Coronary Plaque Boundary Detection in Intravascular Ultrasound Image by Using Hybrid Modified Level Set Method and Fuzzy Inference

Similar documents
2. The differential pressure across control valves must not vary too much

6 λk 0gh 3 þ 1 6 k. þ 1. z ¼ ð1 z ÞH p þ b qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

The Extension of Weight Determining Method for Weighted Zone Scoring in Information Retrieval

Development and validation of a new mass-consistent model using terraininfluenced

Three Dimensional Analysis and Application of A Line Solver for the Re-circulating Flows Using Multi-Grid Methods

Experimental and Numerical Studies on Flocculation of Sand-Mud Suspensions

Optimization Model of Oil-Volume Marking with Tilted Oil Tank

Ana Torre Carrillo Facultad de Ingeniería Civil, Universidad Nacional de Ingeniería Av. Túpac Amaru 210, Lima 25 (Peru)

Work and Energy. A Force F will do work on a particle only when the particle undergoes a displacement in the direction of the force.

The collision avoidance control algorithm of the UAV formation flight

A Note on Software Protection and Social Welfare

Practical design approach for trapezoidal modulation of a radio-frequency quadrupole

DESIGN OF MACHINES AND STRUCTURES

HCR OF HEAT PUMP ROOM AIR CONDITIONER IN CHINA. Beijing , China

Overall stability of multi-span portal sheds at right-angles to the portal spans

Dimitrios Krontsos Technological Educational Institute of Thessaloniki, Greece Thessaloniki 57400, Greece

CALIBRATION ALGORITHM FOR CURRENT-OUTPUT R-2R LADDERS

Spatiotemporal Analysis of Marriage and Marital Fertility in Japan: Using Geographically Weighted Regression

Effect of user fee on patient s welfare and efficiency in a two tier health care market

Dominance Testing for Pro-Poor Growth with an Application to European Growth

Consumer Price Indices

Reflections on the drinking bowl 'Balance'

Can Survey Bootstrap Replicates Be Used for Cross-Validation?

Effect of Twisted-tape Inserts on Heat Transfer in a Tube

Tunnel-soil-pile interaction in the prediction of vibration from underground railways: validation of the sub-models

SLOPE STABILITY ANALYSIS IN SEISMIC AREAS BY A RELIABILITY APPROACH

Study of microrelief influence on optical output coefficient of GaN-based LED

Inventory Decision Model of Single-echelon and Two-indenture Repairable Spares

Understanding the Performance of Cooperative Web Caching Systems

Surface Roughness Measurement on a Wing Aircraft by Speckle Correlation

16.1 Volume of Prisms and Cylinders

BLIND SOURCE SEPARATION BASED ON SPACE-TIME-FREQUENCY DIVERSITY. Scott Rickard, Radu Balan, Justinian Rosca

Open Access Prediction of the Seismic Response of Steel Frames with Concentric Diagonal Bracings

Thermodynamic Analysis Of Steam Ejector Refrigeration Cycle

Evaluation Method of Banking System Stability Based on the Volume of Subsystems

ANALYSIS OF WORK ROLL THERMAL BEHAVIOR FOR 1450MM HOT STRIP MILL WITH GENETIC ALGORITHM

Numerical Study on Vane Seal Lubrication Model Considering Surface Morphology Chuang JIA *, Liang-xi XIE and Zhong-zheng LUO

Math Practice Use a Formula

Lecture 15: Effect modification, and confounding in logistic regression

Coffeemar. Installation and maintenance guide. Contents: Ctra, Marcilla, Km PERALTA, Navarra SPAIN

Volumes of Pyramids. Essential Question How can you find the volume of a pyramid?

