2016 3rd international conference on biomedical and

15
Proceedings of 2016 3 rd International Conference on Biomedical and Bioinformatics Engineering ICBBE 2016 Taipei, Taiwan November 12-14, 2016 ISBN: 978-1-4503-4824-9

Upload: others

Post on 20-Feb-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 2016 3rd International Conference on Biomedical and

Proceedings of

2016 3rd

International Conference on Biomedical and

Bioinformatics Engineering

ICBBE 2016

Taipei, Taiwan

November 12-14, 2016

ISBN: 978-1-4503-4824-9

Page 2: 2016 3rd International Conference on Biomedical and

The Association for Computing Machinery

2 Penn Plaza, Suite 701

New York New York 10121-0701

ACM ISBN: 978-1-4503-4824-9

ACM COPYRIGHT NOTICE. Copyright © 2016 by the Association for Computing Machinery, Inc. Permission

to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee

provided that copies are not made or distributed for profit or commercial advantage and that copies bear this

notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM

must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to

redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Publications Dept.,

ACM, Inc., fax +1 (212) 869-0481, or [email protected]

For other copying of articles that carry a code at the bottom of the first or last page, copying is permitted provided that

the per-copy fee indicated in the code is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers,

MA 01923, +1-978-750-8400, +1-978-750-4470 (fax).

Page 3: 2016 3rd International Conference on Biomedical and

III

Table of Contents

Proceedings of 2016 3rd International Conference on Biomedical and Bioinformatics Engineering

Preface……………………..………………………………………………………….…. ……..……………… V

Conference Committees…………………………………………………………………………………...…..VI

Session 1-Bioinformatics and Computational Biology

A Novel Framework for Repeated Measurements in Diffusion Tensor Imaging

Mohammad Alipoor, Irene Y.H. Gu, Andrew Mehnert, Göran Starck, and Stephan E. Maier

1

Tactile Sensor for Cardiovascular Catheters

Chao-Hsien Huang, Cheng-Hung Shih, and Nai-Jun An

7

Algorithm for Segmentation and Reduction of Fractured Bones in Computer-Aided Preoperative Surgery

Irwansyah, Jiing-Yih Lai, Terence Essomba, and Pei-Yuan Lee

12

Study on Ratcheting Behavior of Trabecular Bone with and without Marrow Under Cyclic Compression

Chao-lei Wei, Li-lan Gao, and Chun-qiu Zhang

19

Analysis of Methicillin-resistant Staphylococcus Aureus Using Apriori, DBSCAN, and K-means Algorithms

Min Young Lee, Taeseon Yoon

23

Influence of Contrast Enhancement Methods in Brain Tumor Detection

Ahsan Khawaja, Maher un Nisa

29

Mechanical Behavior of Articular Cartilage Soaked in Physiological Saline under Cyclic Compressive Loading

Dong-dong Liu, Li-lan Gao, and Xiao-yi Qin

35

Hand Motion Capture for Medical Usage

Kiyoshi Hoshino, Sota Sugimura, Motomasa Tomida, Naoki Igo, Isao Kawano, and Masahiko

Sumitani

40

Quantitative Analysis of Spectroscopy Data for Skin Oximetry

Audrey Huong, Xavier Ngu

46

Measurement of Rotational Eye Movement with Blue Light Irradiation

Kiyoshi Hoshino, Nayuta Ono, Motomasa Tomida, and Naoki Igo

50

Numerical Simulation of the Adenoidectomy Preoperative and Postoperative Upper Airway in Children with OSAHS

Chi Yu, Gang Wang, and Jing Zhang

55

Page 4: 2016 3rd International Conference on Biomedical and

IV

Development of a Swordsmanship Machine Enabling the Inner and Outer Muscles to be Safely Trained while Having Fun

Kiyoshi Hoshino, Chuanhan Cheng

59

Research on the Factors of Colonoscopy Screening Compliance in High-risk Colorectal Cancer Groups

MA Hong-mei, LU Jian-hong

63

FP-AK-QIEA-R for Multi-Objective Optimization Josimar Edinson Chire Saire

67

Session 2- Pharma Medicine and Biological Sciences

Potentiometric Determination of a Regulated Veterinary Drug via MIP-Modified Electrode

Yasmin D.G. Edañol, Marleane Rovi R. Ferrer, Rizzie Kimberly M. Raguindin, and Susan D. Arco

71

The Primary Study of the Relationship between Environmental Factors and Dawn Song in Varied Tits

Tingting Zhao, Jingfeng Lin, Xiande Zhang, Dongmei Wan, and Jiangxia Yin 75

The Impact of Externally Supplied Protein on Root and Phytohormone in Endangered Species Cercidiphyllum Japonicum Cutting Seedling

Shaohui Huang

81

Session 3- Environmental and Chemical Engineering

Preparation and Characterization of Kenaf Derived Heterogeneous Catalyst for Esterification Reaction

Koguleshun Subramaniam, Fei-Ling Pua, Kumaran Palanisamy, and Saifuddin M.Nomanbhay

86

Combination of Photocatalytic and Biological Process for Treatment of Wastewater

Nur Syazrin Amalina Abdullah, Sufian So’aib, Fazlena Hamzah

90

Preliminary Study of Electrolytic Cell a Treatment for Wasted Brine from Resin Regeneration

