ThyroUS-Net: Hybrid CNN with Squeeze-Excitation for Thyroid Nodule Malignancy Classification from Ultrasound

Fahima akter nila^1, and Rafsana Ferdouse²

¹Master’s of public health (Biostatistics and epidemiology) Monroe university. USA

²Master’s of Public Health (New Rochelle Subject: MPH) University: king Graduate School, Monroe University. USA

Cite this paper:
Fahima akter nila and Rafsana Ferdouse. ThyroUS-Net: Hybrid CNN with Squeeze-Excitation for Thyroid Nodule Malignancy Classification from Ultrasound. American Journal of Public Health Research. 2026; 14(2):44-51. doi: 10.12691/ajphr-14-2-5

Abstract

The prevalence of thyroid nodules in general population is very high, but only a small percentage of the nodules become malignant; thus, a proper risk stratification is necessary to prevent unnecessary biopsies and prompt cancer diagnosis. The most common diagnostic modality used to examine the thyroid is ultrasound imaging, but it is a subjective form of interpretation that is prone to inter-observer variability despite standardized systems like ACR TI-RADS. Despite convolutional neural networks (CNNs) achieving encouraging outcomes in the classification of thyroid ultrasounds, most of the current models do not incorporate hybrid multi-scale feature extraction and effective attention in order to improve the discriminative effectiveness. The paper presents a hybrid CNN model called ThyroUS-Net to achieve superior/significantly improved classification of thyroid nodule malignancy. The model combines shallow and deep convolutional networks to extract local texture and global morphological information, and SE blocks are used to recalibrate channels channelwise to highlight the diagnostically meaningful information. The network was tested on the Thyroid Digital Image Database (that is publicly accessible: 428 ultrasound images: 357 benign, 71 malignant). Untrained data augmentation, standard preprocessing, and stratified train-validation-test splitting were used to guarantee sound evaluation. According to the results of the experiment, ThyroUS-Net is superior to the baseline models, such as simple CNN, VGG-16, or ResNet-50, with accuracy, precision, recall, and F1 score of 95, 98.4, 97.1, and AUC of 0.97, respectively. The ablation analysis proved that both the hybrid backbone and SE attention module were significant in boosting performance. Such results indicate that ThyroUS-Net has better diagnostic consistency and better malignancy discrimination, which can be used as a clinical assistance system. The future research will involve multi-center validation, TI-RADS scoring integration, and explainable AI techniques to promote clinical adoption.

Keywords:
Thyroid nodule classification Ultrasound imaging Hybrid CNN Squeeze-Excitation network Deep learning in healthcare Computer-aided diagnosis

This work is licensed under a Creative Commons Attribution 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

References:

