Research Associate (2002-)


Graduated from University of Tokyo in Japan, 1986

MS degree from University of Tokyo in Japan, 1988

Tokyo Research Lab, IBM Research, 1988-1996

Yamato Software development, 1996-2000

Visiting Research, CMU, 2000-2002





Department of Mechanical Engineering

Carnegie Mellon University

5000 Forbes Avenue

Pittsburgh, PA 15213, U.S.A. 

Office: HH B 127

Lab: (412) 268-8880






Automated surface mesh generation

We developed a high quality (non-linear analysis, like crash simulation, capable) mesh generator. This bubble mesh based mesh generation system can handle the realistic scale surface CAD model, it contains four key technologies:

  • Automatic heeling of CAD model error

Define an approximated B-Spline surface covering entire CAD model and re-parameterize original CAD model on it. We achieved robust mesh generation free from CAD model error (tiny gap, overlapping, etc.)

  • Directionality control along CAD primary curvature direction via Square bubble packing

Less than 0.05% invalid quality elements in 340 inner model of an automobile model.


Total elements ~400K, total number of bad elements 130 (quality criteria: minimum angle 30deg, maximum angle 135deg, aspect ratio 4.0, time step 1.0e-6, folding angle of quad element 15deg)

Triangle elements ratio is ~13%.

  • Sizing controllability & anisotropic mesh capability

In bubble packing, we can control the size and aspect ratio of each bubble.

  • Thread-safe implementation

Our mesh generation system is thread-safe and multi-CPU aware, which runs on Windows, Linux, and AIX.






Medical CAD & Robotics

We developed a computer system to help surgical operations, this system supports whole geometrical aspects of surgical procedure.

  • X-ray based 3D surface bone model reconstruction

We reconstruct 3D model from 2 or several X-ray images. We employ the template model and use the GPU to generate synthetic X-ray image. We hieratically subdivide the control lattice of the template model and deform it via FFT to minimize the error of real X-ray image and synthetic one. We can generate 3D model in minimum labor cost in marked contrast to the current CT scan based 3D surface model reconstruction, which is heavily labor intensive.

  • Surgery simulation & planning

Since bone deformity or its orientation is in the 3D space, it is important to navigate it in 3D manner. This navigation helps the surgeon to understand each patient case. Currently this is only done in the imagination of the surgeon, ie. surgical experience is necessary and important.

  • Real time 3D model registration

Finally, in the surgery time, this program register the current x-ray or camera image to the computer model. By using this system, surgeon can handle the case in accurate manner. This simulation is based on the same technology of the bone reconstruction.