Ken'ich Morooka
リンク


Click figure to enlarge

Digial Brain Atlas for Safe and Accurate Stereotactic Neurosugery
A brain atlas is a very important tool for stereotactic neurosurgery to implant an electrode into the target neural structure. Practically, a classical brain atlas has been used by altering three dimensional magnification to fit patients’ brain images. However, it is difficult to apply the classical atlases to the individual Japanese patient because coronal, horizontal and sagittal slices of the atlases are made from different Westerners. This paper reports the detail of technical innovations and preliminary results in constructing a digitalized human brain atlas by the novel techniques in neuroanatomy and computer science.

A cadaver brain, 89-year-old male, was fixed in 10% formalin. The whole brain surface was scanned by a 3D digitizer, which provides the shape geometry data. Using a specially designed brain-cutting machine, the brain embedded in agar was cut into serial 10[mm]-thick blocks. After removing the agar, the blocks were scanned by the digitizer. The blocks were embedded in gelatin. After making four markers by penetrating the surrounding gelatin with needle containing India ink, each block was further sliced into serial 100[μm]-thick sections with a vibrating microslicer. The sections were stained with Nissl- and myelin-stainings. We generate both the models of the brain surface and the blocks by using the geometrical data, and another block models by modifying the section images based on the location of the markers. The relationship between the block models enables us to incorporate the slice images into the brain model.

This project is partially supported by Grant-in-Aid for Scientific Research on Innovative Areas, "Computational Anatomy for Computer-aided Diagnosis and Therapy" MEXT, Japan
Morphign between rabbit and torus
[Demo Movie]
Morphign between torus and block with two holes
[Demo Movie]


3-D Morphing between Objects with Different Topologies Using Deformable Models
3-D morphing is shape transformation from one object to another. This paper presents a new method of morphing objects with different topologies. The basic idea behind our method is to generate an approximate model of a given object by using a deformable model, called the Active Balloon Model(ABM). Since the data structure for each object is similar to that of the original ABM, it is easy to find correspondence between two models by utilizing ABMs. We also propose a new method of generating intermediate models during morphing. This method uses various kinds of time functions according to the types of topological changes. These time functions enable us to control topological changes arbitrarily.

This research in 2006 received Niwa-Takayanagi Award, The Institute of Image Information and Television Egineer.



Click figure to enlarge
Liver deformation [Demo Movie]


Real-Time Nonlinear FEM-based Simulation of Soft Tissue Deformation by Neural Network
Methods have been proposed for simulating the behavior of soft tissues. Among these methods, the finite element method (FEM) is a well-known technique for accurately modeling the behaviors of continuous objects. Compared with other simulation techniques, FEM can achieve a more physically realistic simulation for deformable objects with linear and nonlinear material properties. However, the FEM-based simulation of tissue deformation is very time-consuming because of the need to solve large scale system of equations. Particularly, in the case of soft tissues, the simulation is very complex because the soft tissues have the properties of anisotropy, inhomogeneity, and large-range deformation. Although many approaches have bee developed for solving this problem, real-time FEM analysis is still a challenge.

We presents a new method for simulating the deformation of soft tissue models in the framework of a neural network. The proposed method is based on the superposition of principal deformed models, which have been computed off-line by a nonlinear finite element method (FEM). The neural network is trained to find a relationship between forces and the deformed models when the forces act on the tissue model. The trained network, called neuroFEM, can emulate the nonlinear FEM analysis of tissue behavior. Experimental results show that our method can reduce the computational time while keeping an acceptable accuracy compared with the original computation by nonlinear FEM.

This project is partially supported by Grant-in-Aid for Scientific Research on Innovative Areas, "Computational Anatomy for Computer-aided Diagnosis and Therapy" MEXT, Japan


Click figure to enlarge


Trachea and Esophagus Classification by AdaBoost for Automatic Tracheal Intubation System
Tracheal intubation is a basic procedure for maintaining the airway of a patient who is unconscious or unable to breathe independently. In standard intubation procedure, a laryngoscope blade is used at first to observe the exact position of the vocal cord before inserting an air tube. However, an operator sometimes encounters difficulties in the observation: for example, the operator cannot see the vocal cord directly due to an abnormal downward or back placement of the tongue. This may cause the false intubation such as esophageal intubation or airway complications like dislodgement of vertebrae cervicales and teeth.

For accurate and safe tracheal intubation, we have been developing an automatic laryngoscope system using a camera attached at the tip of an endotracheal stylet. This paper proposes a method for classifying trachea and esophagus images acquired by the camera. The proposed method is based on Adaboost algorithm using three types of image features. These features are derived from the circular patterns of cricoid cartilage observed in tracheal images. From experimental results, the average recognition rate of 1,900 test images is 97.6%, and it is verified that the proposed technique is applicable to the classification of trachea and esophagus images for safe intubation.
Copyright(C) 2011 IRVS Lab., Kyushu Univ., All Rights Reserved.