Investigation and Assessment of Disorder of Ultrasound B-mode Images

Size: px

Start display at page:

Download "Investigation and Assessment of Disorder of Ultrasound B-mode Images"

Baldwin King
6 years ago
Views:

1 Vol. 7, No., February 00 Ivestigatio ad Assessmet of Disorder of Ultrasoud B-mode Images Vidhi Rawat *,Alok jai**,vibhakar shrimali***, *Departmet of Biomedical Egieerig **Departmet of electroics Egieerig *** Departmet of Trg. & Techical Educatio, Govt. of NCT Delhi, Delhi. Samrat Ashok Techological Istitute Vidisha, Idia. Abstract Digital image plays a vital role i the early detectio of cacers, such as prostate cacer, breast cacer, lugs cacer, cervical cacer. Ultrasoud imagig method is also suitable for early detectio of the abormality of fetus. The accurate detectio of regio of iterest i ultrasoud image is crucial. Sice the result of reflectio, refractio ad deflectio of ultrasoud waves from differet types of tissues with differet acoustic impedace. Usually, the cotrast i ultrasoud image is very low ad weak edges make the image difficult to idetify the fetus regio i the ultrasoud image. So the aalysis of ultrasoud image is more challegig oe. We try to develop a ew algorithmic approach to solve the problem of o clarity ad fid disorder of it. Geerally there is o commo ehacemet approach for oise reductio. This paper proposes differet filterig techiques based o statistical methods for the removal of various oise. The quality of the ehaced images is measured by the statistical quatity measures: Sigal-to-Noise Ratio (SNR), Peak Sigal-to-Noise Ratio (PSNR), ad Root Mea Square Error (RMSE). Keywords- fetus image,sigal-to-noise Ratio (SNR), Peak Sigal-to-Noise Ratio (PSNR), ad Root Mea Square Error (RMSE). I. INTRODUCTION Ultrasoud imagig method is suitable to diagose ad progeies []. The accurate detectio of regio of iterest i ultrasoud image is crucial. Sice the result of reflectio, refractio ad deflectio of ultrasoud waves from differet types of tissues with differet acoustic impedace. Usually, the cotrast i ultrasoud image is very low ad boudary betwee regio of iterest ad backgroud are fuzzy []. fetus regio i the ultrasoud image is ot approachable. So the aalysis of ultrasoud image is more challegig oe.noise is cosidered to be ay measuremet that is ot part of the pheomea of iterest.digital image plays a vital role i the early detectio of abormality i the fetus. Processes goig o i the productio ad capture of real sigal. Ultrasoud is safe, radiatio free, faster ad cheaper. Ultrasoud images themselves will ot give a clear view of a affected regio. So digital processig ca improve the quality of raw ultrasoud images. I this work a software tool called Image Processig Tool has bee developed by employig the histogram equalizatio ad regio growig approach to give a clearer view of the affected regios i the abdome. Ultrasoud is applied for obtaiig images of almost the etire rage of iteral orgas, these iclude the kidey, liver, pacreas, bladder, the fetus durig pregacy. Today, B-mode ultrasoud imagig is oe of the most frequetly used diagostic tools, ot oly because of its real time capabilities, as it allows faster ad more accurate procedures, but also because there is low risk to the patiet ad low cost as compared to other image modalities. A proposed algorithm has bee successfully developed to semi-automate oivasive examiatio of the fetus. Such a system ca be helpful i reducig costs, miimizig exposure of the fetus to ultrasoic radiatio, ad providig a uiform examiatio ad iterpretatio of the results. II. OBSTETRICAL ULTRASOUND IMAGING Normally ultrasoud images will cotai a mixture of differet types of oises. Removal of oises is crucial sice ultrasoud images themselves will ot give a clear view of a affected regio. It is easy to remove the oises if the images are digitized. The it is possible to develop software that suits to remove a type of oise. After removig the oises, clear view of affected regios ca be pipoited. I the ideal case of a cotiuous probability distributio of the gray levels, histogram equalizatio produces a uiform histogram. Ultrasoud exams do ot use ioizig radiatio (as used i x-rays). Because ultrasoud images are captured i real-time. Ultrasoud imagig is a oivasive medical test that helps physicias diagose ad treat medical coditios. Obstetrical ultrasoud provides pictures of a embryo or fetus withi a woma's uterus. A. properties of ultrasoography. Most ultrasoud scaig is oivasive (o eedles or ijectios) ad is usually pailess.. Ultrasoud is widely available, easy-to-use ad less expesive tha other imagig methods. 89

