An Optimization Method for Non-IID Federated Learning Based on Deep Reinforcement Learning.

Federated learning (FL) is a distributed machine learning paradigm that enables a large number of clients to collaboratively train models without sharing data. However, when the private dataset between clients is not independent and identically distributed (non-IID), the local training objective is inconsistent with the global training objective, which possibly causes the convergence speed of FL to slow down, or even not converge. In this paper, we design a novel FL framework based on deep reinforcement learning (DRL), named FedRLCS. In FedRLCS, we primarily improved the greedy strategy and action space of the double DQN (DDQN) algorithm, enabling the server to select the optimal subset of clients from a non-IID dataset to participate in training, thereby accelerating model convergence and reaching the target accuracy in fewer communication epochs. In simulation experiments, we partition multiple datasets with different strategies to simulate non-IID on local clients. We adopt four models (LeNet-5, MobileNetV2, ResNet-18, ResNet-34) on the four datasets (CIFAR-10, CIFAR-100, NICO, Tiny ImageNet), respectively, and conduct comparative experiments with five state-of-the-art non-IID FL methods. Experimental results show that FedRLCS reduces the number of communication rounds required by 10-70% with the same target accuracy without increasing the computation and storage costs for all clients.

Effectiveness of S-1-Based Chemoradiotherapy in Patients 70 Years and Older With Esophageal Squamous Cell Carcinoma: A Randomized Clinical Trial.

Importance: Double-agent intravenous chemotherapy concurrent with radiotherapy is the standard of care for patients with inoperable esophageal cancer. However, patients tend to tolerate intravenous chemotherapy less well with age and comorbidities. It is essential to find a better treatment modality that improves survival outcomes without reducing the quality of life. Objective: To evaluate the effectiveness of simultaneous integrated boost radiotherapy (SIB-RT) with concurrent and consolidated oral S-1 chemotherapy for patients aged 70 years and older with inoperable esophageal squamous cell carcinoma (ESCC). Design, Setting, and Participants: This multicenter, phase III randomized clinical trial was conducted between March 2017 and April 2020 in 10 centers in China. Patients with inoperable, locally advanced, clinical stage II to IV ESCC were enrolled and randomized to receive SIB-RT concurrent with and followed by oral S-1 chemotherapy (CRTCT group) or SIB-RT alone (RT group). Data analysis was completed on March 22, 2022. Interventions: In both groups, the planning gross tumor volume was administered with radiation dose of 59.92 Gy and the planning target volume was administered with radiation dose of 50.4 Gy, in 28 fractions each. In the CRTCT group, concurrent S-1 was administered on radiotherapy days, and consolidated S-1 was administered at 4 to 8 weeks after SIB-RT. Main Outcomes and Measures: The primary end point was overall survival (OS) of the intent-to-treat population. Secondary end points were progression-free survival (PFS) and toxicity profile. Results: A total of 330 patients (median [IQR] age, 75.5 [72-79] years; 220 [66.7%] male patients) were included, with 146 patients randomized to the RT group and 184 randomized to the CRTCT group. A total of 107 patients (73.3%) in the RT group and 121 patients (67.9%) in the CRTCT group were clinically diagnosed with stage III to IV disease. At the time of analysis of the 330 patients in the intent-to treat-population (March 22, 2022), OS was improved in the CRTCT group compared with the RT group at 1 year (72.2% vs 62.3%) and 3 years (46.2% vs 33.9%; log-rank P = .02). PFS was similarly improved in the CRTCT group compared with the RT group at 1 year (60.8% vs 49.3%) and 3 years (37.3% vs 27.9%; log-rank P = .04). There was no significant difference in the incidence of treatment-related toxic effects higher than grade 3 between the 2 groups. Grade 5 toxic effects occurred in each group, including 1 patient who experienced myelosuppression and 4 patients with pneumonitis in the RT group and 3 patients with pneumonitis and 2 patients with fever in the CRTCT group. Conclusions and Relevance: These findings suggest that oral S-1 chemotherapy administered with SIB-RT should be considered as an alternative treatment option for patients aged 70 years and older with inoperable ESCC, since it improved survival outcomes without additional treatment-related toxic effects compared with SIB-RT alone. Trial Registration: ClinicalTrials.gov Identifier: NCT02979691.

Pretreatment Low Serum Sodium as a Prognostic Factor for Patients with Esophageal Cancer Treated with Radiotherapy or Chemoradiotherapy.

Low serum sodium levels have been associated with poor prognoses for several cancers. However, the prognostic value of low serum sodium levels in esophageal carcinoma (EC) has not been well elucidated. We examined the prognostic value of low baseline serum sodium levels before radiotherapy or chemoradiotherapy for EC patients. A retrospective analysis of data from EC patients who received radiotherapy or chemoradiotherapy at a single cancer center was performed. Patients were divided into low serum sodium level (≤140.0 mmol/L) or high serum sodium level (>140.0 mmol/L) groups according to the median pretreatment serum sodium level. The Kaplan-Meier model and Cox proportional hazards model were used for survival analyses. The 5-year progression-free survival (PFS) and overall survival (OS) rates in the whole group were 16.9% and 21.8%, respectively. The PFS and OS rates of patients in the low serum sodium levels group were significantly lower than those in the high serum sodium levels group (p < 0.001). A similar association between PFS/OS and sodium levels was observed in the treatment subgroups. The univariate analysis showed that low serum sodium levels, Karnofsky performance status (KPS), clinical N stage, tumor site, clinical stage, and treatment mode were the influencing factors of OS. Multivariate analyses indicated that low baseline serum sodium levels were an independent prognostic marker of poor PFS (HR, 1.744; 95% CI, 1.248-2.437; p = 0.001) and OS (hazard ratio (HR), 2.125; 95% confidence interval (CI), 1.555-2.904; p < 0.001). Pretreatment levels of low serum sodium could be a new and helpful serum biomarker of the prognosis of EC patients receiving radiotherapy or chemoradiotherapy.