Thyroid Biology and Disease

SAT-405 - Can Artificial Intelligence Rule Out Malignancy in Incidental Thyroid Nodules Detected on Computed Tomography?

Saturday, June 13, 2026

9:30 AM - 4:30 PM CT

Location: ENDOExpo; Poster Floor

Abstract Poster Presenter(s)

Luis David Vilatuna, MD

Visiting Research Fellow
Mayo Clilnic
Sangolqui, Ecuador

Co-Author(s)

FR

Fazle Rafsani, n/a

School of Computing and Augmented Intelligence, Arizona State University
Arizona, United States
Ana Cristina Proano, MD

PGY-1
ADVOCATE IL MASONIC MEDICAL CENTER
Chicago, IL, United States
SB

Shubhangi Bagewadi, MBBS

Louisiana State University Health Sciences Center Shreveport-Monroe
Shreveport, United States
AC

Ana Gabriela Claros, MD

Mayo Clinic
Rochester, Minnesota, United States
ML

Maria Lizarazo Jimenez, MD

Care and AI Laboratory, Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic
Rochester, United States
HD

Hari Varshan Dharmendra Mohan Prabu, n/a

School of Computing and Augmented Intelligence, Arizona State University
Tempe, Arizona, United States
IA

Ivan Mateo Alzamora, MD

Care and AI Laboratory, Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic
Rochester, United States
MM

Marcelo Montero, MD

Research Trainee
Care and AI Laboratory, Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic
Rochester, Minnesota, United States
EG

Esteban Gomez Alvarez, MD

Care and AI Laboratory, Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic
Rochester, United States
KG

Kerly Guevara maldonado, MD

Research Fellow
Mayo Clinic
Rochester, Minnesota, United States
Felipe Larios Madrid, MD

Research Fellow
Care and AI Laboratory, Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic
Lima, Peru
David Toro Tobon, MD (he/him/his)

Principal Investigator
Mayo Clinic
Rochester, Minnesota, United States
MA

Misk Al Zahidy, MS

Care and AI Laboratory, Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic
Rochester, United States
MB

Megan E. Branda, MS

Principal Biostatistician
Mayo Clinic
Rochester, Minnesota, United States

Senior Author/Primary Investigator(s)

TW

Teresa Wu, n/a

Arizona State University
Tempe, Arizona, United States
Juan P. Brito, MD

Care and AI Laboratory, Knowledge and Evaluation Research Unit, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Medicine, Mayo Clinic
Rochester, MN, United States

Disclosure(s):

Luis David Vilatuna, MD: No financial relationships to disclose

Ana Cristina Proano, MD: No financial relationships to disclose

Felipe Larios Madrid, MD: No financial relationships to disclose

David Toro Tobon, MD: Immunovant Inc: Consultant ()

Background
Incidental thyroid nodules (ITNs) detected on CT are common and are a major driver of downstream imaging, biopsy, and overdiagnosis of thyroid cancer. Once an ITN prompts ultrasound, additional nodules are often identified, triggering further surveillance or invasive testing with limited clinical benefit. Tools to identify ITNs unlikely to be malignant at the time of CT detection could help clinicians safely avoid unnecessary follow-up and interrupt overuse and overtreatment.
Objective
To evaluate whether an artificial intelligence model can support a rule-out strategy for malignancy in CT-detected ITNs.
Methods
We retrospectively identified CT-detected ITNs (2017–2023) across four Mayo Clinic sites, including patients with ≥12 months follow-up and no prior thyroid imaging, thyroid disease, cancer, or thyroid-related symptoms. Nodules were matched across CT studies and reference standards using size, location, and morphology. Nodule status was classified as benign or malignant based on surgical histopathology or fine-needle aspiration Bethesda classification; ultrasound-based Mayo Clinic risk stratification was used only for nodules with benign features when pathology was unavailable. Indeterminate nodules were excluded. CT slices were cropped to the nodule region and manually segmented by experts in ITK-SNAP; inter-annotator agreement was assessed using Dice coefficient. BioMedCLIP, an image encoder, was fine-tuned with this data and evaluated through a five-fold cross-validation technique with an 80/20 train–test split. Model performance was assessed using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy.
Results
The cohort included 381 patients (62% female; mean age 67.3 ± 15.5 years), mean nodule size was 2.1 ± 1.5 cm and 38% were malignant; 60% of CT scans were contrast-enhanced; the most common examinations were chest CT with and without intravenous contrast (21% and 17%) and neck CT with contrast (14%). In the test sets, BioMedCLIP achieved a sensitivity of 78.6%, specificity of 48.9%, PPV of 47.8%, NPV of 79.3%, and overall accuracy of 60.0%.
Conclusion
In a large, heterogeneous real-world cohort, BioMedCLIP showed moderate NPV for malignancy in CT-detected ITNs, supporting its potential role as a rule-out tool to guide further thyroid imaging. Although overall discrimination was modest, the ability to identify nodules unlikely to be malignant aligns with the clinical goal of reducing unnecessary downstream testing. These findings, derived from routine practice rather than curated datasets, better reflect real-world variability. Future work will expand the dataset, refine rule-out thresholds, and integrate additional contextual data such as radiology reports.

*Unless otherwise noted, all abstracts presented at ENDO must not be released to the press or the public until the date and time of presentation. For oral presentations, the abstracts are embargoed until the session begins. The Endocrine Society reserves the right to lift the embargo on specific abstracts that are selected for promotion prior to or during ENDO.*