INVESTIGATION OF TURBULENT BOUNDARY LAYER OVER FORWARD-FACING STEP BY MEANS OF DNS

Numerical Simulation of Stresses in Thin-rimmed Spur Gears with Keyway B. Brůžek, E. Leidich

Balanced Binary Trees

2 2D 2F. 1pc for each 20 m of wire. h (min. 45) h (min. 45) 3AC. see details J, E

The Policy Performance of NFSF and Slippage in Futures Markets

A SIMPLE CORRECTION TO THE FIRST ORDER SHEAR DEFORMATION SHELL FINITE ELEMENT FORMULATIONS

Engel s Law in the Global Economy: Demand-Induced Patterns of Structural Change, Innovation, and Trade*

The Vapor Compression Cycle in Aircraft Air-Conditioning Systems

Non-contact Respiration Measurement Using Structured Light 3-D Sensor

Structural Changes in Household Income Inequality in Korea: A New Decomposition Including Labor Supply Changes * Chulhee Lee

Revision Topic 12: Area and Volume Area of simple shapes

4.2 Using Similar Shapes

Modeling the Greek Electricity Market

Prediction of steel plate deformation due to triangle heating using the inherent strain method

Calculation of Theoretical Torque and Displacement in an Internal Gear Pump

青藜苑教育 Example : Find te area of te following trapezium. 7cm 4.5cm cm To find te area, you add te parallel sides 7

Spline Regression in the Estimation of the Finite Population Total

Variance Estimation of the Design Effect

Annex 16. Methodological Tool. Tool to determine project emissions from flaring gases containing methane

of Beef Top Sirloin Steaks

Analysing the energy consumption of air handling units by Hungarian and international methods

A New Information Hiding Method for Image Watermarking Based on Mojette Transform

Characteristics and dead-time of GM-tube

10 Fingers of Death: Algorithms for Combat Killing Roger Smith and Don Stoner Titan Corporation

RISING FAMILY INCOME INEQUALITY IN THE UNITED STATES, : IMPACTS OF CHANGING LABOR SUPPLY, WAGES, AND FAMILY STRUCTURE.

A centralized/decentralized control approach for periodic systems with application to chromatographic separation processes

Paddy McGinty's Goat

OPTIMIZATION OF TWO-DIMENSIONAL WING IN GROUND EFFECT CONSIDERING AERODYNAMIC CENTER OF HEIGHT AND LIFT

Oil spill forecasting in the Mediterranean Sea

'""' USAFA/ Coord.{!tr lv~ ""' DFCE... ~A.., USAFA/ DFER. Sign C:.dl A:>.-').l'. 23 \,;'~ rs- 7 USAFA-DF-PA- CJ

Rev. 3/15/13 COMMODITIES & STANDARDS ** NTE CCODE PRODUCT PHOS COLI KEY: 16A Cheddar Cheese 10 ** n less than 1 microgram phenol/gram or equivalent

LIQUEUR FORMULAE 205. Lemon peel No. 80. Add. Nutmeg water. Add enough water to make up to. Almonds (sweet) ANISETTE. Green anis

Coffee Differentiation: Demand Analysis at Retail Level in the US Market

Public Assessment Report. Scientific discussion. Gestoden/Ethinylestradiol 75/20 Stragen Gestoden/Ethinylestradiol 75/30 Stragen

20.1 Heights and distances

Development and application of a rural water supply assessment tool in Brazil

International Plant Protection Convention Page 1 of 10

Reinforcement in concrete piles embedded in sand

Installation the DELTABEAM Frame

Forecasting Harvest Area and Production of Strawberry Using Time Series Analyses

Goal: Measure the pump curve(s)

REAL-TIME TRACKING OF NON-RIGID OBJECTS USING MODIFIED KERNEL-BASED MEAN SHIFT AND OPTIMAL PREDICTOIN

Study of Steam Export Transients in a Combined Cycle Power Plant

Christmas Shopping Blues

TORQUE CONVERTER MODELLING FOR ACCELERATION SIMULATION

Modeling of Heat Losses Within Combustion Chamber of Diesel Engines

MORNING FARE EARLY RISER THE CLASSIC CONTINENTAL COLLEGE OF DUPAGE BREAKFAST BUFFET. Pastry Basket, Assorted Juice & Coffee

Earthquake countermeasures for wood houses to mitigate seismic loss and total cost

Optimization Design of a Piezoelectric Actuator with Orthogonal Theory

Developing a building damage function using SAR images and post-event data after the Typhoon Haiyan in The Philippines

To find the volume of a pyramid and of a cone

Modelling Beta Risk for New Zealand Industry Portfolios

Detecting Melamine Adulteration in Milk Powder

Designing Ranking Systems for Hotels on Travel Search Engines by Mining User-Generated and Crowdsourced Content

Homework 7. problems: 9.33, 9.40, 9.65, 9.78

Ethnic Sorting in the Netherlands

Description of Danish Practices in Retail Trade Statistics.