2 Vol. 7, No., February Ultrasoud scaig gives a clear picture of soft tissues that do ot show up well o x-ray images. 4. Ultrasoud causes o health problems ad may be repeated as ofte as is ecessary. 5. Ultrasoud is the preferred imagig modality for the diagosis ad moitorig of pregat wome ad their ubor babies. 6. Ultrasoud has bee used to evaluate pregacy for early four decades ad there has bee o evidece of harm to the patiet, embryo or fetus. Nevertheless, ultrasoud should be performed oly whe cliically idicated. B. Obstetrical ultrasoud is a useful cliical test to Establish the presece of a livig embryo/fetus. Estimate the age of the pregacy. Diagose cogeital abormalities of the fetus. Evaluate the positio of the fetus. Evaluate the positio of the placeta. Determie if there are multiple pregacies. Determie the amout of amiotic fluid aroud the baby. Check for opeig or shorteig of the cervix or mouth of the womb. Assess fetal growth. C. Limitatios of Obstetrical Ultrasoud Imagig Obstetric ultrasoud caot idetify all fetal abormalities. Cosequetly, whe there are cliical or laboratory suspicios for a possible abormality, a pregat woma may have to udergo oradiologic testig such as amiocetesis (the evaluatio of fluid take from the sac surroudig the fetus) or chorioic villus samplig (evaluatio of placetal tissue) to determie the health of the fetus, or she may be referred by her primary care provider to a gerotologist (a obstetricia specializig i high-risk pregacies). Image ehacemet is especially relevat i mammography where the cotrast betwee the soft tissues. These approaches all use a reversible wavelet decompositio, which may be redudat or ot, ad perform the ehacemet by selective modificatio (Amplificatio) of certai wavelet coefficiets prior to recostructio. Whe the weightig scheme is liear, this approach ca be iterpreted as a multiscale versio of traditioal usharp maskig. B. Image segmetatio Segmetatio is the stage where sigificat commitmet is made durig automated aalysis by delieatig structures of iterest ad discrimiatig them from backgroud tissues. Segmetatio algorithm operate o the itesity texture variatios of the image usig techique that iclude thresholdig,regio growig ad patter recogitio techique such as eural etwork. Image segmetatio is a useful tool i may realms icludig idustry, health care, astroomy, ad various other fields. Segmetatio i cocept is a very simple idea. Simply lookig at a image, oe ca tell what regios are cotaied i a picture. This paper discusses two differet regio determiatio techiques: oe that focuses o edge detectio as its mai determiatio characteristic ad aother that uses regio growig to locate separate areas of the image. IV. PROBLEM DESCRIPTION A image may be defied as a two dimesioal fuctio f(x, y), where x ad y are spatial (plae) coordiates, ad the amplitude of f at ay pair of co-ordiates (x, y) is called the itesity or grey level of the image at that poit [5]. Data sets collected by image sesor are geerally cotamiated by oise. The regio of iterest i the image ca be degraded by the impact of imperfect istrumet, the problem with data acquisitio process ad iterferig atural pheomea. Therefore the origial image may ot be suitable for applyig image processig techiques ad aalysis. Thus image ehacemet techique is ofte ecessary ad should be take as the first ad foremost step before image is aalysed. Aother commo form of oise is data dropout oise geerally referred to as speckle oise. This oise is, i fact, caused by errors i data trasmissio [3, 4]. The corrupted pixels are either set to the maximum value, which is somethig like a sow i image or have sigle bits flipped over. III. ELEMENTS OF BIOMEDICAL IMAGE PROCESSING A. Image Ehacemet Ehacemet algorithms are used to reduce image oise ad icrease the cotrast of structure of iterest. whe i images the distictio betwee ormal ad abormal tissue is occur the accurate iterpretatio may become difficult if oise level are relatively high. Ehacemet improve the quality of image ad facilitates diagosis. Ehacemet techiques are geerally used to provide a clear image for a huma observer. A. Image Data Idepedet Noise It is described by a additive oise model, where the recorded image, i(m, ) is the sum of the true image t(m, ) ad the oise (m, )[5, 8, 9] i(m,)=t(m,)+(m,) The oise (m, ) is ofte zero-mea ad described by its variace. I fact, the impact of the oise o the image is ofte described by the SNR [6], which is give by 90

