Gongzi Zhang

Gongzi Zhang

Auto-generated profile, based on publicly available information

Gongzi Zhang

China Department of Orthopedics, Chinese PLA General Hospital, Beijing 100141, China

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 April 2025 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
JWJunyu Wu
Junyu Wu
HWHe Wang
He Wang
GZGongzi Zhang
Gongzi Zhang
YLYubin Liu
Yubin Liu
Jie ZhaoJie Zhao
Jie Zhao
HCHegao Cai
Hegao Cai
The increasing prevalence of balance disorders presents significant challenges to rehabilitation therapy, prompting the development of rehabilitation robots as effective tools for motion training. To enhance the patient experience and rehabilitation outcomes in human-robot collaborative training, a motion control strategy based on reinforcement learning and genetic algorithms (RL-GA) with somatosensory enhancement is proposed and implemented in a novel rehabilitation robot. This study introduces a mathematical model of the human sensory system to quantify somatosensory feedback related to motion and integrates the washout algorithm (WA) into the robot's control system, facilitating the reproduction of somatosensation and motion simulation. Three typical rehabilitation modes—level walking, stair climbing, and stair descending—are selected, with natural gait features extracted as predefined trajectories. The WA parameters for each mode are optimized using RL-GA in various rehabilitation scenarios. Simulation results demonstrate that the washout filtering optimization method using RL-GA reduces theoretical somatosensory error by approximately 10% to 20% across all three rehabilitation modes, compared to traditional GA optimization. The experimental results further confirm the reliability and feasibility of the proposed method. The proposed approach enhances the realism of robot-assisted motion, thereby theoretically improving training effectiveness and accelerating the rehabilitation process.

31 July 2026 in Science Advances
Verified
Journal Article
Closed access
YLYumeng Lu
Yumeng Lu
Xinjian FanXinjian Fan
Xinjian Fan
CFChengjuan Fan
Chengjuan Fan
GZGongzi Zhang
Gongzi Zhang
HCHui Cui
Hui Cui
Zhiguang WuZhiguang Wu
Zhiguang Wu
Intrathecal delivery for central nervous system (CNS) therapy lacks spatiotemporal control due to spinal canal anatomical constraints, especially at human scale. We developed an adaptive magnetic nanorobotic platform enabling multimode swarm mobility for in vivo spinal canal navigation in mice and nonhuman primates. Programmable chain-like, vortex-like, and ribbon-like swarm were accomplished, and vortex swarm was indicated as optimal strategy for adaptive intrathecal navigation. Driven by a robotic arm–generated magnetic field, the vortex swarms indicate navigation within spinal canal over centimeter distances, which magnetically enhanced accumulation at CNS target sites and prolonged their in vivo retention, thereby substantially enhancing their therapeutic efficacy compared with conventional passive intravascular delivery, with minimal side effects. Tests in nonhuman primates indicate that the nanorobotic swarm migration is from the gap between the bundles of cauda equina and then the spin cord, revealing their morphological adaption at the cauda equina interface to circumvent anatomical constraints.

16 October 2026 in Science Advances
Verified
Journal Article
Open access
Zhiguang WuZhiguang Wu
Zhiguang Wu
YWYalin Wang
Yalin Wang
HZHongyue Zhang
Hongyue Zhang
Xinjian FanXinjian Fan
Xinjian Fan
MLMingtong Li
Mingtong Li
Jie ZhaoJie Zhao
Jie Zhao
An urgent requirement in chimeric antigen receptor (CAR) immune cell therapy [e.g., CAR–natural killer (NK) cells] for solid tumors is to engineer the cells for precise tumor targeting. Meanwhile, medical microrobots can be navigated toward specific target sites using external magnetic fields. Here, by bridging these approaches, we turn CAR-NK cells into magnetic biohybrid microrobots (CAR-NK microrobots) that achieve controllable long-distance targeted navigation at therapeutic doses (∼10 6 cells) in both murine and nonhuman primate animal models in vivo. CAR-NK microrobots are developed through genetic engineering with CAR structures and loading with magnetic nanoparticles for field-based navigation. Different magnetic field inputs reconfigure their swarms between vortex and ribbon modes, allowing dense aggregation for targeted navigation and further broad dispersion for large-area coverage. Using programmable magnetic fields, in vivo CAR-NK microrobot–based therapy is demonstrated for metastatic leptomeningeal carcinomatosis in brain’s cerebrospinal fluid space. After intrathecal injection, the microrobots are magnetically guided long distance through the spinal canal toward the brain in vortex mode and subsequently spread over the target site in ribbon mode, as visualized by real-time angiographic imaging in a nonhuman primate model, providing a proof-of-feasibility demonstration in a nonhuman primate anatomical setting.