Point Pollution Sources Dimensioning

Measurement and Study of Soil ph and Conductivity in Grape Vineyards

Transcription:

Coronary Plaqe Bondary Detecton n Intravaclar Ultraond Image by Ung Hybrd Modfed Level Set Metod and Fzzy Inference Syafl nam ab Ej Ucno ac and Norak Setake a a Gradate Scool of Scence and Engneerng Yamagc Unverty 1677-1 Yoda Yamagc 753-851 Jaan {r501a cno netake}@yamagc-.ac.j b Deartment of Matematc Unverty of Brajaya Jl. Veteran Malang 65145 Indonea c Fzzy Logc Sytem Inttte 680-41 Kaaz Izka 80-0067 Jaan BSTRCT T aer decrbe a bondary detecton of coronary laqe by a ybrd of a modfed level et metod and a fzzy nference. Te tandard level et metod to detect an mage bondary commonly e an mage gradent for calclatng a eed fncton. Bt te eed fncton of te level et cannot ork ell tn an ntravaclar ltraond (IVUS) mage c te target of t aer. Terefore e rooed a metod for coronary bondary detecton n IVUS mage by ng a modfed level et metod. In tat metod te mage gradent n te eed fncton bttted by te egted mage earablty. Hoever ome regon of te IVUS mage often become adoed and ten contan no textre nformaton de to te reence of te gde re. T te modfed level et metod fal to detect te laqe bondary n toe regon. To olve t roblem e frter rooe n t aer a ybrd of te modfed level et metod and te T-S fzzy model. Te reent metod a been more ccefl n te accracy of laqe bondary detecton. KEYWORDS Coronary laqe bondary detecton ntravaclar ltraond mage modfed level et metod T-S fzzy model. 1 INTRODUCTION cte coronary yndrome (CS) one of te leadng reaon for otalzaton n te orld. It caed by a rtre of vlnerable laqe c blt nde te coronary artere cang eart attack. One of te medcal magng metod to dagnoe CS te ntravaclar ltraond (IVUS) metod [1]. IVUS mage are ed for te nner and oter coronary laqe bondary detecton to evalate te qanttatve aement of te coronary laqe comoton. Detectng coronary laqe bondare a very dffclt tak for medcal doctor. In order to redce te orkload of te medcal doctor an atomatc coronary laqe bondary detecton metod t g accracy a become a necety. Prevo ork [] and [3] ere rooed to olve t roblem. Te fzzy nference model emloyed n te revo ork Takag-Sgeno (T-S) fzzy model [4]. Level et metod ave been dely ed n mage egmentaton and ave everal advantage over oter egmentaton metod c a te nake metod regon grot and treoldng. Te tandard level et metod commonly e a gradent for calclatng a eed fncton bt te eed fncton of te level et cannot ork ell tn an IVUS mage. Terefore e rooed a metod for coronary laqe bondary detecon by ng te egted mage earablty metod and te level et metod [5]. In [5] te mage gradent n te eed fncton bttted by te egted mage earablty. Hoever t faled to detect te laqe bondary for everal regon of te mage a on n Fg. 1(b). Te laqe bondary detecton n IVUS mage very dffclt becae a regon of te IVUS mage often become adoed and ten contan no textre nformaton de to te reence of te gde re a on n Fg 1. T te metod [5] fal to detect te laqe bondary. ISBN: 978-0-9891305--3 013 SDIWC 69

