Deep-Learning Algorithm for Indeterminate Thyroid Nodules: Comparison and Combination with Molecular Testing
Abstract
Thyroid nodules with indeterminate cytology (Bethesda III) carry an intermediate risk of malignancy, and molecular testing is widely used to decide between surgery and surveillance, but it is costly and not universally available. We investigated whether a convolutional neural network (CNN) applied to routine ultrasound images can stratify Bethesda III nodules, and how it compares and combines with molecular testing. We retrospectively identified 466 nodule records from 418 patients who underwent molecular testing (Afirma or ThyroSeq). A CNN pre-trained on determinate nodules was adapted to indeterminate nodules with soft labels and transfer learning. After removal of duplicate records, 119 nodules with surgical pathology were used for development and 47 were held out for testing. On the test set, the deep learning (DL) model achieved an AUC of 0.73 (95% CI 0.51-0.94), compared with 0.55 (0.38-0.71) for molecular testing (P = 0.15). At a cut-off inherited from the original algorithm, the DL model was more specific than molecular testing (0.95 vs 0.32; P < 0.001), with a higher positive predictive value (0.67 vs 0.21), a lower sensitivity (0.44 vs 0.78; P = 0.25) and a similar negative predictive value (0.88 vs 0.86). All six DL-positive test nodules were molecular-positive, and the DL output split the molecular-positive nodules into a high-risk (4/6 malignant) and a lower-risk group (3/27 malignant; P = 0.01). Surgical verification was strongly driven by the molecular result (99% of positive vs 19% of negative results), and molecular-test accuracy depended on the platform. A logistic regression of the DL output, nodule size and the molecular result dropped the DL term and did not improve on the DL output alone. Ultrasound DL complemented molecular testing for Bethesda III nodules, in particular by identifying molecular-positive nodules at lower risk of malignancy.
Related
- TRIDENT (Taxonomic Resolution and IDentification using Environmental dNa Traces): An Optimized Algorithm for Vertebrate Taxonomic Assignments in eDNA Metabarcoding, Integrating Molecular, Taxonomic, and Ecological Criteria
- Deep learning-based detection of construction waste in complex scenarios with an improved lightweight algorithm
- Integrating hippopotamus optimization algorithm into machine learning frameworks for predicting suspended sediment concentrations
- Revisiting Identification Thresholds of Specific Learning Disorders ( SLD ) in Gifted Children ( GC ). A Comparison of Three Groups: Twice‐Exceptional ( TE ) Children, GC Without SLD and Non‐ GC With SLD
- A redundant encoding algorithm for artificial sensory information speeds learning and improves multisensory-guided navigation
BibTeX
@misc{yang2026deeplearning,
title = {{Deep-Learning Algorithm for Indeterminate Thyroid Nodules: Comparison and Combination with Molecular Testing}},
author = {Jichen Yang and Jeffrey Ashton and Jikai Zhang and Maciej A. Mazurowski and Benjamin Wildman-Tobriner},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.05400},
eprint = {2611.05400},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.05400},
}
RIS
TY - UNPB
TI - Deep-Learning Algorithm for Indeterminate Thyroid Nodules: Comparison and Combination with Molecular Testing
AU - Jichen Yang
AU - Jeffrey Ashton
AU - Jikai Zhang
AU - Maciej A. Mazurowski
AU - Benjamin Wildman-Tobriner
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.05400
UR - https://arxiv.org/abs/2611.05400
AB - Thyroid nodules with indeterminate cytology (Bethesda III) carry an intermediate risk of malignancy, and molecular testing is widely used to decide between surgery and surveillance, but it is costly and not universally available. We investigated whether a convolutional neural network (CNN) applied to routine ultrasound images can stratify Bethesda III nodules, and how it compares and combines with molecular testing. We retrospectively identified 466 nodule records from 418 patients who underwent molecular testing (Afirma or ThyroSeq). A CNN pre-trained on determinate nodules was adapted to indeterminate nodules with soft labels and transfer learning. After removal of duplicate records, 119 nodules with surgical pathology were used for development and 47 were held out for testing. On the test set, the deep learning (DL) model achieved an AUC of 0.73 (95% CI 0.51-0.94), compared with 0.55 (0.38-0.71) for molecular testing (P = 0.15). At a cut-off inherited from the original algorithm, the DL model was more specific than molecular testing (0.95 vs 0.32; P < 0.001), with a higher positive predictive value (0.67 vs 0.21), a lower sensitivity (0.44 vs 0.78; P = 0.25) and a similar negative predictive value (0.88 vs 0.86). All six DL-positive test nodules were molecular-positive, and the DL output split the molecular-positive nodules into a high-risk (4/6 malignant) and a lower-risk group (3/27 malignant; P = 0.01). Surgical verification was strongly driven by the molecular result (99% of positive vs 19% of negative results), and molecular-test accuracy depended on the platform. A logistic regression of the DL output, nodule size and the molecular result dropped the DL term and did not improve on the DL output alone. Ultrasound DL complemented molecular testing for Bethesda III nodules, in particular by identifying molecular-positive nodules at lower risk of malignancy.
ER -