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.

Moon JH, et al. Endocrinol Metab. 2018 Jun;33(2):278-286. https://doi.org/10.3803/EnM.2018.33.2.278

Background: Transoral endoscopic thyroidectomy vestibular approach (TOETVA) is a scarless alternative to open thyroidectomy (OT). This systematic review and meta-analysis aimed to synthesize evidence comparing the intraoperative and postoperative outcomes of TOETVA and OT. Methods: A systematic literature search of PubMed, Web of Science, the Cochrane Library, and Google Scholar was performed to identify studies comparing the outcomes of TOETVA and OT published before February 2023. The outcomes of interest were operative time, intraoperative blood loss, hospital stay, postoperative pain, number of central lymph nodes retrieved, number of metastatic central lymph nodes, and incidences of transient and permanent recurrent laryngeal nerve injury, transient and permanent hypocalcemia, hematoma, and infection. Results: Thirteen studies published between 2016 and 2022, involving a total of 2889 patients (TOETVA, n=1085; OT, n=1804) were included in this systematic review and meta-analysis. Meta-analysis showed that the TOETVA group had a significantly longer overall operative time (weighted mean difference [WMD] 55.19; 95% CI, 39.15, 71.23; P<0.001), longer hospital stay (WMD, 0.27; 95% CI, 0.14, 0.39; P<0.001), and lower pain scores on postoperative day 1 (WMD, −1.41; 95% CI, −2.79, −0.03; P=0.04) than the OT group. Other intraoperative and postoperative outcomes were not significantly different between the groups. Conclusion: TOETVA has a similar safety profile to OT with less postoperative pain, making it an appropriate and more cosmetically appealing alternative to OT for select patients.