Patcharin Racho, Sasiwimon Namgool, Warutai Dejtanon

94

Session 4- Coastal and Urban Engineering

Submerged Breakwater Hydrodynamic Modeling for Wave Dissipation and Coral Restorer Structure

Safari Mat Desa, Othman A. Karim, and Azuhan Mohamed

98

Research on Sea Reclamation and Urban Sustainable Development

Shen Yilei

102

Author Index 106

Page 5: 2016 3rd International Conference on Biomedical and

V

Preface

This volume contains papers presented at the 2016 3rd

International Conference on Biomedical and

Bioinformatics Engineering, which was held during November 12-14, 2016 in Taipei, Taiwan.

ICBBE provides a scientific platform for both local and international scientists, engineers and

technologists who work in all aspects of biomedical and bioinformatics engineering. In addition to the

contributed papers, internationally known experts from several countries are also invited to deliver

keynote and plenary speeches at ICBBE 2016.

The volume includes 22 selected papers which were submitted to the conference from universities,

research institutes and industries. Each contributed paper has been peer-reviewed by reviewers who

were collected organizing and technical committee members as well as other experts in the field from

different countries. The proceedings tend to present to the readers the newest researches results and

findings in the field of biomedical and bioinformatics engineering.

Much of the credit of the success of the conference is due to topic coordinators who have devoted their

expertise and experience in promoting and in general co-ordination of the activities for the organization

and operation of the conference. The coordinators of various session topics have devoted a considerable

time and energy in soliciting papers from relevant researchers for presentation at the conference.

The chairpersons of the different sessions played important role in conducting the proceedings of the

session in a timely and efficient manner and the on behalf of the conference committee, we express

sincere appreciation for their involvement. The reviewers of the manuscripts, those by tradition would

remain anonymous, have also been very helpful in efficiently reviewing the manuscripts, providing

valuable comments well within the time allotted to them. We express our sincere and grateful thanks to

all reviewers.

ICBBE 2016 Organizing Committee

November 12-14, 2016

Page 6: 2016 3rd International Conference on Biomedical and

VI

Conference Committees

Conference Chairs Prof. David Zhang, Hong Kong Polytechnic University, Hong Kong

Assoc. Prof. Kuo-Yuan Hwa, National Taipei University of Technology, Taiwan

Conference Program Chairs Prof. Xuefeng Tong, Tongji University, China

Prof. Irene Yu-Hua Gu, Dept. of Signals and Systems, Chalmers Univ. of Technology, Sweden

Prof. Congo Tak Shing CHING, Graduate Institute of Biomedical Engineering, National Chung

Hsing University, Taiwan

Prof. Manoj R. Tarambale, Marathwada Mitra Mandal’s College of Engineering, Pune, India

Technical Committees Prof. Ming-Wei Lin, Institute of Biomedical Informatics, National Yang Ming University,

Taiwan

Prof. Chuang-Chien Chiu, Feng Chia University, Taiwan, R.O.C.

Prof. Mohamed Osama Abdelaal Rabie Elshazly, Cairo University, Egypt

Prof. Edwin Wang, National Research Council Canada, McGill University, Canada

Prof. Sujata Dash, Bijupatnaik University of Technology, Orissa, India

Prof. Helmut Zarbl, Rutgers, The State University of New Jersey, USA

Prof. Mohammed Bougataya, UQO Quebec Canada

Prof. Rita Singh Majumdar, Dept of Biotechnology, Sharda University, Greater Noida, India

Prof. Jun F. (James) Liang, Stevens Institute of Technology, New Jersey, USA

Prof. Muhammad Nawaz Iqbal, Pakistan Engineering Council, Pakistan

Prof. Ajitkumar Gorakhanath Patil, S.B.M.Polytechnic, Mumbai, India

Prof. Bimal Kumar Sarkar, Department of Physics, Galgotias University, India

Prof. Zhen Xie, Tsinghua National Lab for Information Science and Technology, Tsinghua

University, China

Prof. Alexander Polyakov, Sevastopol National Technical University, Russia

Prof. Pedro Joaquin Gutierrez-Yurrita, Instituto Politecnico Nacional, Mexico

Prof. Nagendra Kumar Kaushik, Plasma Bioscience Research Center, Kwangwoon University,

Seoul, South Korea

Assoc. Prof. Lee, Yuan-Chii Gladys, Taipei Medical University, Taiwan

Assoc. Prof. P.SHANMUGHAVEL, Department of Bioinformatics, Bharathiar University

Coimbatore, India

Dr. Wen-Ling Chan, Department of Bioinformatics and Medical Engineering, Asia University,

Taiwan

Dr. Muhammad Arshad Malik, Department of Bioinformatics & Biotechnology, International