[1]	M. Uludag et al., “Management of Thyroid Nodules,” Sisli Etfal Hastanesi Tip Bulteni, vol. 57, no. 3, pp. 287–304, 2023.
[2]	H. A. Al-Hakami, R. Alqahtani, A. Alahmadi, D. Almutairi, M. Algarni, and T. Alandejani, “Thyroid Nodule Size and Prediction of Cancer: A Study at Tertiary Care Hospital in Saudi Arabia,” Cureus, vol. 12, no. 3, Mar. 2020.
[3]	D. W. Chen and M. R. Haymart, “Unravelling the rise in thyroid cancer incidence and addressing overdiagnosis,” Nature Reviews Endocrinology, Sep. 2025.
[4]	D. Wang, C. Xie, X. Zheng, and M. Li, “Diagnostic accuracy of ultrasound in hyperthyroidism: A comprehensive review of recent studies,” Journal of Radiation Research and Applied Sciences, vol. 18, no. 2, p. 101370, Feb. 2025.
[5]	M. Itani, R. Assaker, M. Moshiri, T. J. Dubinsky, and Manjiri Dighe, “Inter-observer Variability in the American College of Radiology Thyroid Imaging Reporting and Data System: In-Depth Analysis and Areas for Improvement,” vol. 45, no. 2, pp. 461–470, Feb. 2019.
[6]	Ibomoiye Domor Mienye, T. G. Swart, G. Obaido, M. Jordan, and P. Ilono, “Deep Convolutional Neural Networks in Medical Image Analysis: A Review,” Information, vol. 16, no. 3, p. 195, Mar. 2025.
[7]	Y. Zhu, P. Jin, J. Bao, Q. Jiang, and X. Wang, “Thyroid ultrasound image classification using a convolutional neural network,” Annals of Translational Medicine, vol. 9, no. 20, pp. 1526–1526, Oct. 2021.
[8]	M. Ennab and H. Mcheick, “Enhancing interpretability and accuracy of AI models in healthcare: a comprehensive review on challenges and future directions,” Frontiers in Robotics and AI, vol. 11, Nov. 2024.
[9]	Y. Xu, M. Xu, Z. Geng, J. Liu, and B. Meng, “Thyroid nodule classification in ultrasound imaging using deep transfer learning,” BMC Cancer, vol. 25, no. 1, Mar. 2025.
[10]	M. Savelonas, “An Overview of AI-Guided Thyroid Ultrasound Image Segmentation and Classification for Nodule Assessment,” Big Data and Cognitive Computing, vol. 9, no. 10, p. 255, Oct. 2025.
[11]	J. Chi, E. Walia, P. Babyn, J. Wang, G. Groot, and M. Eramian, “Thyroid Nodule Classification in Ultrasound Images by Fine-Tuning Deep Convolutional Neural Network,” Journal of Digital Imaging, vol. 30, no. 4, pp. 477–486, Jul. 2017.
[12]	M. Guo, “An intelligent diagnosis method for thyroid nodules using UNet++ integrated with ResNet and transformer,” Discover Artificial Intelligence, vol. 6, no. 1, Dec. 2025.
[13]	K. K. Kumar and P. S. Shelokar, “An SVM method using evolutionary information for the identification of allergenic proteins,” Bioinformation, vol. 2, no. 6, pp. 253–256, Jan. 2008.
[14]	N. Bouchekout et al., “A novel hybrid deep learning and chaotic dynamics approach for thyroid cancer classification,” Scientific Reports, vol. 15, no. 1, pp. 40471–40471, Nov. 2025.
[15]	A. I. Aramendia, “The U-Net: A Complete Guide,” Medium, Feb. 01, 2024. https://medium.com/@alejandro.itoaramendia/decoding-the-u-net-a-complete-guide-810b1c6d56d8.
[16]	J. Hu, L. Shen, S. Albanie, G. Sun, and E. Wu, “Squeeze-and-Excitation Networks,” Sep. 2017.
[17]	Z. Ullah, M. Hong, T. Mahmood, and J. Kim, “Systematic Integration of Attention Modules into CNNs for Accurate and Generalizable Medical Image Classification,” Mathematics, vol. 13, no. 22, p. 3728, Nov. 2025.
[18]	L. Chen et al., “ThyroidNet: A Deep Learning Network for Localization and Classification of Thyroid Nodules,” Computer Modeling in Engineering & Sciences, vol. 139, no. 1, pp. 361–382, Dec. 2023.
[19]	Y. Fu, L. Guo, and F. Huang, “A lightweight CNN model for pepper leaf disease recognition in a human palm background,” Heliyon, vol. 10, no. 12, pp. e33447–e33447, Jun. 2024.
[20]	K. Bahmane, S. Bhattacharya, and A. B. Chaouki, “Evaluation of a Hybrid CNN Model for Automatic Detection of Malignant and Benign Lesions,” Medicina, vol. 61, no. 11, p. 2036, Nov. 2025.
[21]	S.-H. Tsang, “Review: SENet — Squeeze-and-Excitation Network, Winner of ILSVRC 2017 (Image Classification),” Medium, May 08, 2019. https://medium.com/data-science/review-senet-squeeze-and-excitation-network-winner-of-ilsvrc-2017-image-classification-a887b98b2883 (accessed Feb. 25, 2026).
[22]	A. Erdoğan, “Squeeze-and-Excitation Networks,” Medium, Oct. 16, 2022. https://medium.com/@atakanerdogan305/squeeze-and-excitation-networks-c4e1ad7d8a3d.
[23]	X. Wang et al., “ThyroNet-X4 genesis: an advanced deep learning model for auxiliary diagnosis of thyroid nodules’ malignancy,” Scientific Reports, vol. 15, no. 1, Feb. 2025.
[24]	S. R. Shah, S. Qadri, H. Bibi, S. M. W. Shah, M. I. Sharif, and F. Marinello, “Comparing Inception V3, VGG 16, VGG 19, CNN, and ResNet 50: A Case Study on Early Detection of a Rice Disease,” Agronomy, vol. 13, no. 6, p. 1633, Jun. 2023.
[25]	A. Rashed, T. Medhat, and A. Elgarayhi, “Enhancing automatic diagnosis of thyroid nodules from ultrasound scans leveraging deep learning models,” Scientific Reports, vol. 15, no. 1, Nov. 2025.
[26]	M. Hirokawa et al., “Application of deep learning as an ancillary diagnostic tool for thyroid FNA cytology,” Cancer Cytopathology, Dec. 2022.
[27]	G. Naga Sujini and Sivadi Balakrishna, “Automated thyroid nodule classification in ultrasound imaging using a hybrid vision transformer and Wasserstein GAN with gradient penalty,” Scientific Reports, vol. 15, no. 1, Nov. 2025.