Researchers proposed a 2-stage method incorporating pipelines of thyroid nodules segmentation and classification of medullary thyroid carcinoma (MTC), individually, as published in Medical Physics. Their method achieves accurate segmentation of thyroid nodules and facilitates the recognition of MTC better than experienced doctors, they said.

MTC is a rare type of cancer but the second most malignant thyroid cancer with a high rate of mortality. It is, therefore, of great importance to accurately recognize it. Yet, even experienced experts often fail to distinguish it from other thyroid nodules based on ultrasound images.

A team of researchers led by Liqin Huang from the College of Physics and Information Engineering at Fuzhou University in China presented a computer-aided method that can help recognize MTC on ultrasound images and help them differentiate from benign nodules and papillary thyroid carcinoma, the most common type of thyroid cancer.

Continue Reading

Read more about MTC diagnosis

“Our method is a two-stage schema with two important components including a cascaded coarse-to-fine segmentation network and a knowledge-based classification network,” the researchers wrote. They tested their method in 248 medullary thyroid carcinoma samples, 240 benign nodule samples, and 239 papillary thyroid carcinoma samples.

They reported that their cascaded segmentation network reached a dice similarity coefficient of 0.776, an intersection over union value of 0.689, 0.778 precision, and 0.821 recall for thyroid nodule segmentation suggesting a performance enhancement on thyroid nodule segmentation.

Moreover, by incorporating prior knowledge, the method achieved a mean accuracy of 82.1% in classifying MTC nodules, papillary thyroid carcinoma nodules, and benign nodules. The researchers reported that their method showed higher performance in recognizing MTC with an accuracy of 86.8%, which is superior to the diagnostic accuracy of experienced doctors at around 70%.

“The improvement introduced by our designed component shows that our method exhibits great performance on the task of MTC detection,” the authors concluded.


Pan L, Cai Y, Lin N, Yang L, Zheng S, Huang L. A two-stage network with prior knowledge guidance for medullary thyroid carcinoma recognition in ultrasound images. Med Phys. Published online February 1, 2022. doi:10.1002/mp.15492