Islamic University, Pakistan

Page 7: 2016 3rd International Conference on Biomedical and

Author Index

A

Ahsan Khawaja 29

Andrew Mehnert 1

Audrey Huong 46

Azuhan Mohamed 98

C

Chao-Hsien Huang 7

Chao-lei Wei 19

Cheng-Hung Shih 7

Chi Yu 55

Chuanhan Cheng 59

Chun-qiu Zhang 19

D

Dong-dong Liu 35

Dongmei Wan 75

F

Fazlena Hamzah 90

Fei-Ling Pua 86

G

Gang Wang 55

Göran Starck 1

I

Irene Y.H. Gu 1

Irwansyah 12

Isao Kawano 40

J

Jiangxia Yin 75

Jiing-Yih Lai 12

Jing Zhang 55

Jingfeng Lin 75

Josimar Edinson Chire Saire 67

K

Kiyoshi Hoshino 40, 50, 59

Koguleshun Subramaniam 86

Kumaran Palanisamy 86

L

Li-lan Gao 19, 35

LU Jian-hong 63

M

MA Hong-mei 63

Maher un Nisa 29

Marleane Rovi R. Ferrer 71

Masahiko Sumitani 40

Min Young Lee 23

Mohammad Alipoor 1

Motomasa Tomida 40, 50

N

Nai-Jun An 7

Naoki Igo 40, 50

Nayuta Ono 50

Nur Syazrin Amalina Abdullah 90

O

Othman A. Karim 98

P

Patcharin Racho 94

Pei-Yuan Lee 12

R

Rizzie Kimberly M. Raguindin 71

S

Safari Mat Desa 98

Saifuddin M.Nomanbhay 86

Sasiwimon Namgool 94

Shaohui Huang 81

Shen Yilei 102

Sota Sugimura 40

Stephan E. Maier 1

Sufian So' aib 90

Susan D. Arco 71

T

Taeseon Yoon 23

106

Page 8: 2016 3rd International Conference on Biomedical and

Terence Essomba 12

Tingting Zhao 75

W

Warutai Dejtanon 94

X

Xavier Ngu 46

Xiande Zhang 75

Xiao-yi Qin 35

Y

Yasmin D.G. Edañol 71

107

Page 9: 2016 3rd International Conference on Biomedical and

Algorithm for Segmentation and Reduction of Fractured Bones in Computer-Aided Preoperative Surgery

Irwansyah, Jiing-Yih Lai, Terence Essomba Mechanical Engineering Department

National Central University Taoyuan City 320, Taiwan

[email protected]

Pei-Yuan Lee Orthopedic Department

Show Chwan Memorial Hospital Changhua 500, Taiwan

ABSTRACT

CT images have extensively been used for the diagnosis of serious

bone injuries, such as comminuted fracture, and for the

preoperative planning of orthopedic surgery. In order to enhance

the use of CT images in preoperative planning, it is necessary to

develop a 3D modeling and simulation tools to acquire more

information regarding the real surgery. Bone segmentation and

reduction are two important simulation tools in computer-aided

preoperative surgery. By improving both algorithms can affect to

the feasibility of 3D preoperative planning for fractured bones and

provide more valuable information prior to decision making. We

provide a technique to model, segment and recover fractured bone

from CT images. A multi-region segmentation algorithm is

proposed to reduce processing time and get efficiency. A manual

and a semi-automatic bone reduction algorithm are proposed to

deal with different kinds of fractured bone cases. Two semi-

automatic bone reduction algorithms, multi-point and mirror

positioning, are proposed to improve the efficiency of manual

reduction. The manual reduction is still necessary to deal with

cases which cannot be caught by the semi-automatic algorithms.

Several realistic examples by using real patients’ CT images are

provided to illustrate the feasibility of the proposed method.

CCS Concepts • Applied computing→Health care information systems

Keywords

Bone reduction; bone segmentation; computer aided preoperative

surgery; 3D bone reconstruction.

1. INTRODUCTION Reconstruction of 3D model from computed - tomography (CT)

scan of patient has been routinely used to assist a clinician

deciding an appropriate operating procedure. In order to construct

3D model, the user should extract 2D image contour from each

slices of CT images that save digital images and communication

in DICOM format and apply an algorithm to convert those

contours to a 3D bone model. Two kinds of process that play

important key to generate 3D bone model are segmentation and

reduction. Segmentation is a process to recognize and separate the

bones. In segmentation process some drawbacks such as how to

segment multiple fragments of bone simultaneously, keep the gray

value of different regions of bone that may not be constant,

identify the transition intensity value near the joint area that

generally appears to be fuzzy, and some areas within the bone that

may have similar intensity to the surrounding soft tissue are still

the issues to be solved [1, 2]. In case of fracture, bones are more

difficult to identify because bone fragments may have arbitrary

shape and can belong to any bone in the nearby area [2]. The

observation of fractured lines and fragments using 2D images

such information can be analyzed with a 3D tool.

Bone segmentation approaches can be classified into three types:

interactive, semi-automatic and fully automatic [3]. The

interactive approach is typically applied on 2D images. The

approach works by detecting every contour slice of the bone of

interest. It is time consuming and may leads to errors. Semi-

automatic approach could provide and maintain accuracy as well

as completeness of the bone model. The fully-automatic is the best

method but most difficult. 3D region growing is a common

method applied for bone segmentation. The model constructed by

this method is good enough for visualization but lacks of accuracy

and completeness.

