KKKenji Kawashima

Kenji Kawashima

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Kenji Kawashima has a total of 65 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    orGAN: Bridging Synthetic and Real Data for Surgical AI via Consistent Image–Label Generation

    by Niran Nataraj, Maina Sogabe, Kenji Kawashima

    \textbf{Background and Objective:} Deep learning in medical imaging faces challenges including restricted data diversity, ethical constraints, high data acquisition costs, and limited accurate annotations. Accurately identifying and localizing bleeding sources during surgery is especially difficult due to scarce, well-annotated, realistic datasets. We propose orGAN, a framework for generating realistic, annotated synthetic surgical images of bleeding scenarios to mitigate these limitations. \textbf{Methods:} The orGAN pipeline uses an enhanced StyleGAN3 architecture tailored to silicone-based “mimicking organs” that emulate tissue properties. The system synthesizes bleeding images with specified bleeding coordinates and simultaneously extracts coordinates and masks to provide artifact-free supervision signals. An inpainting network refines images, and Bleeding Alert Map (BAM) models are trained on a balanced 1:1 dataset blending orGAN synthetic data with “mimicking organ” images. The software is implemented as a modular, reproducible pipeline with clear interfaces for downstream training. \textbf{Results:} The enhanced StyleGAN3 achieved an Inception Score increase of 2.24, and the inpainting network maintained a mean SSIM of 0.98. BAM trained on the blended dataset reached 90\% image-level bleeding detection accuracy and up to 99\% frame-level accuracy in both ex-vivo and in-vivo evaluations, outperforming models trained solely on real or solely on synthetic data. \textbf{Conclusions:} orGAN provides an ethical and economical approach to generating annotated surgical datasets, coupling high-fidelity image synthesis with automatic supervision (coordinates and masks). The modular design enables reuse beyond one-off experiments and supports bleeding-localization workflows and broader AI applications in surgery.

  • Controlling Pneumatic Bending Actuator With Gain-Scheduled Feedforward and Physical Reservoir Computing State Estimation

    by Junyi Shen, Tetsuro Miyazaki, Kenji Kawashima

    Hysteresis brings challenges to both the control and state perception of soft robots. This work proposes a real-time gain-scheduled feedforward proportional controller design and a Physical Reservoir Computing (PRC) model to address hysteresis effects in the motion control and unobstructed state estimation of a dual pneumatic artificial muscle (PAM) soft bending actuator. The dual-PAM soft actuator comprises an active PAM used for actuation and a pressurized-and-sealed passive PAM serving as a physical reservoir and used for computation. The physical reservoir's state is reflected by the passive PAM's inner pressure and used for bending state estimation. Experiments exhibit the physical reservoir state's nonlinear responses to the active PAM's actuation inputs. The proposed feedforward controller improves the soft actuator's responsiveness in hysteresis dead zones by dynamically adjusting the feedforward proportional gain. The proposed controller outperforms a linear approximation-based feedforward controller in motion control, and the PRC-based bending state estimation model achieves higher accuracy than a comparative Echo State Network (ESN) with 1,000 neurons. The presented strategies are expected to benefit the precise motion control and unobstructed state estimation of soft actuators.


  • Open Access

    Control Pneumatic Soft Bending Actuator With Feedforward Hysteresis Compensation by Pneumatic Physical Reservoir Computing

    by Junyi Shen, Tetsuro Miyazaki, Kenji Kawashima

    The nonlinearities of soft robots bring control challenges like hysteresis but also provide them with computational capacities. This letter introduces a fuzzy pneumatic physical reservoir computing (FPRC) model for feedforward hysteresis compensation in motion tracking control of soft actuators. Our method utilizes a pneumatic bending actuator as a physical reservoir with nonlinear computing capacities to control another pneumatic bending actuator. The FPRC model employs a Takagi-Sugeno (T-S) fuzzy logic to process outputs from the physical reservoir. The proposed FPRC model shows equivalent training performance to an Echo State Network (ESN) model, whereas it exhibits better test accuracies with significantly reduced execution time. Experiments validate the FPRC model's effectiveness in controlling the bending motion of a pneumatic soft actuator with open-loop and closed-loop control system setups. The proposed FPRC model's robustness against environmental disturbances has also been experimentally verified. To the authors' knowledge, this is the first implementation of a physical system in the feedforward hysteresis compensation model for controlling soft actuators. This study is expected to advance physical reservoir computing in nonlinear control applications and extend the feedforward hysteresis compensation methods for controlling soft actuators.


