Related Experiment Video
Updated: Oct 22, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
2.1K
False Negative Rates in Benign Thyroid Nodule Diagnosis: Machine Learning for Detecting Malignancy.
Alexander J Idarraga1, George Luong1, Vivian Hsiao1
1University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin.
The Journal of Surgical Research
|August 31, 2021
Summary
Machine learning can predict false negative fine-needle aspiration (FNA) results for thyroid nodules. A Random Forest model showed promise in identifying malignancy, aiding earlier cancer treatment.
Area of Science:
- Endocrinology
- Oncology
- Medical Informatics
Background:
- Thyroid nodules are highly prevalent, with most being benign.
- Fine-needle aspiration (FNA) cytology is key for malignancy risk assessment.
- False negative FNA diagnoses can delay critical thyroid cancer treatment.
Purpose of the Study:
- To develop a machine learning model for identifying false negative FNA results.
- To predict malignancy risk using non-invasive clinical data.
Main Methods:
- Retrospective review of thyroid nodule cases with ultrasound and FNA.
- Utilized linear, non-linear, and ensemble machine learning models (scikit-learn).
- Employed 10-fold cross-validation with repetition, comparing models via AUROC.
Main Results:
- A total of 604 subjects were included, with 38 malignancies.
- The Random Forest model achieved the highest AUROC (0.64) for malignancy prediction.
- Model performance improvement over other algorithms was not statistically significant.
Conclusions:
- A Random Forest model can predict thyroid nodule malignancy better than chance.
- Model thresholds can be adjusted to balance false positives and negatives.
- Prospective studies are needed to validate the model's performance.