Our previous studies have been contributing to segmentation

algorithm for constructing 3D images and have come out with

most of bone structure to be segmented effectively by the

proposed bone segmentation algorithm. Reference [4] proposed a

region growing algorithm for segmentation of the volume data.

The two consecutive procedures considered are applied

simulatneously, image processing and triangulation. An

interactive region growing method is applied to guaranty that the

targeted bone structure growing is succesful. Reference [3]

presented an optimized method for multiple-bones segmentation

that includes automatic seed region definition, iterative region

growing as well as re-combination and re-segmentation

procedures. This approach beneficial for solving over flow

problem whereas multiple regions points combined together

which makes different bones too difficult to separate accurately.

Several researchers also have been working on an accurate,

complete, efficient and segmentation algorithm for constructing

3D images. Reference [2] proposed a less user interaction method

for segmentation and labeling bone fragments from CT images.

The region growing based method is employed to easily detect

and solve overgrowing cases. Reference [5] presented a new cross

validation based segmentation algorithm which automatically

extracts multiple-level bone structures using a combination of

anatomical knowledge and computational techniques. The

Permission to make digital or hard copies of all or part of this work for

personal or classroom use is granted without fee provided that copies

are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights

for components of this work owned by others than ACM must be

honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior

specific permission and/or a fee. Request permissions from

[email protected]. ICBBE '16, November 12-14, 2016, Taipei, Taiwan

© 2016 ACM. ISBN 978-1-4503-4824-9/16/11…$15.00

DOI: http://dx.doi.org/10.1145/3022702.3022703

12

Page 10: 2016 3rd International Conference on Biomedical and

algorithm successfully detects the bones and is effective as well as

robust in term of the quantitative measurements. Reference [6]

presented an interactive tool for separating bone fragments in CT

volumes. The proposed method combining direct volume

rendering with interactive 3D texture painting interface that

allows the user to identify and mark bone structures instantly. The

user enables place seed point directly either on the rendered bone

surfaces or individual CT slices. The random walks segmentation

algorithm applied to separate marked bones. Reference [7]

proposed an active contour segmentation process to identify the

exact location of the fracture region on the images. The

identification process of bone regions refers to differences the

gray level pixels on the images.

Reduction is a surgical procedure to restore a dislocated or

fractured part to its former place. To maintain stability of the

fracture, specific implant used as fixation device such as plates,

screws, nails or other implants which may be external or internal.

It is important to verify the accuracy of reduction by clinical tests

and X-ray, especially in the case of joint dislocations. The

reduction procedures still have variety of problems depending on

the type of bone and fracture. Aligning the two bone fragment in

order to recover their original posititon is one of the simple

fracture reduction case. If the fracture generates more than two

fragments, then manual alignment becomes a difficult task. For

that purpose some approaches consider to match fracture zones

and algorithms to calculate this zone are generated.

Reference [8] presented a review about the current approaches

employed to match and register bone fragments in complex

fractures. For matching procedure, they classified the reducing

complex bone fractures methods into three kinds; interactive tools,

match fracture areas, and register templates. After the matching

procedure, most of the proposed studies perform a final alignment

to callibrate the result. The alignment is performed by registering

fracture zones such as surface, lines or points to each other.

Reference [1] presented an integrated surgical simulation process

including 3D modelling, visualization, segmentation, bone

reduction, fixation and data output. In order to provide efficient

segmentation process, a bone semi-automatic segmentation is

employed by combining three procedures, multi-region growth,

region re-segmentation and region re-combination. The proposed

bone segmentation algorithm has shown bone structure segmented

successfully [3]. In bone reduction process, manual bone

reduction takes long operating time. Each of the fragment bone

transform piece by piece and the final result visually evaluated.

Reference [9] presented an extended study to resolve some

drawback in previous system and provide an effective 3D

preoperative planning surgery. Semi-automatic bone reduction to

overcome the inefficiency of manual bone reduction for multiple

fragments is introduced. In semi automatic bone reduction, multi-

point positioning and mirror positioning algorithms are proposed.

For the final alignment a registration algorithm is employed to fit

each broken fragment with the reference bone. Reference [10]

presented an extended study to investigate the feasibility of an

integrated preoperative planning system. The operating times

required for each step of preoperative planning and simulation is

measured and several patients’ images data are illustrated such as

femur, tibia, foot, proximal, scapula, distal hand and elbow.

In this paper, we provide a technique to model and segment bone

tissue from CT images, with different bone segments being

modeled individually. A multi-region segmentation algorithm is

proposed to determine seed points automatically, and propagate

all regions simultaneously. In addition, a bone reduction process

allows recovering the original status of broken bones. We provide

a manual and a semi-automatic bone reduction algorithm to deal

with different kinds of situation. The semi-automatic bone

reduction algorithm can be implemented when the CT images of

the un-broken side is available. Otherwise, the manual bone

reduction algorithm can be implemented. Bone segmentation and

reduction are two important simulation tools in computer-aided

preoperative planning for simulating the bone reduction process in

real surgery. Improving the algorithm of bone segmentation and

reduction may affect to the feasibility of 3D preoperative planning

for fractured bones and provide more valuable information prior

to decision making. In the following we organized the paper by

describing the entire stages of computer aided preoperative

surgery modules, the fractured bone segmentation and reduction

system, segmentation and reduction algorithm, realistic example

clinical cases to illustrate the feasibility of the proposed algorithm.