3 Vol. 7, No., February 00 Where, ad t SNR = t are the variaces of the true image ad the recorded image respectively. I may cases, additive oise is evely distributed over the frequecy domai (white oise), whereas a image cotais mostly low frequecy iformatio. Therefore, such a oise is domiat for high frequecies ad is geerally referred as Gaussia oise ad it is observed i atural images [0, ]. B. Image Data Depedet Noise Data-depedet oise (e.g. arisig whe moochromatic radiatio is scattered from a surface whose roughess is of the order of a wavelegth, causig wave iterferece which results i image speckle), it is possible to model oise with a multiplicative, or o-liear, model. These models are mathematically more complicated; hece, if possible, the oise is assumed to be data idepedet. ) Speckle Noise Aother commo form of oise is data dropout oise geerally referred to as speckle oise. This oise is, i fact, caused by errors i data trasmissio [3, 4]. The corrupted pixels are either set to the maximum value, which is somethig like a sow i image or have sigle bits flipped over. This kid of oise affects the ultrasoud images [4]. Speckle oise has the characteristic of multiplicative oise [5]. Speckle oise follows a gamma distributio ad is give as α g / a g e f ( g) = α ( α )! a The primary objective of the image ehacemet is to adjust the digital image so that the resultat image is more suitable tha the origial image for a specific applicatio [5, 8, ad 9]. There are may image ehacemet techiques. They ca be categorized ito two geeral categories. The first category is based o the direct maipulatio of pixels i a image, for istace: image egative, low pass filter (smoothig), ad high pass filter (sharpeig). Secod category is based o the positio maipulatio of pixels of a image, for istace image scalig.i the first category, the image processig fuctio i a spatial domai ca be expressed as g (x, y) = T (f(x, y)) Where, T is the trasformatio fuctio, f (x, y) is the pixel value of iput image, ad g(x, y) is the pixel value of the processed image [5, 8, 9]. The media, mea, high pass, low pass filterig techiques have bee applied for deoisig the differet images [5, 9]. A. Max Filter The max filter plays a key role i low level image processig ad visio. It is idetical to the mathematical morphological operatio: dilatio [9]. The brightest pixel gray level values are idetified by this filter. It has bee applied by may researchers to remove pepper oise. Though it removes the pepper oise it also removes the block pixel i the border [5]. This filter has ot yet applied to remove the speckle i the ultrasoud medical image. Hece it is proposed for speckle oise removal from the ultrasoud medical image. it is expressed as: f ( x, y) = max{ g( s, t)} ORIGINAL ULTRASOUND IMAGE OF FETUS ) Salt ad Pepper Noise This type of oise is also caused by errors i data trasmissio ad is a special case of data dropout oise whe i some cases sigle, sigle pixels are set alteratively to zero or to the maximum value, givig the image a salt ad pepper like appearace [6]. Uaffected pixels always remai uchaged. The oise is usually quatified by the percetage of pixels which are corrupted. It is foud i mammogram images [7]. It probability distributio of fuctio is i [8]. V. SPATIAL FILTER Figure. origial image 9

4 Vol. 7, No., February 00 MEDIAN FILTER SALT & PEPPER NOISE Figure. Media filter output. It reduces the itesity variatio betwee adjacet pixels. Implemetatio of this method for smoothig images is easy ad also reducig the amout of itesity variatio betwee oe pixel ad the ext. The result of this filter is the max selectio processig i the sub image area. B. Mi Filter The mi filter plays a sigificat role i image processig ad visio. It is equivalet to mathematical morphological operatio: erosio [9]. It recogizes the darkest pixels gray value ad retais it by performig mi operatio. This filter was proposed for removig the salt oise from the image by researchers. Salt oise has very high values i images. The operatio of this filter ca be expressed as: Figure 3. Salp ad Pepper Noice output WIENER FILTER f ( x, y) = mi{ g( s, t)} It removes oise better tha max filter but it removed some white poits aroud the border of the regio of the iterest [5].I this filter each output pixel value ca be calculated by selectig miimum gray level value of the chose widow. C. Stadard Deviatio Filter Normally the iterpretatios of the images are quite difficult, sice the backscatter causes the uwated oise. The stadard deviatio was proposed to remove the oise i radar satellite images []. This filter has ot proposed to remove the speckle oise from the ultrasoud medical images to the best of our kowledge. The stadard deviatio filter [5] calculates the stadard deviatio for each group of pixels i the sub image, ad assigs this pixel i the output image. By usig a stadard deviatio filter, we are able to recogize some patters. It is expressed as f ( x, y) = x x ( ) rc r= c= Figure 4. filtered image Where, x is the total umber of pixels i a sub image, the idices of the sub image, is the value of the pixel at row r ad colum c i the sub image ad x is the mea of pixel values i the widow. It measures of heterogeeity i the sub image at cetered over each pixel. Stadard deviatio filter is applied to detect the chages i sub images [5]. A small mask was used for the filter i order to obtai sharp edges. A size of 3x3 pixels was supposed to be sufficiet. The filter geerates a ew image based o the value of the stadard deviatio. 9

