Clinical Thyroidology® for the Public

Summaries for the Public from recent articles in Clinical Thyroidology
Table of Contents | PDF File for Saving and Printing

THYROID NODULES
The AI “goalkeeper”: preventing unnecessary biopsies in the era of thyroid cancer overdiagnosis

12

Instagram Youtube LinkedIn Facebook X

BACKGROUND
Thyroid nodules are common, occurring in up to 50% of individuals that get any imaging of the neck. The concern about a thyroid nodule is whether the nodule contains a cancer. Many nodules are small and most types of thyroid cancer grow very slowly. Nodules are evaluated as to whether they are cancer by a thyroid biopsy. Doctors use ultrasound to decide who needs a biopsy, but this decision can be different from one doctor to another. Because doctors do not want to miss cancer, many nodules that are actually safe still get biopsied. This leads to stress, extra procedures, and higher costs.

Artificial intelligence (AI) has been used to help standardize the ultrasound characterization of thyroid nodules. In this study, researchers wanted to see if an AI computer program evaluating ultrasound images could safely reduce unnecessary biopsies.

THE FULL ARTICLE TITLE
 Ni JH et al. Optimizing thyroid nodule management with artificial intelligence: multicenter retrospective study on reducing unnecessary fine needle aspirations. JMIR Med Inform 2025;13:e71740; doi: 10.2196/71740. PMID: 40737551.

SUMMARY OF THE STUDY
The study looked at ultrasound images from over 4,500 adults with thyroid nodules. The participants had an average age of 49.4 years and 75.3% were female.

All nodules had already been recommended for biopsy by doctors. All nodules either had benign biopsy results with >12 months of stable follow-up of the nodule or had the nodule surgically removed with pathology results. A computer program reviewed the images and decided which nodules were likely benign.

The computer correctly identified most benign nodules and could have reduced unnecessary biopsies from about 70% to about 9%. The computer performed better than both junior and senior doctors. However, 123 cancers (8%) were misdiagnosed as benign, with 56.9% being papillary microcancers (small cancers <1 cm)

WHAT ARE THE IMPLICATIONS OF THIS STUDY?
This study suggests that AI characterization works well as a clinical “goalkeeper,” significantly outperforming radiologists and potentially preventing nearly 90% of unnecessary biopsies in the study group. Fewer unnecessary biopsies mean less pain, less worry, and fewer medical visits, while still keeping patients safe. However, the misclassification of a subset of cancers underscores that AI-benign designations should trigger ongoing ultrasound follow up, known as active surveillance, rather than clinical discharge, particularly in those with intermediate- and high-risk nodules. Overall, AI characterization of ultrasound images can help doctors make better decisions and avoid unnecessary biopsies but cannot replace medical judgment. Further studies to clarify the role of AI in ultrasound imaging are needed.

— Joanna Miragaya, MD

ABBREVIATIONS & DEFINITIONS

Thyroid nodule: an abnormal growth of thyroid cells that forms a lump within the thyroid. While most thyroid nodules are non-cancerous (Benign), ~5% are cancerous.

Thyroid Ultrasound: a common imaging test used to evaluate the structure of the thyroid gland. Ultrasound uses soundwaves to create a picture of the structure of the thyroid gland and accurately identify and characterize nodules within the thyroid. Ultrasound is also frequently used to guide the needle into a nodule during a thyroid nodule biopsy.

Thyroid fine needle aspiration biopsy (FNAB): a simple procedure that is done in the doctor’s office to determine if a thyroid nodule is benign (non-cancerous) or cancer. The doctor uses a very thin needle to withdraw cells from the thyroid nodule. Patients usually return home or to work after the biopsy without any ill effects.

Table of Contents | PDF File for Saving and Printing