2. COMPUTER AIDED PREOPERATIVE

SURGERY The integrated preoperative planning system provides an

environment for 3D visualization and manipulation. The system is

named PhysiGuide. It is able to display and manipulate both 3D

medical images and computer aided design (CAD) models

simultaneously [10]. The system consists in several modules

including 2D and 3D display, bone segmentation, bone resection,

bone reduction, and fixation. Figure 1 shows the flow diagram of

the integrated preoperative planning system main modules.

Figure 1. Integrated preoperative planning system modules.

3. METHODS Improving the algorithm of bone segmentation and reduction

bring significant progress to the feasibility of 3D preoperative

planning for fractured bones and provide more valuable

information prior to decision making for surgery practitioners. We

provide a technique to model and segment bone tissue from CT

images, with different bone segments being modeled individually.

A multi-region segmentation algorithm is proposed to determine

seed points automatically, and propagate all regions

simultaneously. In addition, a bone reduction process allows

recovering the original status of broken bones. We provide a

manual and a semi-automatic bone reduction algorithm to deal

with different kinds of situation. The semi-automatic bone

CT/MRI images

Voxel-based

Representation

2D Segmentation

3D Region Growing

3D Fractured bones

View

3D Fractured bones

Modeling

3D

Reco

nstru

ction

Mo

delin

g an

d A

pp

lication

Volumetric

Representation

Contour-based

Generation

Volume RenderingVolumetric

Representation

3D Imaging

Representation

CAD-based Medical

Modeling

13

Page 11: 2016 3rd International Conference on Biomedical and

reduction algorithm can be implemented when the CT images of

the un-broken side is available. Otherwise, the manual bone

reduction algorithm can be implemented. Several realistic

examples by using real patients’ CT images are provided to

illustrate the feasibility of the proposed method.

3.1 Fractured Bone Segmentation The proposed bone segmentation algorithm includes three

procedures: multiple region growing, region re-segmentation and

region re-combination. The overall procedures are implemented

interactively through the user interface and can be performed

iteratively. The system accepts both X-ray and CT images which

are stored in DICOM format. Each slice of CT image has

512512 pixels and a gray value of 12 bits. The image data are

converted into volume data. It is essentially 3D matrices that

record the gray value of each voxel on the images. A volume

rendering technique is employed to show the ISO-surface of the

volume data. The skin, bone and soft tissue can be displayed more

realistically by shifting a threshold. Two kinds of threshold, initial

threshold (Ti) and target threshold (Tt) are determined first. The

procedure of the proposed algorithm for segmenting fractured

bone is depicted in Figure 2.

Figure 2. Flow diagram of bone segmentation.

Initial threshold is assigned for initial seed region growing. As the

initial threshold is increased, more seed regions are generated. It

means more possibilities to separate unclear boundaries but will

lead to over segmentation. Meanwhile, the target threshold is

provided to determine surface boundaries of the final region. If

target threshold is huge, the region growing will be stopped before

reaching the desired boundaries. It may result tiny and incomplete

region. If the target threshold is too small, the region growing will

become over flow. It means several regions may be wrongly

connected and part of the surface boundary is over grown.

Iterative 3D region growing aims to grow multiple bones

simultaneously and emphasize accurate segmentation of the

closed surface to the joint area. Two steps of multi region

segmentation procedures are performed in pairs: initial seed

region growing and simultaneous multiple regions growing.

Whenever a seed point is found, it follows the propagation of the

seed points to obtain a seed region. The procedure is fully

explained in Algorithm 1.

Algorithm 1, multiple region growth:

1. All seed regions are given

2. Find all the voxels on the region boundary (fronts/Fj) for all

seed regions (m=1)

3. The gray value of the bone is larger than its surrounding

tissues (Tm = Ti - m T)

4. Compute the average gray value of all voxels on each fronts

(Gavg, i)

5. Arrange the front in terms of Gavg, i, from maximum to

minimum (j=1)

6. Take the voxels in Fj

7. Expand each voxel along its six neighborhoods, stop the

expansion if the gray value of its neighborhoods (G) < Tm

8. Once the voxels on the front have been test, continuing by

test the voxel of next front

9. When all front has been tested, the inner loop of the iteration

is completed (j-j+1)

10. The index m is increased by 1 to reduce the intermediate

target threshold (Tm) for T

11. If Tm Tt, the process go back to first step for the next cycle

of expansion with an upgraded Tm

12. If Tm Tt, the process stop (m-m+1)

13. Region growing finished.

3.2 Fractured Bone Reduction The purposed bone reduction is to relocated fracture bone into

original position. The 3D bone reduction is used to shift the 2D

images approach limitation in term of fractured fragments,

fractured lines and assembling illustration. Manual and semi-

automatic bone reduction methods are provided to recover the

displacement of fractured bones by transforming to the original

positions and orientation in space. In manual bone reduction,

every single bone fragment is transformed through a user interface.

Three kinds of transformation is applied such as translating the

part on the view plane, rotating the movable part along the surface

normal to the viewing plane and rotating the movable part relative

to the part coordinate. The manual reduction result is judged

visually.

