Nodules are a common finding within the thyroid with up to 67% of adults found to have them. However, only 4-7% of nodules are found to be malignant on cytology [1~2]. Given the large number of patients presenting with thyroid nodules it poses a diagnostic
problem determining which nodules require intervention and which do not. Artificial intelligence (AI) is becoming increasingly utilised in a healthcare environment, with the Care Quality Commission (CQC) stating in 2018 that “diagnostic imaging will be ‘revolutionised’ by machine learning”. AI aims to mitigate the issue of inter-operator or inter-observer reliability, which is a welldocumented limitation of ultrasound as a modality. There is growing support of AI in a radiology setting with Buda et al (2019) demonstrating AI may match performance of radiologists using an American grading system for nodules.
This study evaluates a novel technology created by SAMSUNG MEDISON, CO. LTD., Korea termed S-DetectTM. This software aims to classify nodules discovered with ultrasound to give an indicator of malignancy, and as such guide the further management of the patient.
S-DetectTM has been recently utilised in the application of breast and thyroid imaging, with studies finding that there is good agreement and high sensitivity [3~5]. However, these studies have not utilised BTA guidelines in the classification of nodules which is the current standard of practice within most healthcare providers in the UK.