Tongji University

About: Tongji University is a education organization based out in Shanghai, China. It is known for research contribution in the topics: Computer science & Population. The organization has 76116 authors who have published 81176 publications receiving 1248911 citations. The organization is also known as: Tongji & Tóngjì Dàxué.

Safety and efficacy of first-line bevacizumab-based therapy in advanced non-squamous non-small-cell lung cancer (SAiL, MO19390): a phase 4 study

Abstract: Summary Background Results of two phase 3 trials have shown first-line bevacizumab in combination with chemotherapy improves clinical outcomes in patients with advanced or recurrent non-squamous non-small-cell lung cancer (NSCLC). The SAiL (MO19390) study was undertaken to assess the safety and efficacy of first-line bevacizumab combined with standard chemotherapy regimens in clinical practice. Methods Between August, 2006, and June, 2008, patients with untreated locally advanced, metastatic, or recurrent non-squamous NSCLC were recruited to this open-label, single group, phase 4 study from centres in 40 countries. Eligible patients had histologically or cytologically documented inoperable, locally advanced, metastatic, or recurrent disease (stage IIIB–IV); an Eastern Cooperative Oncology Group performance status of 0–2; and adequate haematological, hepatic, and renal function. Patients received bevacizumab (7·5 or 15 mg/kg every 3 weeks) plus standard chemotherapy for up to six cycles, followed by single-agent bevacizumab until disease progression. The primary endpoint was safety; analysis was by intention to treat (ITT). This study is registered with ClinicalTrials.gov, number NCT00451906. Findings At the final data cutoff (July 24, 2009), an ITT population of 2212 patients was assessed. The incidence of clinically significant (grade ≥3) adverse events of special interest was generally low; thromboembolism occurred in 172 (8%) patients, hypertension in 125 (6%), bleeding in 80 (4%), proteinuria in 67 (3%), and pulmonary haemorrhage in 15 (1%). 57 (3%) patients died because of these adverse events, with thromboembolism (26 patients, 1%) and bleeding (17, 1%) as the most common causes. The most common grade 3 or higher serious adverse events deemed by investigators to be associated with bevacizumab were pulmonary embolism (28 patients; 1%) and epistaxis, neutropenia, febrile neutropenia, and deep vein thrombosis (all of which occurred in 13 patients [1%]). Bevacizumab was temporarily interrupted after 28 (2%) of 1347 bleeding events and 72 (7%) of 1025 hypertension events, and permanently discontinued after 110 (8%) bleeding events and 40 (4%) hypertension events. No new safety signals were reported. Interpretation Our results confirm the manageable safety profile of first-line bevacizumab in combination with various standard chemotherapy regimens for treatment of advanced non-squamous NSCLC. Funding F Hoffmann-La Roche Ltd.

Human-like autonomous car-following model with deep reinforcement learning

Abstract: This study proposes a framework for human-like autonomous car-following planning based on deep reinforcement learning (deep RL). Historical driving data are fed into a simulation environment where an RL agent learns from trial and error interactions based on a reward function that signals how much the agent deviates from the empirical data. Through these interactions, an optimal policy, or car-following model that maps in a human-like way from speed, relative speed between a lead and following vehicle, and inter-vehicle spacing to acceleration of a following vehicle is finally obtained. The model can be continuously updated when more data are fed in. Two thousand car-following periods extracted from the 2015 Shanghai Naturalistic Driving Study were used to train the model and compare its performance with that of traditional and recent data-driven car-following models. As shown by this study’s results, a deep deterministic policy gradient car-following model that uses disparity between simulated and observed speed as the reward function and considers a reaction delay of 1 s, denoted as DDPGvRT, can reproduce human-like car-following behavior with higher accuracy than traditional and recent data-driven car-following models. Specifically, the DDPGvRT model has a spacing validation error of 18% and speed validation error of 5%, which are less than those of other models, including the intelligent driver model, models based on locally weighted regression, and conventional neural network-based models. Moreover, the DDPGvRT demonstrates good capability of generalization to various driving situations and can adapt to different drivers by continuously learning. This study demonstrates that reinforcement learning methodology can offer insight into driver behavior and can contribute to the development of human-like autonomous driving algorithms and traffic-flow models.