In semi-automatic bone reduction, two algorithms are employed:

multi-point positioning and mirror positioning. For the multi-point

positioning method, a number pairs of points along the fractured

contour are selected in sequence. A coordinate transformation

algorithm is used for aligning a piece of fragment respect to its

counterpart [11]. The accuracy of the alignment depends mainly

on the accuracy of the point pairs selected. The main factors affect

to this algorithm is the feature points on the fractured contour may

not be visible to distinct and selected. For the mirror positioning

method, the symmetry property of the bone is used to align the

broken fragment with respect to the normal half of bone. A mirror

plane is determined automatically based on the middle plane of

the broken fragment. The normal half of the bone is mirrored onto

the broken side and serves as a reference. The alignment of each

broken fragment is performed regarding to reference bone. The

entire of broken fragments are continuously transformed to

recover the original position in space. The accuracy of this

algorithm is affected by two factors, symmetry property

influencing the error of transformation and the algorithm does not

CT Images

Multiple region

growing

Generation of automatic

seed regions

Simultaneously growth of

multiple regions

Post-processing

Region re-combination Region re-segmentation

Determination of

thresholds, Ti and Tt

ISO-surface from CT

Image

STL model

14

Page 12: 2016 3rd International Conference on Biomedical and

work for tiny fragments. In serious injuries with comminuted

fracture bone, both manual and semi-automatic bone reduction

could be permitted applied simultaneously. The manual bone

reduction is required to deal with cases which cannot be dealt with

by the semi-automatic algorithms. In common practice, the mirror

positioning is first conducted following by multi-pairs points and

adjusting manually. The procedure of the proposed algorithm for

reduction fractured bone is shown in Figure 3.

Figure 3. Flow diagram of bone reduction.

Determining position and orientation of an object in 3D space

may refer to the alignment of the part coordinates to match the

model coordinates. To determine the coordinate transformation

matrix between the reference point and the target point,

the singular value decomposition (SVD) method is applied. A

schematic illustration describes the generating steps of three

measurement points method to evaluate the surface normal is

shown in Figure 4.

Figure 4. Three measurement points method. (a) The unit disc

with the three canonical unit vectors, (b) Unit disc

transformed by rotating axis [11].

Three points P1, P2 and P3 are assigned in a circle on the XY

plane. The center is on the origin of the circle and a small radius

assigned freely (). Any point on the circle is described (cos,

sin, 0)T, where =120.

The coordinate transformation matrix T can be decomposed into a

translation vector D and a rotation matrix R, as shown below [11]:

[

]

if Pt = (px, py, pz)

T, the translation vector D = (px, py, pz)T.

The rotation matrix R, can be obtained by computing the Z axis

( rotating to reverse direction . The steps of computation

rotation matrix R:

- The back-off direction = (nx, ny, nz)

- The rotation axis = ( = (-ny, nx, 0)T

- The angle between ( and

- The rotation matrix:

[

]

- The homogeneous transformation matrix:

[

]

The coordinate transformation is performed to convert three

points P1, P2 and P3 into the 3D space using the homogeneous

transformation matrix T. The new three points are denoted as P’1,

P’2 and P’3.

The part coordinate setup in the manufacturing process is adopted

to solve the positioning problem in reduction broken bone

fragments. The proposed algorithm consists in two stages: rough

and fine positioning. The rough positioning is implemented first to

obtain a series of coordinate transformations between the part

coordinates and model coordinates. First, the part coordinates are

brought into the neighborhood of the model coordinate. Next, the

fine positioning is implemented. It is an iterative measurement

and organized as automatic procedure. The procedure for part

registration process is fully explained in Algorithm 2.

Algorithm 2, part registration:

1. Find the initial coordinate transformation matrix by defining

the original measurement and the reference points.

-Transform part coordinate first

-Transform measurement point to the new part coordinate

2. Optimize of the surface points and the part coordinate

3. Upgrade the part coordinate

4. Re-measure the part

5. Compute the transformation matrix T

6. Upgrade the part coordinate with T

7. If the error (

[∑ ‖ ‖

] is not

satisfied, the process go back to re-measure the part step.

8. Once the error satisfied, the process stops.

9. Part registration finished

Manual bone reduction

Manually

transformation

Fractured Bone Reduction

Semi-automatic bone

reduction

Visually

validation

Multi point

positioning

Mirror

positioning

Selecting reference

points

Alignment fragment

(Coordinate Transform Algorithm)

Recognizing

symmetry properties

Determine

mirror plane

Selecting the normal

bone as reference

Alignment fragment

(Registration Algorithm)

CT Images

ISO-surface from CT

Image

Fractured Bone

Segmentation

InterfaceImplant/screw

model

Database

CAD model

(2)

(3)

(1)

15

Page 13: 2016 3rd International Conference on Biomedical and

4. DISCUSSION In this study, several actual clinical cases by using real patients’

CT images are provided to illustrate the feasibility of the proposed

method. In segmentation of multi-part of bones, the most difficult

problem is to separate clearly bones near the joint area. Most

methods are not efficient enough as each bone must be processed

one by one. We provide a technique to model and segment bone

tissue from CT images, with different bone segments being

modeled individually.