5 Vol. 7, No., February 00 Statistical VI. STATICAL MEASUREMENT Table; formulas applies for Statistical Measuremet Measuremet MSE RMSE SNR PSNR S.No Formula ( f ( i, j) F( i, j)) MSE = MN ( f ( i, j) F( i, j)) RMSE = MN SNR = 0 log 0 PSN R = 0 log 0 e 55 RM SE VII. COMPUTATIONAL RESULT Table; Comparative aalysis. FILTERING METHOD RMSE SNR PSNR. Media filter Variace filter Correlatio filter Mea filter VIII. CONCLUSION The performace of oise removig algorithms is measured usig quatitative performace measures such as PSNR, SNR, ad RMSE as well as i term of visual quality of the images. May of the methods fail to remove speckle oise preset i the ultrasoud medical image, sice the iformatio about the variace of the oise may ot be idetified by the methods. Performace of all algorithms is tested with ultrasoud image regard to fetus. REFERENCES [] Yaog Zhu, Stuart Williams, Reyer Zwiggelaar, Computer Techology i Detectio ad Stagig of Prostate Carcioma: A review, Medical image Aalysis 0(006),pp [] J. G. Abbott ad F. L. Thurstoe, "Acoustic speckle: Theory ad experimetal aalysis," UltrasoicImag., vol., pp , 979. [3] D Hykes, W.Hedrick ad D.Starchma, Ultrasoud Physics ad Istrumetatio, Churchill New,985 [4] R.C. Gozalez ad R.E. Woods: 'Digital Image Processig', Addiso- Wesley Publishig compay, 99. [5] ] Image Processig Fudametals Statistics, Sigal to Noise Ratio, 00. [6] R.C. Gozalez ad R.E. Woods: 'Digital Image Processig', Addiso- Wesley Publishig compay, 99. [7] A.K. Jai, fudametal of digital image processig. Eglewood cliffs, NJ Pretice-Hall, 989. [8] ] Prostate Carcioma: A review, Medical image Aalysis 0(006),pp 78-99,006. [9] H. GUO, J E Odegard, M.Lag, R.A.Gopiath,I.W.Selesick, ad C.S. Burrus, Wavelet based Speckle reductio with applicatio to SAR based ATD/R, First It I Cof. o image processig, vol., pp ,Nov 994. [0] Z. Wag ad D. Hag, Progressive Switchig Media Filter for the Removal of Impulse Noise from Highly Corrupted Images," IEEE Tras. o Circuits ad Systems-II: Aalog ad Digital Sigal processig, vol. 46, o., pp. AUTHORS PROFILE V..Rawat received her B.E i Electrical Egieerig from Rajiv Gadhi Techological uiversity,bhopal ad Master Degree i Istrumetatio from Devi Ahiliyabai Uiversity,Idore ad her field of iterest is Istrumetatio,Biomedical Egieerig ad Image proceesig. She is a member of Biomedical society of Idia. Alok jai received his B.E i Electroics & istrumetatiol Egieerig from Samrat Ashok Techological istitute,vidisha i 988, ad Master Degree i Computer Sciece from Roorkee i 99,.he obtaied his Ph.D degree from Thapar istitute of egieerig ad techology,patiala i 006. His field of iterest is sigal processig,filter baks,powerelectroics. Vibhakar shrimali is received his B.E. (Electroics & Comm. Egg.) i 988 from MBM Egg.College Jodhpur (Raj), M.E. (Electroics & Comm.Egg.) i 003 from Delhi College of Egieerig (Delhi) ad Ph.D. from IIT, Roorkee i 009. His field of iterest is Electroics & Commuicatio, Medical Electroics, Digital Image Processig ad Rehabilitatio Egieerig. 93

CHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER

95 CHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER 5.1 GENERAL Ru-legth codig is a lossless image compressio techique, which produces modest compressio ratios. Oe way of icreasig the compressio ratio of a ru-legth