(a) (b) Fgre 1. Gde re ado roblem. (a) Gde re ado n IVUS Image. (b) Bondary detecton error on gde re ado. We alo rooed a metod for mage bondary detecton by ng te modfed level et metod [6]. In [6] e ave modfed te eed fncton to make te level et metod more ccefl n detectng te bondare regardle of extng noe. In t aer e frter rooe a metod for bondary detecton by ng a ybrd of te modfed level metod and te T-S fzzy model. We detect te laqe bondary otde te gde re ado regon by ng te modfed level et metod. fter tat te bondary relt of te level et ed to nterolate te bondary nde te gde re regon by ng te T-S fzzy model. Te effectvene of te reent metod a verfed trog te exerment ng real IVUS mage. CORONRY PLQUE BOUNDRY CLCULTIONS IN IVUS IMGE Te IVUS metod one of te alcaton of ltraond tecnology c a many alcaton n medcal dagno. Te IVUS metod an ngeno metod ed to oberve te nde of te blood veel trog te rrondng blood colmn from te otde valzng te coronary laqe n vvo. Te mage on n Fg. called a B-mode mage c contrcted of te amltde nformaton of te receved ltraond rado freqency (RF) gnal. Te amled RF gnal are tranformed nto ntente and te ntente n all radal drecton are ed to form a tomograc cro-ectonal mage of a coronary artery. Te bondare of laqe need to be calclated for te dagno of laqe. Fg. o to bondare of laqe.e. lmnal bondary (LB) and adventtal bondary (B). Fgre. IVUS B-mode mage..1 notroc Dffon Flter Te anotroc dffon flter a orgnally rooed by Perona and Malk [7]. Te dcrete veron of PMD defned by: I ( n1) I g( I ) I (1) ere ( x y) and are te coordnate of te xel of concern and t negborng xel (n) (n) reectvely. I an ntenty at I t an teraton cont n. rereent te dffon drecton. te nmber of xel n te negborng area. a arameter. monotoncally decreang fncton of te gradent of an mage ally adoted a g (). Te gradent of te mage calclated by: I I I 1 g( z) z 1 K (3) ere K a arameter c control te trengt of dffon. g () take a large vale at te regon ere te ntenty gradent are lo. Vce vera t take a lo vale at te regon ere te ntenty gradent are g. Te mage earablty of IVUS mage on n e.g. Fg 6.. Image Searablty Te earablty [8] for xel on n Fg. 3 defned by: ISBN: 978-0-9891305--3 013 SDIWC 70

n( I I ) nb ( I B I ) (4) S ( I I ) k1 ere n and n B rereent te nmber of te xel n te regon of and B reectvely. I and I B rereent te average of ntente n te regon of and B. I tand for te average of te ntente n te combned regon of and B. S and I k are te nmber of te xel and te level of ntenty of te k-t xel n te combned regon of and B. Te egted mage earablty for xel defned by: k Imax I IB (5) Imax Imax ere I max te maxmm ntenty of te ole IVUS mage. atfe 0 1 and t take a large vale en to regon are earated from eac oter. 3 PROPOSED METHOD We rooe a metod for te laqe bondary detecton by ng a ybrd of te modfed level et metod and te T-S fzzy model. Fgre 3. Calclaton of te mage earablty. 3.1 Gde Wre Detecton IVUS mage often become adoed and ten contan no textre nformaton de to te reence of te gde re. T te level et metod fal to detect te laqe bondary. To olve t roblem n te frt te te gde re ado old be detected. Te rocedre of detectng te gde re ado regon mmarzed n te follong te: 1) Convert te B-mode mage from Cartean coordnate ytem on n Fg. 1(a) to olar coordnate ytem on n Fg 1(b). ) Te locaton of te gde re ado ll be atomatcally detected. In t te te mean gray-level of every colmn of te olar coordnate ytem calclated a on n Fg. 4 (te lne crve). From Fg. 4 t can be oberved tat te gde re adoed locaton arond te colmn a te mnmm vale of te mean gray-level (tck black lne). Te locaton of te gde re ado a on n area beteen to daed lne determned by ng te mnmm vale of te mean gray-level. 3. Plaqe Bondary Detecton by Ung te Modfed Level Set Metod Let conder a dynamc arametrc contor C ( x( t) y( t)) te crve evolton of c defned by: C / t FN (6) ere t a et ont n tme a crve arameter N te nard normal vector to te crve C and F te eed fncton tat control te moton of te contor. Te crve evolton n (6) n term of a arameterzed contor can be converted to a level et formlaton by embeddng te dynamc contor C a te zero level et of a tme deendent level et fncton ( x y t). mng tat te embeddng level et fncton take te negatve vale nde te zero level contor and te otve vale otde te nard normal vector can be exreed a N / ere a gradent oerator. Ten te crve evolton n (6) converted to: ISBN: 978-0-9891305--3 013 SDIWC 71