  • Open Access

    Gait Perception via Actual and Estimated Pneumatic Physical Reservoir Output

    by Junyi Shen, Tetsuro Miyazaki, Swaninda Ghosh, Toshihiro Kawase, Kenji Kawashima

    Accurately identifying user needs in terms of assist timing and magnitude presents challenges for wearable power‐assist limb devices. Traditional approaches to gait perception—such as estimating joint angles and walking conditions—often rely on electronic sensors and neural networks, which can compromise wearability and impose high computational demands. Physical reservoir computing (PRC), which utilizes the inherent nonlinearity of physical systems for data processing, offers a promising alternative. This study proposes a novel self‐estimated physical reservoir computing (SEPRC) model that improves traditional PRC models for gait perception using a wearable pneumatic physical reservoir. A core feature of the new model is the self‐estimation structure, wherein the outputs of the physical reservoir are mutually estimated. Experimental evaluations indicate that the SEPRC model outperforms traditional PRC in clustering time‐series reservoir output sequences with the same dimensionality. This enhanced clustering performance is subsequently leveraged in gait perception by incorporating Takagi–Sugeno fuzzy logic for joint angle estimation and a softmax activation function for walking condition recognition. The newly proposed time‐sequence processing approach facilitates the traditional PRC model to achieve higher accuracy in gait perception and greater robustness against the user's walking pattern variations while preserving PRC's hardware simplicity.


  • Study on Force Control of Pneumatically Driven Arm Motion Assist Suit

    by Yoshihide Tomita, Tetsuro Miyazaki, Maina Sogabe, Kenji Kawashima


  • Open Access

    Detection and Cancellation of Multiplicative FDI Attack on Bilateral Encrypted Control System

    by Katsumasa Kosha, Tetsuro Miyazaki, Kaoru Teranishi, Kiminao Kogiso, Kenji Kawashima

    Teleoperation of remote assist robots has recently advanced owing to the development of communication technology. However, anonymous malicious attacks may intercept or falsify the network control system. Therefore it is necessary to improve the security against cyber-attacks. As a countermeasure, an encrypted control method has been applied to prevent intercepting and detect the falsification of control parameters, as well as control signals by performing control operations with encrypted signals and parameters. However, it is difficult to detect False Data Injection (FDI) attacks, which exploit encryption malleability, and countermeasures against such attacks are necessary. Therefore, this study proposes a novel method to detect and cancel the effect of the FDI attack, which aims at the encrypted bilateral control system’s malleability. The assumed FDI attack falsifies the plaintext by multiplying a constant factor, which is realized by monitoring the entire system’s energy change and estimating the attack parameters. The proposed method can detect and cancel the impact of the FDI attack within finite steps. Moreover, we verified the proposed method for the bilateral control system using ElGamal encryption and experimentally confirmed its effectiveness against the FDI attack.


  • Open Access

    Ensuring communication redundancy and establishing a telementoring system for robotic telesurgery using multiple communication lines

    by Yusuke Wakasa, Kenichi Hakamada, Hajime Morohashi, Takahiro Kanno, Kotaro Tadano, Kenji Kawashima, Yuma Ebihara, Eiji Oki, Satoshi Hirano, Masaki Mori

    Assuring communication redundancy during the interruption and establishing appropriate teaching environments for local surgeons are essential to making robotic telesurgery mainstream. This study analyzes robotic telesurgery with telementoring using standard domestic telecommunication carriers. Can multiple carriers guarantee redundancy with interruptions? Three commercial optical fiber lines connected Hirosaki University and Mutsu General Hospitals, 150 km apart. Using Riverfield, Inc. equipment, Hirosaki had a cockpit, while both Mutsu used both a cockpit and a surgeon’s console. Experts provided telementoring evaluating 14 trainees, using objective indices for operation time and errors. Subjective questionnaires addressed image quality and surgical operability. Eighteen participants performed telesurgery using combined lines from two/three telecommunication carriers. Manipulation: over 30 min, lines were cut and restored every three minutes per task. Subjects were to press a switch when noticing image quality or operability changes. Mean time to task completion was 1510 (1186–1960) seconds: local surgeons alone and 1600 (1152–2296) seconds for those under remote instructor supervision, including expert intervention time. There was no significant difference (p = 0.86). The mean error count was 0.92 (0–3) for local surgeons and 0.42 (0–2) with remote instructors. Image quality and operability questionnaires found no significant differences. Results communication companies A, B, and C: the A/B combination incurred 0.17 (0–1) presses of the environment change switch, B/C had 0, and C/A received 0.67 (0–3), showing no significant difference among provider combinations. Combining multiple communication lines guarantees communication redundancy and enables robotic telementoring with enhanced communication security.