Related Experiment Video
Updated: Nov 3, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
2.1K
Indeterminate thyroid cytology: detecting malignancy using analysis of nuclear images
Caroline Y Hayashi1, Danilo T A Jaune1, Cristiano C Oliveira2
1Department of Internal Medicine, Botucatu Medical School, Sao Paulo State University (Unesp), Botucatu, São Paulo, Brazil.
Endocrine Connections
|June 2, 2021
Summary
Computerized analysis of nuclear images (CANI) accurately classifies indeterminate thyroid nodules. This method improves diagnostic accuracy for atypical or follicular lesions, aiding in malignancy assessment.
Area of Science:
- Medical Imaging
- Computational Pathology
- Oncology
Background:
- Thyroid nodules classified as atypia of undetermined significance/follicular lesion of undetermined significance (AUS/FLUS) or follicular neoplasm/suspected follicular neoplasm (FN/SFN) pose diagnostic challenges.
- Computerized analysis of nuclear images (CANI) offers a potential tool for improving the classification of these indeterminate thyroid cytology cases.
Purpose of the Study:
- To evaluate the efficacy of CANI in correctly classifying AUS/FLUS and FN/SFN thyroid nodules for malignancy.
- To assess the diagnostic accuracy of CANI in differentiating benign from malignant thyroid lesions within these indeterminate categories.
Main Methods:
- Analysis of 101 thyroid nodules (68 AUS/FLUS, 33 FN/SFN) from 97 patients using ImageJ software for morphometric and texture parameters of follicular cell nuclei.
- Application of classification and regression trees (gini model) to evaluate the predictive capacity of nuclear parameters for malignancy.
- Assessment of method reproducibility for 12 nuclear parameters using the intraclass coefficient of correlation.
Main Results:
- Significant differences between benign and malignant nodules were observed in entropy (AUS/FLUS) and fractal analysis, coefficient of variation of roughness, and CV-entropy (FN/SFN).
- CANI achieved high classification accuracy, correctly identifying 90.0% of malignant AUS/FLUS nodules and 100.0% of malignant FN/SFN nodules.
- Overall classification accuracy was 94.1% for AUS/FLUS and 97.0% for FN/SFN nodules, with substantial to nearly complete reproducibility for most parameters.
Conclusions:
- CANI demonstrates a high capacity for accurately classifying indeterminate thyroid nodules (AUS/FLUS and FN/SFN) for malignancy.
- This computational method shows promise as a valuable tool to enhance diagnostic accuracy in cases of indeterminate thyroid cytology.
- The findings support the integration of CANI into clinical practice for better thyroid nodule management.