In Figure 5, a B1 type of fractured pelvic pointed to illustrate the

segmentation process, which indicates that the target threshold is

grown independently. The common target threshold assigned is Tt

= 600, Figure 5(a) shows that the regions are not grown

completely. Then, the target threshold value is reduced to Tt = 185,

Figure 5(b) shows that all region are completely grown. A

displacement of hip bone near to sacroiliac joint can be observed

respect to 2D of CT image. The target threshold for bone

segmentation is set by the user interface and initial threshold is

fixed on 180. However, the overflow occurs in many regions that

make the broken fragments are not displayed clearly. For

example, the broken pubis is only displayed in a single color.

Therefore, re-segmentation is required to visible recognize the

broken fragments.

Figure 5. Pelvic case (Male 18Y), the fractured bone

segmentation, (a) in-complete segmented, (b) a completed

segmented of 3D image.

In order to show all particular fractured bones, the re-

segmentation is applied. The region re-segmentation procedure

applied semi-automatically, where the user only specifies the

target region and initial threshold. The re-segmentation can be

implemented repeatedly until all regions are properly segmented.

Region re-combination procedure is employed to combine

separated regions into a single region. The region combination

algorithm blends the adjacent region into another. It can be

employed continually until all undesired regions are blended

properly. The accuracy of the segmentation is evaluated by

observing the number of colors in the bone region. Figure 6

depicts bone region segmentation result that shows incorrect

separation process. Each of regions is identified by its own color.

A single region of sacrum separated into four regions and five

regions of hip bone. This occurs after simultaneous multiple

region growing, where all segmented regions are displayed in

different colors. It means one particular region of the normal bone

ideally should have only one color.

Figure 6. The semi-automatic Pelvic segmentation process, (a)

re- segmentation and (b) re- combination.

The integrity of broken fragments is mainly concerned. In this

case, a pelvic is unstable fracture due to lateral compression. To

recover this fracture, pelvic is pushed inward. The left hip bone

around the sacroiliac joint has displaced from the original position.

Manual and automatic reduction is used simultaneously as

depicted in Figure 7. The mirror positioning employed a normal

bone or a larger part of bone as a reference and mirrored it onto

the broken side. The displaced fragment of the pubis is aligned

along its displaced line and the right hip bone is aligned to

become symmetric with the left hip bone. In Figure 7(a), manual

reduction is required to adjust the positions and angles of the

displaced parts. The part of broken bone is move piece by piece

through a user interface with respect to the viewing plane. The

translation is either along the XY, XZ or YZ plane. In addition,

for rotation is rotation in X-axis, rotation in Y-axis and rotation in

Z-axis. The result of the manual bone reduction is judged visually.

A mirror plane is determined from the sacrum. Normal bone (right

hip bone) is mirrored to the broken bone (left hip bone) to serve as

a reference bone. Then, each of the broken bones is aligned with

respect to the reference bone, as shown in Figure 7(b). All broken

bone fragments are transformed to fit the original position in

space.

Figure 7. The Pelvic reduction, (a) repositioning displaced hip

joint, (b) mirror positioning technique.

Figure 8 depicts the result of pelvic reduction process and

alignment using the registration algorithm. The integrity of the

pelvic rings is the primary concerns. The hip bone is adjusted to

become symmetric with the reference.

(a)

(b)

Displaced area

before reduction

(a)

(1)

(2)

(3)(4)

(1)(2)

(3)

(4)(5)

(1)

(2)

(3)(4)

(5)

(1) (1)(1)

(b)

Displaced area

14 mm 5.5 mm

(a) (b)

(a)

(b)

(c)

(d)(e)

(a’)

(b’)

(c’)

(d’)

(e’)

16

Page 14: 2016 3rd International Conference on Biomedical and

Figure 8. The Pelvic registration

The other clinical of fractured bone cases examined to present the

performance of method proposed. Diagnosing of fractured bone

classification refers to the AO classification [12]. Figure 9

presents the Tibia Plateau with the fracture type 41-B1. The

fractured bone registration split of the lateral surface in 2

fragments recovered. Figure 10 depicts the lateral Tibia Plateau

split depression. The fracture type classification is 41-B3,

separated fractured bone into 5 fragments. The Tibia Proximal

with complete articular and multi-fragmentary is diagnosed 41-

C3.2 as shown in Figure 11.

Figure 9. Ilustration of the reduction results, Tibia Plateau

type 41-B1 FR (Male 16Y), the fractured bone registration

with 2 fragments recovered.

Figure 10. Ilustration of the reduction results, Tibia Plateau

type 41-B3 FR (Male 60Y), the fractured bone registration

with 5 fragments recovered.

Figure 11. Ilustration of the reduction results, Tibia Proximal

type 41-C3.2 FR (Male 48Y), the registration with 6 fragments

recovered.

The broken Distal Femur is diagnosed type 33-C2 with intact

wedge. Fractured femur has complete articular fractures and

multi-fragmentary. Figure 12 presents the registration of fractured

bone split within 4 fragments.

Figure 12. Ilustration of the reduction results, Distal Femur

type 33-C2 FR (Male 17Y), the registration with 4 fragments

recovered.

