Algorithms 2009, 2, ; doi: /a OPEN ACCESS

Size: px
Start display at page:

Download "Algorithms 2009, 2, ; doi: /a OPEN ACCESS"

Transcription

1 Algorthms 009,, ; do:0.3390/a04350 OPEN ACCESS algorthms ISSN Artcle CADrx for GBM Bran Tumors: Predctng Treatment Response from Changes n Dffuson-Weghted MRI Jng Huo, *, Kazunor Okada, Hyun J. Km, Whtney B. Pope, Jonathan G. Goldn, Jeffrey R. Alger and Matthew S. Brown UCLA Department of Radologcal Scences, 94 Westwood Blvd., Sute 650, Los Angeles, CA 9004, USA; E-Mal: mbrown@mednet.ucla.edu (M.S.B.) San Francsco State Unversty Computer Scence Department, Thornton Hall 9, 600 Holloway Avenue, San Francsco, CA , USA; E-Mal: kazokada@sfsu.edu (K.O.) * Author to whom correspondence should be addressed; E-Mal: jhuo@mednet.ucla.edu. Receved: August 009; n revsed form: September 009 / Accepted: 3 November 009 / Publshed: 6 November 009 Abstract: The goal of ths study was to develop a computer-aded therapeutc response (CADrx) system for early predcton of drug treatment response for globlastoma multforme (GBM) bran tumors wth dffuson weghted (DW) MR mages. In conventonal Macdonald assessment, tumor response s assessed nne weeks or more post-treatment. However, we wll nvestgate the ablty of DW-MRI to assess response earler, at fve weeks post treatment. The apparent dffuson coeffcent (ADC) map, calculated from DW mages, has been shown to reveal changes n the tumor s mcroenvronment precedng morphologc tumor changes. ADC values n treated bran tumors could theoretcally both ncrease due to the cell kll (and thus reduced cell densty) and decrease due to nhbton of edema. In ths study, we nvestgated the effectveness of features that quantfy changes from pre- and post-treatment tumor ADC hstograms to detect treatment response. There are three parts to ths study: frst, tumor regons were segmented on Tw contrast enhanced mages by Otsu s thresholdng method, and mapped from Tw mages onto ADC mages by a 3D regon of nterest (ROI) mappng tool usng DICOM header nformaton; second, ADC hstograms of the tumor regon were extracted from both pre- and fve weeks post-treatment scans, and ftted by a two-component Gaussan mxture model (GMM). The GMM features as well as standard hstogram-based features were extracted. Fnally, supervsed machne learnng technques were appled for classfcaton of responders or non-responders. The approach was evaluated wth a dataset of 85 patents wth GBM under chemotherapy, n whch 39 responded and 46 dd not,

2 Algorthms 009, 35 based on tumor volume reducton. We compared adaboost, random forest and support vector machne classfcaton algorthms, usng ten-fold cross valdaton, resultng n the best accuracy of 69.4% and the correspondng area under the curve (Az) of Keywords: CADx; DW-MRI; bomarker; adaboost; random forest; support vector machne. Introducton Computer aded dagnoss (CADx) can be defned as a dagnoss that s made by a radologst who uses the output from a computerzed analyss of medcal mages as a second opnon n both detectng lesons and makng dagnostc decsons []. One am of the typcal CADx system s to extract and analyze the characterstcs of bengn and malgnant lesons n an objectve manner to ad the radologst. Here, the dagnostc decson relates to treatment response and early classfcaton of drug responders versus non-responders, and we name our proposed system as computer-aded therapeutc response (CADrx) system. Globlastoma multforme (GBM) s the most aggressve and lethal prmary bran tumor n human. Ant-angogeness drugs are ncreasngly beng explored n clncal trals as therapeutc optons. In a phase II n vvo clncal tral, the conventonal way to assess treatment response s the tumor sze change after chemotherapy or radotherapy based on Macdonald crtera and evaluated on T-weghted contrast enhanced (TwCE) MR mages. However, effcacy can only be evaluated at least 8 0 weeks after treatment. Dffuson weghted magnetc resonance magng (DW-MRI) has the potental to work as a surrogate bomarker to reveal changes n the tumor mcroenvronment that precede morphologc tumor changes []. DW-MRI depends on the mcroscopc moblty of water. Ths moblty, classcally called Brownan moton, s due to thermal agtaton and s hghly nfluenced by the cellular envronment of water. Because water dffuson s strongly affected by molecular vscosty and membrane permeablty between ntra- and extracellular compartments, DW-MRI can be used to characterze hghly cellular regons of tumors versus acellular regons. Treatment response detecton can be manfested as a change n tumor cellularty, whch may precede tumor sze changes. Thus, fndngs on DW-MRI could be an early sgn of bologc changes. [3] The purpose of ths study s to use apparent dffuson coeffcent (ADC), derved from DW-MR mages, for early predcton of the tumor volume change on a later scan. There are two man parts to ths computer-aded treatment response evaluaton system. Frst, a sem-automated segmentaton algorthm s appled to segment the GBM bran tumors on TwCE mages. Then, the tumor ROI s mapped onto derved ADC maps and the hstogram of tumor ADC values wll be extracted for automatc treatment response predcton. Computer-aded detecton and segmentaton of GBM bran tumors s a challengng problem and n Table we present a concse revew of the pror art n automatc tumor segmentaton. Fuzzy clusterng and knowledge-based analyss are popular methods explored by the early poneers [4-6]. Voxel-based classfcaton method usng statstcal pattern classfcaton technques are explored by others [7-5].