Fgre 4. Gde re ado detecton. Fgre 8. Bondary detecton relt by ng a ybrd of te level et metod and te T-S fzzy model. / t F (7) c referred to a a level et evolton eqaton. Crve evolton tat ed by te level et (x) defned by: Fgre 5. IVUS Image after gnorng gde re ado. Fgre 6. Wegted mage earablty and ntal contor of te modfed level et. / t dv( d ( ) ) ( ) dv( g / ) g ( ) (8) ere a drac delta fncton dv a dvergence oerator and g a eed fncton. eed fncton gven by: g 1/(1 ( G * I ) (9) ere G te Gaan Flter and I te mage [9]. Te rocedre of laqe bondary detecton rooed mmarzed n te follong te: 1) Detect te gde re ado regon (angle ndex beteen a and c). ) Merge area c to 55 and 0 to b by lacng te latter on te rgt de a on n Fg. 5. 3) Extend te mage by rerodcng bot de of te mage bt onto te correondng oote end a te level et often fal to detect te edge of te mage. 4) Redce te eckle noe by ng PMD flterng. 5) Calclate te egted mage earablty tat ed for calclatng te eed fncton [5] c on n Fg. 6. 6) Calclate te modfed eed fncton by: Fgre 7. Bondary detecton relt by ng te modfed level et metod. 1 g 1 b (10) ISBN: 978-0-9891305--3 013 SDIWC 7

ere b a contant c control te moton of te contor. We e te eed fncton c rooed n [6]. Bt n t aer e bttte te mage gradent t te egted mage earablty. 7) Gve te ntal contor of te level et a on n Fg. 6. 8) Calclate te contor evolton by ng (8). 9) Calclate te ne contor. 10) Reeat te 8 and 9 ntl t converge or te maxmm nmber of teraton reaced. Te laqe bondary detected by ng te modfed level et on n Fg. 7. 3.3 Plaqe Bondary Detecton by Ung te T-S Fzzy Model Te laqe bondary detecton by te T-S fzzy model on by black lne on te angle nterval [a c] n Fg. 8. 4 EXPERIMENTL RESULTS In te exerment e ed tree IVUS mage and te rooed metod a comared t te revo metod [] [3]. We ame tat f ntal contor dfferent te laqe bondary by te rooed metod ll rodce dfferent olton. Terefore te roceng for eac mage a reeated 5 tme (5 rn) t dfferent ntal contor. Te dered bondare ere baed on te dfference of mage brgtne. Te laqe bondary n te gde re ado regon nferred by T-S fzzy model. Te laqe bondary aroxmated by te ere of te follong fzzy rle: If x ten f ( x ) a x b (11) ere a fzzy et t te member fncton (MSF) ( x ) x correond to te angle ndex and f ( x ) a lnear fncton. In te antecedent art of te fzzy rle te comlementary tranglar MSF are ed. Te - t rle t tand for an aroxmaton of te laqe bondary by a lnear fncton n te nterval [ z 1 z 1]. Te nferred bondary gven by: (a) ˆ 1 1 y( x ) ( x ) f ( x ) ( x ) f ( x ). (1) Te otmm coeffcent n te coneqent art of te fzzy rle are determned t e of te leat qare metod o a to mnmze te follong error crteron: E ( y yˆ ( x )) (13) ere y laqe bondary tat detected by te modfed level et metod on angle ndex. (b) Fgre 9. Comaron of te laqe bondary detecton. (a) IVUS mage to be roceed. (b) Bondary detecton relt. Black lne and te lne ndcate te detected bondare by te rooed metod and te dered bondary reectvely. ISBN: 978-0-9891305--3 013 SDIWC 73