1 January 2021 in SSRN Electronic Journal
Verified
Preprint
Open access
PDF file
GZGongzi Zhang
Gongzi Zhang
MLMiao Liu
Miao Liu
SZShuwei Zhang
Shuwei Zhang
XWXiang Wang
Xiang Wang
NHNing Hu
Ning Hu
LZLicheng Zhang
Licheng Zhang
Background: Intertrochanteric fracture is a significant risk factor for elder death and disability. Though current guidelines remain controversial for the treatment strategies by limited evidence, intramedullary fixation has overtaken the use of extramedullary fixation as the preferred technique. We aimed to assess the effect of different treatments on death and secondary fracture, and evaluate their cost-effectiveness. Methods: We conducted a population-based cohort study of all intertrochanteric fracture patients (≧60 years) in Beijing from 2008 to 2016, using population-based health database from the Beijing Municipal Health Commission and Beijing Center for Disease Control and Prevention (CDC) . We used multivariable Cox-regression analysis and semi-competing risk model to calculate the clinical effects and determine whether surgical technique was an independent and significant risk factor for death and secondary fracture. We further used inverse probability of treatment propensity-score weighting and did subgroup analyses by age and comorbidities. A Markov model was constructed to evaluate the incremental cost-effectiveness ratios and determine the most cost-effective treatment. Findings: During a median 5.49 years follow-up, a total of 36488 intertrochanteric fracture patients were included. From 2008 to 2016, the proportion of intramedullary fixation increased from 27.7% to 54.4%, while the extramedullary fixation decreased from 42.5% to 20.7%. After adjusted, extramedullary fixation had significantly lower mortality than intramedullary fixation by 8.8% (hazard ratio [HR] 0.839, 95% CI 0.789-0.891), especially for elder patients with fewer comorbidities. After adjusted, extramedullary fixation had significantly lower secondary fracture incidence than intramedullary fixation by 7.4% (hazard ratio [HR] 0.791, 95% CI 0.715-0.875). Extramedullary fixation was the most cost-effective strategy with an incremental cost-effectiveness ratio of $31,354.20 per QALY compared with conservative treatment. Interpretation: Compared to other operative treatments, the extramedullary fixation was associated with lower mortality and secondary fracture incidence, and was the most cost-effective strategy. Funding: National Natural Science Foundation of China (61771489), Program of Beijing Municipal Science and Technology Commission (Z191100004419008). Declaration of Interest: None to declare. Ethical Approval: This study was approved by the biomedical ethics board of the Chinese PLA General Hospital (Ethical Approval No.: S2019-014-01)

9 July 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
JWJunyu Wu
Junyu Wu
TQTaoming Qian
Taoming Qian
GZGongzi Zhang
Gongzi Zhang
HPHongye Peng
Hongye Peng
ZMZhuoqi Man
Zhuoqi Man
YWYanhui Wei
Yanhui Wei
End-effector-based rehabilitation robots can provide controlled lower-limb training for patients with balance disorders, but gait reconstruction must balance motion-perception fidelity with mechanical displacement constraints. To address this trade-off, this study proposes a reinforcement learning-assisted NSGA-II method, termed RL-NSGAII, for washout-algorithm-based gait reconstruction. A bi-objective optimization model is formulated to minimize theoretical vestibular motion-perception error and end-effector displacement simultaneously, while constraining additional tilt motion for safety and comfort. Q-learning is embedded into NSGA-II to adaptively select crossover and mutation parameters, and a hypervolume-increment reward is introduced to improve Pareto-front convergence and distribution quality. Simulation experiments are conducted on 12 representative gait trajectories covering level walking, stair ascent, and stair descent. The proposed method is compared with NSGA-II, MOEA/D, RVEA, and MOPSO over 80 independent runs. Results show that RL-NSGAII achieves the best mean hypervolume and inverted generational distance values in six scenarios, obtains the best overall weighted rank, and provides a favorable compromise between perceptual-error reduction and motion-amplitude limitation. The resulting Pareto solutions offer alternative gait-reconstruction schemes for different rehabilitation preferences, supporting individualized robotic gait training for patients with varying levels of balance impairment.