3 Algorthms 009, 35 Most of the studes above use multple MRI sequences (Tw, Tw, proton densty weghted, and Flar) for the automatc tumor and edema detecton and segmentaton. Lu et al. [6] developed an nteractve system adaptng the fuzzy connectedness usng multple MRI sequences. Dube et al. [7,8] used texture features and segmentaton by the weghted aggregaton (SWA) method for the GBM tumor segmentaton on TwCE mages whch s smlar to part of our study. In our study, we developed sem-automated method to segment tumors on TwCE mages; n addton, we mapped the tumor contours onto ADC maps. Table. Summary of related methods n bran tumor segmentaton. The type abbrevatons are NC: Nasopharyngeal carcnoma; MNG: Menngomas; MG - malgnant glomas; MS multple scleross. Authors Technque Type Image sequences # of tumors Lu et al. [6] Sem-automated fuzzy GBM Tw, Tw+c, Flar 5 clusterng Phlps et al. [4] Fuzzy clusterng GBM PD,Tw, Tw+c Clark et al. [5] Fuzzy clusterng and GBM PD,Tw, Tw+c 7 knowledge-based analyss Fletcher-Heath Fuzzy clusterng and Bran PD,Tw, Tw wth 4 et al. [6] knowledge-based analyss tumor no contrast Prastawa et al. [7] Learn dstrbuton of normal Bran Tw, Tw (wth or 3 tssues/outler detecton as tumors tumors wthout contrast) Kaus et al. [8] Adaptve template -moderate LGG/M Tw+c, sagttal 0 technque wth atlas pror G vew Lee et al. [] Condtonal random feld and Bran Tw, Tw+c, Tw 7 support vector machne tumors Ho et al. [9] 3D level set GBM Tw+c, Tw, Tw 3 Vntsk et al. [0] k-nearest neghbor MS and MG PD, Tw, Tw, magnetzaton transfer 9 Zhu & Yan et al. [] Hopfeld neural network Bran tumors NA Zhang et al. [3] Support vector machne NC Tw, Tw+c 9 Corso et al. [5] SWA-segmentaton by GBM Tw, Tw, Tw+c, 0 weghted aggregaton. Flar Dube et al. [7] SWA wth texture features GBM Tw+c NA Ne et al. [4] Spatal accuracy-weghted hdden Markov feld and EM to solve the problem of hgh and low resoluton problem Glomas Hgh:Tw, Tw+c Low:Tw, Flar 5 Computer-aded dagnoss (CADx) n GBM bran tumor s an actve research area, and many promsng MR methods have been developed for detectng and characterzng cancer, ts treatments and adverse effects, e.g. T-weghted MR, T-weghted MR, MR spectroscopy, perfuson-weghted MR, and dffuson-weghted MR. In our study, we focused on T-weghted and DW-MRI. Tumor sze