The comminuted fracture calcaneus is diagnosed 82-C type.

Figure 13 depicts the fractured calcaneus that registered 5

fragments for recovery to original position.

Figure 13. Ilustration of the reduction results, Calcaneus type

82-C FR (Male 42Y), the fractured bone registration with 5

fragments recovered.

5. CONCLUSION This paper presents a technique to model, segment and recover

fractured bone from CT images. A multi-region segmentation

algorithm has been used to reduce processing time and get

efficiency. In region growing, initial threshold and target

threshold are played to determine boundary of region. The target

threshold value should be controlled on common range value to

ensure all region growth completely. A manual and a semi-

automatic bone reduction algorithm have been performed to deal

with different kinds of fractured bone. Two semi-automatic bone

reduction algorithm, multi point and mirror positioning, were

applied to improve the efficiency of manual reduction. Multi-point

positioning, a number pairs of points is selected sequentially along

the fractured contour. The accuracy of the transformation highly

depends on the accuracy of the point pairs selected. The manual

reduction still required to deal with cases which cannot be caught

by semi-automatic algorithm. The proposed algorithm of bone

segmentation and reduction have contributed to improve the

feasibility of 3D preoperative planning for fractured bones and

provide more valuable information prior to clinician decision

making.

6. REFERENCES [1] Lee, P. Y., Lai, J. Y., Hu, Y. S., Huang, C. Y., Tsai, Y. C.

and Ueng, W. D. 2012. Virtual 3D planning of Pelvic

fracture reduction and implant placement. Biomedical

Engineering: Applications, Basis and Communications. 24, 3

(June 2012), 1250007-1 - 1250007-18.

DOI= http://dx.doi.org/10.4015/S101623721250007X.

[2] Paulano, F., Jimenez, J. J., Pulido, R. 2014. 3D segmentation

and labeling of fractured bone from CT images. The Visual

Computer. 30, 6 (June 2014), 939–948.

DOI=10.1007/s00371-014-0963-0.

Displaced area

after reduction

118.5 mm

17

Page 15: 2016 3rd International Conference on Biomedical and

[3] Huang, C. Y., Lee, P. Y., Lai, J. Y., Luo, L. J., Tsai, Y. C.

and Lin, S. C. 2011. Simultaneous segmentation of bone

regions using Multiple-Level Threshold. Computer Aided

Design and Application. 8, 2 (April 2011), 269-288.

DOI= 10.3722/cadaps.2011.269-288.

[4] Huang, C. Y., Luo, L. J., Lee, P. Y., Lai, J. Y., Wang, W. T.

and Lin, S. C. 2010. Efficient segmentation algorithm for 3D

bone models construction on medical images. Journal of

Medical and Biological Engineering. 31, 6 (May 2010), 375-

386. DOI= 10.5405/jmbe.734.

[5] Wu, J., Belle, A., Hargraves, R. H., Cockrell, C., Tang, Y.

and Najarian K. 2014. Bone segmentation and 3D

visualization of CT images for traumatic Pelvic injuries.

International Journal of Imaging Systems and Technology,

24,1 (February 2014), 29–38. DOI= 10.1002/ima.22076.

[6] Nysjö, J., Malmberg, F., Sintorn, I. M. and Nyström, I. 2015.

Bone split-A 3D texture painting tool for interactive bone

separation in CT images. Journal of WSCG. 23, 2. 157-166.

ISSN 1213-6972.

[7] Joshi, M. U., Gandhe, S. T. 2016. Bone fracture detection

using active contour segmentation. International Journal of

Applied Engineering Research. 11, 6. 4230-4234.

ISSN 0973-4562.

[8] Jiménez-Delgado, J.J., Godino, F.P., Pulido, R., Ramírez, R.,

Jiménez-Pérez, J.R. 2016. Computer assisted preoperative

planning of bone fracture reduction: Simulation techniques

and new trends. Medical Image Analysis. 30, (May 2016).

30–45. DOI= http://dx.doi.org/10.1016/j.media.2015.12.005

[9] Lee, P. Y., Lai, J.Y., Yu, S. H., Huang, C. Y., Hu, Y. S. and

Feng, C. L. 2013. Computer-assisted fracture reduction and

fixation simulation for pelvic fractures. Journal of Medical

and Biological Engineering. 34, 4 (September 2013). 368-

376. DOI= doi: 10.5405/jmbe.1605.

[10] Irwansyah, Lai, J. Y., Lee, P. Y., Chung, C. Y. 2016.

Development and clinic study of an integrated preoperative

planning system for orthopedic surgery. In Proceeding of the

XIV International Symposium on 3D Analysis of Human

Movement (Taipei, Taiwan, July 18-21, 2016).141-144.

[11] Lai, J. Y., Chen, K. J. 2007. Localization of parts with

irregular shape for CMM inspection. The International

Journal of Advanced Manufacturing Technology. 32, 11

(May 2007). 1188–1200. DOI= http://doi: 10.1007/s00170-

006-0430-9.

[12] Rüedi, T. P., Murphy, W. M. 2000. AO Principles of

Fracture Management. Thieme/AO publishing. Stuttgart.

New York. ISBN: 3-13-117441-2 (GTV).

18