Deep Learning-based framework shown as a valuable tool to determine the composition of thyroid nodules

The authors of this retrospective multicenter study proposed a deep learning-based framework to identify the composition of thyroid nodules while also assessing their malignancy risk. Their research demonstrated that convolutional neural networks (CNNs) were able to assist in the diagnosis of thyroid nodules and reduce the rate of unnecessary fine-needle aspiration.

Key points:

• Thyroid solid nodules have a high probability of malignancy.
• Our models can improve the differentiation between benign and malignant solid thyroid nodules.
• The differential performance of one model was superior to that of senior radiologists. Applying this could reduce the rate of unnecessary fine-needle aspiration of solid thyroid nodules.

Article: Deep learning to assist composition classification and thyroid solid nodule diagnosis: a multicenter diagnostic study

Authors: Chen Chen, Yitao Jiang, Jincao Yao, Min Lai, Yuanzhen Liu, Xianping Jiang, Di Ou, Bojian Feng, Lingyan Zhou, Jinfeng Xu, Linghu Wu, Yuli Zhou, Wenwen Yue, Fajin Dong & Dong Xu