4 Algorthms 009, 353 change on Tw mages s the only magng bomarker that s accepted by the FDA as a surrogate endpont of clncal outcome after chemotherapy and radotherapy for phase III trals [9]. Dffuson MRI has been explored as early detecton of human GBM bran tumor treatment response early therapeutc responses before the tumor sze changes. Table presents a revew of the recent studes that used DWI for GBM early predcton of treatment response. Ross et al reported ADC value ncrease sgnfcantly n effectve therapeutc nterventon n pre-clncal studes and presented two patents to support ths hypothess n a prelmnary clncal study [,0]. Mardor et al. [] appled both low and hgh b-value and used mean ADC and dffuson ndex for treatment response evaluaton. Moffat et al calculated voxel-by-voxel tumor ADC value changes over tme and dsplayed t as a functonal dffuson map for correlaton wth clncal response [,3]. They reported that the number of voxels wth ncreased ADC s related to treatment effcacy. Our prevous work [4] showed promsng results for usng ADC hstogram analyss, and we explored a more sophstcated classfer and desgned experments to show the advantages of the two-component hstogram modelng. Table. Summary of related methods n GBM tumor treatment response usng DWI. Authors # Of Patents Chenevert et al. [0] Ross et al. [] Mardor et al. [] 0 Moffat et al. [] 0 Hamstra et al. [3] 34 Machne learnng and statstcal pattern recognton have great contrbutons to the bomedcal communty because they can mprove the senstvty and/or specfcty of detecton and dagnoss of dsease, whle at the same tme ncreasng objectvty of the decson-makng process [6]. The need for machne learnng s perhaps greater than ever gven the dramatc ncrease n medcal data beng collected, new detecton, and dagnostc modaltes beng developed as well as the complexty of the data types and mportance of multmodal analyss. In all of these cases, machne learnng can provde new tools for nterpretng the hgh-dmensonal and complex datasets wth whch the clncan s confronted [6]. In our study, we explored three dfferent classfcaton methods: AdaBoost, random forest, and support vector machne. The AdaBoost algorthm, ntroduced by Freund and Schapre [7], s an teratve algorthm that can boost weak classfers nto a strong classfer and mprove the fnal accuracy. In each teraton, a feature s workng as a weak classfer and the best feature s selected to mnmze the average tranng error. Afterwards, the weghts on tranng samples are redstrbuted n such a way that the weght of accurately classfed samples wll be reduced whle the weght of ll classfed samples s rased. Therefore, AdaBoost focuses on the most dffcult ones [8]. The fnal classfer aggregates the selected weak classfer from each teraton, and the weght for each weak classfer depends on ts error rate. However, AdaBoost can be senstve to nose and may ntroduce the overfttng problem. Random forests (RF) s a classfer that combnes many decson trees [9]. Each tree depends on values of a random vector sampled ndependently and wth equal dstrbuton. Each tree casts a unt vote for the most popular case at nput, and random forests outputs the class that s the mode of the

5 Algorthms 009, 354 classes output by ndvdual trees. Breman suggests the generalzaton error for forests converges to a lmt as the number of trees n the forest becomes large [30]. The error of a forest of tree classfers depends on the strength of the ndvdual trees n the forest and the correlaton between them. Usng a random selecton of features to splt each node yelds error rates that compare favorably to Adaboost but are more robust wth respect to nose. Support vector machnes (SVMs) are a set of related supervsed learnng methods used for classfcaton and regresson [3,3]. Vewng nput data as two sets of vectors n an n-dmensonal space, an SVM wll construct a separatng hyperplane n that space, one whch maxmzes the margn between the two data sets. To calculate the margn, two parallel hyperplanes are constructed, one on each sde of the separatng hyperplane, whch are "pushed up aganst" the two data sets. Intutvely, a good separaton s acheved by the hyperplane that has the largest dstance to the neghborng data ponts of both classes, snce n general the larger the margn the lower the generalzaton error of the classfer [33]. SVM have been reported to work well for pharmaceutcal data analyss [34]. There are two man challenges n ths work. One challenge s the two competng effects n ADC changes after treatment. In general, water movement nsde cells s more restrcted than outsde. Thus, ncreased cell densty tends to lower ADC values, whereas ncreased edema (more ntersttal water) results n hgher ADC values. Therefore, theoretcally, ADC values n treated bran tumors could not only ncrease due to the cell kll (and thus reduced cell densty), but also decrease due to nhbton of edema. None of the lsted studes above have specfed the separate effects. Thus, we appled a twocomponent model to ft the tumor ADC hstogram [5]. The other challenge s that t s dffcult to drectly dentfy GBM bran tumors on ADC maps. We developed a sem-automated framework to acheve that goal. There are several contrbutons n ths work. Frst, we developed a computer-aded method to semautomatcally dentfy tumors on ADC maps. Second, we explored the changes of dfferent statstcal features of the whole tumor ADC hstogram. Moreover, we appled a two-component Gaussan mxture modelng to ft the tumor ADC hstogram to overcome the two competng effects. Next, we used earth mover s dstance (EMD) to drectly measure the dstance between the pre- and posttreatment tumor ADC hstograms. Fnally, we ntroduced machne learnng technque to do feature selecton and classfcaton to classfy responders and non-responders. Ths paper s organzed as follows: Secton descrbes the mage acquston and patent group, Secton 3 descrbes the sem-automated dentfcaton of GBM tumors on ADC maps, and Secton 4 descrbes the hstogram feature extracton and classfcaton. The Result Secton reports the performance of the tumor mappng on ADC maps, and the results of our comparatve study for three dfferent classfers. The fnal secton offers a dscusson of the expermental results as well as the future work.. Image Acquston.. Patent cohort A total of 85 patents wth GBM treated by ant-angogeness drugs were ncluded n our prelmnary study from our research database. Images n ths database were acqured as part of

