The impact of deep learning on diagnostic performance in the differentiation of benign and malignant thyroid nodules
Medical Ultrasonography · 2024
DOI: 10.11152/mu-4432 ↗Kurz gefasst
Fachartikel in der Zeitschrift Medical Ultrasonography, 2024. Verfasst von Esat Kaba und 9 weiteren Autor:innen. Beteiligte Einrichtung: Recep Tayyip Erdogan University, Department of Radiology, Rize. esatkaba04@gmail.com. Hauptthemen laut MEDLINE: Thyroid Nodule, Deep Learning, Thyroid Neoplasms, Image Interpretation, Computer-Assisted. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
Aims
This study aims to use deep learning (DL) to classify thyroid nodules as benign and malignant with ultrasonography (US). In addition, this study investigates the impact of DL on the diagnostic success of radiologists with different experiences. Material and methods: This study included 576 US images of thyroid nodules. The dataset was divided into 80% training and 20% test sets. Four radiologists with different levels of experience classified the images in the test set as benign-malignant. A DL model was then trained with the train set and predicted benign-malignant for the test set. Then, the output of the DL model for each nodule in the test set was presented to 4 radiologists, who were asked to make a benign-malignant classification again considering these DL results.
Results
The accuracy of the DL model was 0.9391. The accuracy for junior resident (JR) 1, JR 2, senior resident (SR), and senior radiologist (Srad) before DL-assisting were 0.7043, 0.7826, 0.8435, and 0.8522 respectively. The accuracy in DL-assisted classifications was 0.9130, 0.8696, 0.9304, and 0.9043 for JR 1, JR2, SR, and Srad, respectively. DL assistance changed the decisions of less experienced radiologists more than more experienced radiologists.
Conclusion
The DL model has superior accuracy in classifying thyroid nodules as benign-malignant with US images than radiologists with different levels of experience. Additionally, all radiologists, and most notably less experienced radiology residents, increased their accuracy in DL-assisted predictions.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Recep Tayyip Erdogan University, Department of Radiology, Rize. esatkaba04@gmail.com.
- Recep Tayyip Erdogan University, Department of Radiology, Rize.
- Osmaniye Korkut Ata University, Department of Electrical and Electronics Engineering, Osmaniye, Kahramanmaraş Sütçü İmam University, Department of Electrical and Electronics Engineering, Kahramanmaraş.
- Kahramanmaraş Sütçü İmam University, Department of Electrical and Electronics Engineering, Kahramanmaraş.
- Samsun University, Department of Radiology, Samsun.
Zitieren
APA
Kaba, E., Solak, M., Varlık, A. T., Çubukçu, Y., Sağır, L., Sünnetci, K. M., Alkan, A., Gündoğdu, H., Çeliker, F. B., & Beyazal, M. (2024). The impact of deep learning on diagnostic performance in the differentiation of benign and malignant thyroid nodules. Medical Ultrasonography. https://doi.org/10.11152/mu-4432
Vancouver
Kaba E, Solak M, Varlık AT, Çubukçu Y, Sağır L, Sünnetci KM, et al. The impact of deep learning on diagnostic performance in the differentiation of benign and malignant thyroid nodules. Medical Ultrasonography. 2024. doi:10.11152/mu-4432
BibTeX
@article{Kaba2024,
author = {Kaba, Esat and Solak, Merve and Varlık, Ayşenur Topçu and Çubukçu, Yusuf and Sağır, Lütfullah and Sünnetci, Kubilay Muhammed and Alkan, Ahmet and Gündoğdu, Hasan and Çeliker, Fatma Beyazal and Beyazal, Mehmet},
title = {The impact of deep learning on diagnostic performance in the differentiation of benign and malignant thyroid nodules},
journal = {Medical Ultrasonography},
year = {2024},
doi = {10.11152/mu-4432}
}
EUVIMED