Background: Federated learning is a decentralized approach to machine learning; it is a training strategy that overcomes medical data privacy regulations and generalizes deep learning algorithms. Federated learning mitigates many systemic privacy risks by sharing only the model and parameters for training, without the need to export existing medical data sets. In this study, we performed ultrasound image analysis using federated learning to predict whether thyroid nodules were benign or malignant. Objective: The goal of this study was to evaluate whether the performance of federated learning was comparable with that of conventional deep learning. Methods: A total of 8457 (5375 malignant, 3082 benign) ultrasound images were collected from 6 institutions and used for federated learning and conventional deep learning. Five deep learning networks (VGG19, ResNet50, ResNext50, SE-ResNet50, and SE-ResNext50) were used. Using stratified random sampling, we selected 20% (1075 malignant, 616 benign) of the total images for internal validation. For external validation, we used 100 ultrasound images (50 malignant, 50 benign) from another institution. Results: For internal validation, the area under the receiver operating characteristic (AUROC) curve for federated learning was between 78.88% and 87.56%, and the AUROC for conventional deep learning was between 82.61% and 91.57%. For external validation, the AUROC for federated learning was between 75.20% and 86.72%, and the AUROC curve for conventional deep learning was between 73.04% and 91.04%. Conclusions: We demonstrated that the performance of federated learning using decentralized data was comparable to that of conventional deep learning using pooled data. Federated learning might be potentially useful for analyzing medical images while protecting patients’ personal information.

Background: Permanent hypoparathyroidism after total thyroidectomy is a rare but potentially serious iatrogenic complication. The aim of this study was to investigate the rate of recovery from postoperative, permanent hypoparathyroidism in patients undergoing thyroidectomy without parathyroid autotransplantation. Methods: This study was a prospective case series with a postoperative follow-up of up to 3 years. We enrolled patients with thyroid cancer who underwent total thyroidectomy with central compartment dissection, with or without lateral neck dissection, and who had postoperative permanent hypoparathyroidism, defined as serum levels of intact parathyroid hormone (PTH) <15 pg/mL for at least 1 year. In the postoperative follow-up period, the serum levels of PTH and calcium were measured regularly. Recovery from permanent hypoparathyroidism was defined as return to normal serum levels of PTH (15–65 pg/mL) and calcium (8.5–10.1 mg/dL) without calcium and/or vitamin D supplementation. Results: In the 1467 patients who underwent total thyroidectomy, 22 presented with permanent postoperative hypoparathyroidism. In 5 of these 22 patients, the PTH levels increased steadily and returned to normal in 27.6±2.9 months, after which supplementation of calcium and vitamin D could be discontinued. Conclusions: Although recovery from permanent hypoparathyroidism is rare, patients should be monitored for serum PTH levels so that unnecessary treatments such as calcium and vitamin D supplementation can be avoided.

Anaplastic thyroid cancer (ATC) constitutes less than 2% of total thyroid cancers but accounts for 20-40% of thyroid cancer-related deaths. Cancer stem cell drug resistance represents a primary factor hindering treatment. This study aimed to develop targeted agents against thyroid malignancy, focusing on individual and synergistic effects of HNHA (histone deacetylase), lenvatinib (FGFR), and sorafenib (tyrosine kinase) inhibitors. Patients with biochemically and histologically proven papillary thyroid cancer (PTC) and ATC were included. Cell samples were obtained from patients at the Thyroid Cancer Center, Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, Korea. PTC and ATC cells were treated with lenvatinib or sorafenib, alone or in combination with HNHA. Tumor-bearing mice (10/group) were administered 10 mg/kg lenvatinib (p.o.) or 40 mg/kg sorafenib (p.o.), alone or in combination with 25 mg/kg HNHA (i.p.) once every three days. Gene expression in patient-derived PTC and ATC cells was compared using a microarray approach. Cellular apoptosis and proliferation were examined by immunohistochemistry and MTT assays. Tumor volume and cell properties were examined in the mouse xenograft model. HNHA-lenvatinib combined treatment induced markers of cell cycle arrest and apoptosis and suppressed anti-apoptosis markers, epithelial-mesenchymal transition (EMT), and the FGFR signaling pathway. Combined treatment induced significant tumor shrinkage in the xenograft model. HNHA-lenvatinib combination treatment thus blocked the FGFR signaling pathway, which is important for EMT. Treatment with HNHA-lenvatinib combination was more effective than either agent alone or sorafenib-HNHA combination. These findings have implications for ATC treatment by preventing drug resistance in cancer stem cells.