6 Algorthms 009, 355 multcenter GBM treatment tral. Tumors were dagnosed by board-certfed radologsts as responders or non-responders to drugs based on Macdonald crtera from follow-up scans (8 0 weeks after baselne). The Macdonald crtera defne tumor response by use of tumor sze change, sterods, and neurologcal functons. There are four response categores: complete response (CR): dsappearance of enhancng tumors, off sterods, and neurologcally stable or mproved. Partal response (PR): >50% reducton n sze of enhancng tumor, sterods stable or reduced, neurologcally stable or mproved. Progressve dsease (PD): >5% ncrease n sze of enhancng tumor or any new tumor, or neurologcally worse, and sterods stable or ncreased. Stable dsease (SD): all other stuatons. In our study, we used tumor volume to evaluate tumor szes. More than 50% ncrease n volume s consdered to be the PD based on the neuro-radologst s suggeston [44]. Snce GBM s a rapdly progressng dsease, we classfed PD as non-responders and CR, PR and SD as responders. As a result, 39 were responders and 46 were non-responders. The DW-MRI scans were performed 5 7 weeks apart between baselne and follow-up scans. The patents n ths study were pooled from sx medcal stes scanned on 9 dfferent scanner models (GE/Semens) ncludng both.5 T and 3 T scanners. The magng protocol for TwCE s 3D volume n the axal plane wth flp angle-spoled gradent echo sequence (FSPGR) or magnetzaton-prepared rapd gradent-echo (MP-RAGE) sequence, 5 mm slce thckness, mm by mm pxel sze, and n-plane resoluton. The magng protocol for the DW mages s ether DWI or dfferent tensor magng (DTI), 700,000 s/mm for b-value, 3 30 for the number of dffuson senstzaton probng drectons, 5 7 mm slce thckness,.797mm by.797 mm pxel sze, and or 8 8 n-plane resoluton. We developed a qualty assurance technque to evaluate the consstency of ADC measurements from multple scanners and multple vsts by use of ROI analyss wth normal appearng whte matter [35]. Our study [35] showed that there s no sgnfcant dfference n ADC measurement among the dfferent scanner models used. For the between-vst reproducblty, ADC measurement was found to be reproducble wth consstent mage protocols... ADC map dervaton All ADC maps were calculated from DW-MR mages wth the same n-house software usng a twopont method as shown n the followng equaton: ADC = -ln[s(b)/s(0)]/b () wth b beng the dffuson senstvty factor rangng between 700 and,000 s/mm, S(0) and S(b) beng the mage ntensty when b = 0 and b = 700,000 s/mm. For DWI mages, we calculated ADC maps from DW mages by equaton (). For DTI, we calculated ADC for each orentaton and averaged them as the fnal ADC map. Fgure (b) shows an example of a derved bran ADC map.

