ISSN 1662-4009 (online)

ESPE Yearbook of Paediatric Endocrinology (2021) 18 3.15 | DOI: 10.1530/ey.18.3.15

ESPEYB18 3. Thyroid Thyroid Cancer (1 abstracts)

3.15. Lymph node metastasis prediction of papillary thyroid carcinoma based on transfer learning radiomics

Yu J, Deng Y, Liu T, Zhou J, Jia X, Xiao T, Zhou S, Li J, Guo Y, Wang Y, Zhou J & Chang C

Nat Commun. 2020;11:4807. doi: 10.1038/s41467-020-18497-3.

This interesting paper illustrates the potential of machine learning to improve the sensitivity and specificity of routine techniques in clinical practice if large cohorts for training and validation of models are used. Yu et al. present a transfer learning model (machine learning approach to solve research problems by reusing information from previously learned tasks for the learning of new tasks) in radiology of lymph node metastasis in papillary thyroid cancer patients.

Lymph node metastasis (LNM) are a major concern in patients with papillary thyroid carcinoma (PTC). While preoperative radiologic detection rate by ultrasound of lateral cervical LNM is reliable, identification rate of LNM in the central cervical area is as low as 30%. These authors studied a main cohort of 1013 PTC cases with unifocal lesions and a second dataset of 368 PTC patients with multifocal lesions from the same hospital. Finally, a third independent data set came from two further hospitals with 513 PTC patients with unifocal lesions. Within the datasets, the authors used different ultrasound machines and the results of three independent radiologists. They compared sensitivity by ROC curves and probability threshold curves and showed high average area under the curve (AUC) of 0.90 in the main cohort, and 0.93 in the two validation cohorts using their transfer learning model for LNM detection, showing clear advantage over routinely used methods (clinical statistical model, traditional radiological model, non-transfer learning model).

Over the last year, such radiomics approaches have also been developed for thyroid nodule diagnostics in adult as well as very recently for paediatric patients [1, 2].

Reference: 1. Peng S, Liu Y, Lv W, Liu L, Zhou Q, Yang H, Ren J, Liu G, Wang X, Zhang X, Du Q, Nie F, Huang G, Guo Y, Li J, Liang J, Hu H, Xiao H, Liu Z, Lai F, Zheng Q, Wang H, Li Y, Alexander EK, Wang W, Xiao H. Deep learning-based artificial intelligence model to assist thyroid nodule diagnosis and management: a multicentre diagnostic study. Lancet Digit Health. 2021 Apr;3(4):e250–e259. doi: 10.1016/S2589-7500(21)00041-8.2. Radebe L, van der Kaay DCM, Wasserman JD, Goldenberg A. Predicting Malignancy with Pediatric Thyroid Nodules: Early Experience in Machine Learning for Clinical Decision Support. J Clin Endocrinol Metab. 2021 Jun 23:dgab435. doi: 10.1210/clinem/dgab435.

Article tools

My recent searches

No recent searches.