A proof-of-concept investigation into predicting follicular carcinoma on ultrasound using topological data analysis and radiomics

Andrew M. Thomas; Ann C. Lin; Grace Deng; Yuchen Xu; Gustavo Fernandez Ranvier; Aida Taye; David S. Matteson; Denise Lee

doi:10.1556/1647.2025.00256

Back

A proof-of-concept investigation into predicting follicular carcinoma on ultrasound using topological data analysis and radiomics

Journal article

Open access

Peer reviewed

A proof-of-concept investigation into predicting follicular carcinoma on ultrasound using topological data analysis and radiomics

Andrew M. Thomas, Ann C. Lin, Grace Deng, Yuchen Xu, Gustavo Fernandez Ranvier, Aida Taye, David S. Matteson and Denise Lee

Imaging (Budapest), Vol.17(1), pp.39-48

06/26/2025

DOI: 10.1556/1647.2025.00256

Files and links (1)

url

https://doi.org/10.1556/1647.2025.00256View

Published (Version of record) Open Access

Abstract

Background Sonographic risk patterns identified in established risk stratification systems (RSS) may not accurately stratify follicular carcinoma from adenoma, which share many similar US characteristics. The purpose of this study is to investigate the performance of a multimodal machine learning model utilizing radiomics and topological data analysis (TDA) to predict malignancy in follicular thyroid neoplasms on ultrasound. Patients & Methods This is a retrospective study of patients who underwent thyroidectomy with pathology confirmed follicular adenoma or carcinoma at a single academic medical center between 2010 and 2022. Features derived from radiomics and TDA were calculated from processed ultrasound images and high-dimensional features in each modality were projected onto their first two principal components. Logistic regression with L2 penalty was used to predict malignancy and performance was evaluated using leave-one-out cross-validation and area under the curve (AUC). Results Patients with follicular adenomas (n = 7) and follicular carcinomas (n = 11) with available imaging were included. The best multimodal model achieved an AUC of 0.88 (95% CI: [0.85, 1]), whereas the best radiomics model achieved an AUC of 0.68 (95% CI: [0.61, 0.84]). Conclusions We demonstrate that inclusion of topological features yields strong improvement over radiomics-based features alone in the prediction of follicular carcinoma on ultrasound. Despite low volume data, the TDA features explicitly capture shape information that likely augments performance of the multimodal machine learning model. This approach suggests that a quantitative based US RSS may contribute to the preoperative prediction of follicular carcinoma.

topological data analysis

follicular carcinoma

multimodal

thyroid nodule

machine learning

Details

Title: Subtitle: A proof-of-concept investigation into predicting follicular carcinoma on ultrasound using topological data analysis and radiomics
Creators: Andrew M. Thomas - University of Iowa
Ann C. Lin - Icahn School of Medicine at Mount Sinai
Grace Deng - Cornell University
Yuchen Xu - Cornell University
Gustavo Fernandez Ranvier - Icahn School of Medicine at Mount Sinai
Aida Taye - Icahn School of Medicine at Mount Sinai
David S. Matteson - Cornell University
Denise Lee - Icahn School of Medicine at Mount Sinai
Resource Type: Journal article
Publication Details: Imaging (Budapest), Vol.17(1), pp.39-48
DOI: 10.1556/1647.2025.00256
ISSN: 2732-0960
eISSN: 2732-0960
Publisher: AKADEMIAI KIADO ZRT
Grant note: NSF Grant: DMS-2114143
This article was supported in part by NSF Grant DMS-2114143.
Language: English
Electronic publication date: 02/11/2025
Date published: 06/26/2025
Academic Unit: Statistics and Actuarial Science
Record Identifier: 9984790977402771

Metrics

13 Record Views

See more details