7 Algorthms 009, 356 Fgure. (a) An example of the tumor segmented on a TwCE mage; (b) An example of the tumor ROI mapped from TwCE to ADC map; (c) An example of the tumor ADC hstogram ftted by two-component Gaussan mxtures. (a) (b) (c) 3. Sem-Automated Image Analyss on ADC Maps All patents were scanned by both TwCE MR mages and DW-MR mages. Snce t s dffcult to segment tumors accurately on derved ADC maps, we segmented tumors on TwCE mages frst, and then mapped the tumor contours onto the correspondng ADC maps. 3.. Tumor segmentaton on TwCE MR mages All tumors were segmented on TwCE mages va a sem-automated method usng the Otsu s thresholdng algorthm [36] and seeded regon growng [37]. Frst, radologsts drew a lne from nsde of the tumor to the outsde of the tumor on the approxmate center slce of the tumor. Then ntensty values along the lne were collected to form a bmodal hstogram, and the Ostu s thresholdng method was used to fnd the optmal thresholdng value. Afterwards, a 3D seeded regon growng was appled to obtan the segmentaton results n the whole volume. Threshold-based segmentaton methods are a standard approach to calculaton of tumor volume. The concept behnd the Otsu s thresholdng method [36] s to fnd the threshold that mnmzes the weghted wthn-class varaton σ as n equaton (), consderng the two-class segmentaton nto object and background: w t t) = P( ) = σ ( t) = q ( t) σ ( t) + q q ( σ ( t) = [ µ ( t)] w t = ( t) σ ( t) P( ) q ( t) µ ( t) = t P( ) q ( t) = wth q( t) as the class probablty, µ ( t ) as class mean, σ ( t ) as class varance, c =, as two dfferent c c () classes, t s varable for the ntensty value, and P(.) s the probablty densty functon. Gven an ntal class mean and varance, the algorthm wll do an exhaustve search by alterng the thresholdng value to fnd the optmal thresholdng value. Afterwards, seeded regon growng [37] usng the optmal thresholdng value was appled to get the tumor contours n the 3D volume. Radologsts revewed the results and made manual correctons when necessary. Fgure (A) shows an example of a segmented tumor on a TwCE mage.

8 Algorthms 009, Tumor mappng from TwCE mages to ADC maps It s dffcult for radologsts to drectly delneate the tumor contours on ADC maps, and the scanner-provded Tw mages and the derved ADC maps are not nherently co-regstered, because they have dfferent slce thckness, dfferent feld of vew (FOV), and dfferent mage resolutons. Therefore, a 3D ROI mappng tool was developed to map the tumor ROIs from TwCE mages onto ADC maps based on the scanner geometry. Compared to the co-regstraton technque, the mappng tool only transformed voxels wthn the tumor ROI rather than the whole mage volume; thus t was more computatonally effcent. However, the mappng tool could not correct for patent moton; thus a board-certfed radologst was requred to vsually check the mapped results and perform manual correctons when necessary. The mappng tool used an affne transformaton wth the parameters extracted from the DICOM header based on physcal locatons. Equaton 3 shows the way to calculate the 3D physcal locaton voxelwse.,j.k s the physcal voxel sze read from the tag pxel spacng and calculated from slce locaton ; X x,y,z, Y x,y,z s mage orentaton read from the tag mage orentaton whch specfes the orentaton of the mage frame rows and columns, Z x,y,z s the z-drecton orentaton calculated from X x,y,z, Y x,y,z, S x,y,z s read from the tag patent poston whch specfes the physcal locaton of the patent s anteror-left-upper corner;, j, k are voxel ndex; and P x,y,z are the calculated physcal locaton of the voxel n mllmeters. The transformaton matrces are calculated for both source and target ROI respectvely. For each voxel n the source ROI, the physcal locaton s frst calculated, and then the nverse operaton s performed to calculate the correspondng voxel coordnates of the target ROI. Fnally, radologsts vsually check the contours on ADC maps and manually correct the tumor contours on ADC when necessary. Fgure (B) shows an example of the mapped tumor ROI on the ADC map from the TwCE mage. Equaton 3. The physcal locaton calculaton of a voxel (,j,k). Px Py Pz = X x X y X z 0 Y x Y y Y y 0 j j j Z Z Z 0 y z x k k k S x S y j S k z 4. Feature Extracton and Classfcaton The dfferences between the features extracted from pre- and post-treatment tumor ADC hstograms are used as the nput to a tumor response classfer. 4.. Observatons Fgure shows examples of tumor ADC hstograms for both pre-and post-treatment wth responders and non-responders. The upper hstogram shows the ADC value dstrbuton before the drug treatment, whle the lower one shows the ADC value dstrbuton after the drug treatment. On the

9 Algorthms 009, 358 left s an example of a volumetrcally respondng tumor, whle on the rght s an example of a nonrespondng tumor. From the fgure, we observe that not only the locaton but also the shape of the responder s hstogram changes after treatment. The two Gaussan mxture components change as well. 4.. General hstogram features Dfferent statstcal features from tumor ADC hstograms were extracted. Accordng to clncal studes, the ADC value should change after treatment. In our data set, we observed that the hstograms exhbt change not only n locaton, but also n shape. Therefore, we ntroduced the extracton of dfferent ADC hstograms features and explored changes n ther pattern. The features are: mean, standard devaton, skewness, kurtoss, medan, IQR (nterquartle range), 5% percentle, and 75% percentle. Fgure. Examples of hstograms from two tumors and two tme ponts: (a), (c): example of a respondng tumor for pre- and post-treatment respectvely; (b), (d): example of a nonrespondng tumor for pre- and post-treatment respectvely Features from GMM Two-component Gaussan mxture modelng was appled to each tumor ADC hstogram and the two-component features were extracted. Due to the competng effects of tumor cell densty and edema, we made the assumpton that the obtaned tumor ADC hstogram was composed of two components relatng to tumor cellularty and edema. We assumed that the component wth lower peak s nfluenced by tumor cellularty, and the component wth hgher peak by edema effects. We used a two component