Table 1. RMSE of bondary detecton relt for mage 1. ( m) Metod LB B Hybrd of te Modfed Level Set Metod and te T-S Fzzy Model 9. 0.1 T-S Fzzy Model [] 13.7 8.4 T-S Fzzy Model Otmzed by Partcle Sarm Otmzaton [3] 1. 8.5 Table. RMSE of bondary detecton relt for mage. ( m) Metod LB B Hybrd of te Modfed Level Set Metod and te T-S Fzzy Model 1.5 18.9 T-S Fzzy Model [] 8.0 35.4 T-S Fzzy Model Otmzed by Partcle Sarm Otmzaton [3] 3.8 33. Table 3. RMSE of bondary detecton relt for mage 3. ( m) Metod LB B Hybrd of te Modfed Level Set Metod and te T-S Fzzy Model 15.6 14.3 T-S Fzzy Model [] 0.0 30. T-S Fzzy Model Otmzed by Partcle Sarm Otmzaton [3] 19.9 40.1 Table 1 and 3 o te root mean qare error (RMSE) beteen te dered and te detected bondare. Te RMSE of te ybrd of te modfed level et metod and te T-S fzzy model mc better tan or revo metod [] [3] for all mage. Fg. 9 o te bet detecton relt by te rooed metod. It oberved from Fg. 9 tat te detected bondare (black) by te rooed metod are cloe to te dered bondare (te). Tee relt o tat te rooed metod ork ell. 5 CONCLUSIONS We ave rooed a metod for coronary laqe detecton tn an IVUS mage by ng a ybrd of te modfed level et metod and te T-S fzzy model. Te rooed metod a mc better detecton erformance tan ever n term of te calclaton accracy. cknoledgment: T ork a orted by te Grant-n-d for Scentfc Reearc (B) of te Jaan Socety for Promoton of Scence (JSPS) nder te Contract No. 3300086. Syafl nam old alo lke to tank DGHE otgradate colar of Indonea for ortng tay at Yamagc Unverty Jaan. 6 REFERENCES [1] B. N. Potkn. L. Bartorell J. M. Geert R. F. Nevlle Y. lmagor W. C. Robert and M. B. Leon Coronary rtery Imagng t Intravaclar Hg- Freqency Ultraond J. Crclaton vol.81 1990.1575 1585. [] T. Koga S. Icyama E. Ucno N. Setake T. Hro and M. Matzak Flly tomatc Bondary Extracton of Coronary Plaqe n IVUS Image by notroc Dffon and T-S Tye Fzzy Inference dvance n Intellgent and Soft Comtng Srnger- Verlag ISC 75 010.139 147. [3] S. nam H. Maa E. Ucno and N. Setake Parameter Tnng by PSO for Fzzy Inference-Baed Coronary Plaqe Extracton n IVUS Image Proceedng of SCIS & ISIS 01 Jaan.146-149 November 01. [4] T. Takag and M. Sgeno Fzzy Identfcaton of Sytem and It lcaton to Modelng and Control IEEE Tranacton on Sytem Man and Cybernetc vol.smc 15 1985.116 13. [5] S. nam H. Maa N. Setake and E. Ucno Coronary Plaqe Bondary Extracton n IVUS Image by Ung Wegted Image Searablty and Level Set Metod Proceedng of te 14t SOFT Ky Cater nnal Conference Jaan.37-38 December 01. [6] S. nam E. Ucno and N. Setake Image Bondary Detecton Ung te Modfed Level Set Metod and a Dffon Flter Proceedng of KES 013 Jaan.161-169 Setember 013. [7] P. Perona and J. Malk Scale-Sace and Edge Detecton Ung notroc Dffon IEEE Tranacton on Pattern naly and Macne Intellgence vol.1 1990. 69-639. [8] K. Fk Edge Extracton Metod Baed on Searablty of Image Featre IEICE Tranacton on Informaton Sytem vol. E78-D 1995.1533-1538. [9] C. L C. X C. G and M. D. Fox. Dtance Reglarzed Level Set Evolton and It lcaton to Image Segmentaton IEEE Tranacton on Image Proceng vol.19 010.343-354. ISBN: 978-0-9891305--3 013 SDIWC 74