair agreement” with sonographers in determining the presence or absence of thyroid disease. Although the software is likely to overclassify as positive, it appears unlikely to miss malignant nodules. Therefore, the S-Detect™ software has a high potential to be used as an adjunct to current clinical practice.
For a practical application guide, see the Quick Guide to S-Detect™ for Thyroid .