10 Algorthms 009, 359 GMM as shown n Equaton 3 to ft the ADC hstogram for both baselne and follow-up scans and appled EM algorthm to estmate GMM parameters, wth x as the ntensty values, a as the weght of the components, µ and σ as the Gaussan parameters. f ( x) = = αg ( x), wth G ( x µ ) σ ( x) = e σ π (4) The EM algorthm can be used to estmate the parameters of a parametrc mxture model dstrbuton: the weght of the components a, the Gaussan parameters µ, and σ. It s an teratve algorthm wth two steps: an expectaton step (E-step) and a maxmzaton step (M-step). In the E-step, wth the current parameter estmates of the mxture components, the algorthm calculates the expectaton values for the membershp varables of all data ponts. In the (m+) teraton, the expectaton s: p ( m+ ) n = α = ( m) α G ( m) ( m) G ( x) ( m) ( x) (5) In M-step, the algorthm maxmzes the expectaton value and updates the correspondng parameters. The followng solutons can be developed: µ = N ( m+ ) n= N n= x p n p ( m+ ) n ( m+ ) n ( σ ) = N ( m+ ) n= ( x µ ) ( m+ ) ( m+ ) n n N n= p ( m+ ) n p α N ( m+ ) = p n n n= (6) The features we obtaned from the GMM-EM were named as lower peak mean (LPM), lower peak varance (LPV), lower peak proporton (LPP), hgher peak mean (HPM), hgher peak varance (HPV) and hgher peak proporton (HPP). Fgure shows examples of tumor ADC hstograms ftted by GMM wth low ADC and hgh ADC curves overlad. Combnng GMM features wth the statstcal features, we obtaned 4-dmensonal feature vectors for both pre- and post-treatment tumor hstograms. Afterwards, we calculated the rate of change between the pre- and the post-treatment tumor hstogram. Therefore, we had a 4-dmensonal vector as the dfference feature vector Earth Mover s Dstance Fnally, we appled the earth mover's dstance (EMD) [38,39] as a metrc to drectly evaluate the dstance between the pre- and post-treatment tumor ADC hstograms. Informally, f the hstograms are nterpreted as two dfferent ways of plng up a certan amount of drt over the regon D, the EMD s the mnmum cost of turnng one ple nto the other; where the cost s assumed to be amount of drt moved tmes the dstance by whch s moved. The calculated EMD value was appended as the 5th element n the dfference feature vector. The calculated 5-dmensonal vector was the nput feature vector for classfcaton.

11 Algorthms 009, Classfcaton In ths study for classfcaton, we nvestgated three classfcaton technques wth dfferent characterstcs: AdaBoost, random forests (RF) and support vector machne (SVM). We employed three classfers to avod basng the results by selecton of a sngle classfcaton method. The reason we choose them s that the frst two classfers both nclude a feature selecton mechansm. By applyng these two classfcaton technques, we are seekng the best features that would separate responders from non-responders. SVM s reported to outperform several of the most frequently used machne learnng technques n structure actvty relatonshp (SAR) analyss. [34] In ths study, all classfers were mplemented n the open source data mnng software Weka [4]. Ther performance was evaluated usng 0-fold cross valdaton method. Three experments were performed. Frst, the conventonal method of usng mean ADC for treatment response classfcaton was appled []. Second, the AdaBoost, RF classfer, and SVM were appled to the dfference feature vectors of general statstcal hstogram features wthout GMM features, and results from the three classfers were compared. Fnally, the three classfers were appled usng all statstcal features ncludng the GMM features, and the results were compared, and the results of accuraces from dfferent classfcaton technques were compared wth conventonal method of ADC mean changes by the test of proporton. 5. Results 5.. Segmentaton Performance Fgure 3 shows four examples of segmentaton on TwCE mages and the mapped results on the derved ADC maps. For quanttatve evaluaton of the tumor segmentaton mappng results, we randomly selected 3 subjects baselne data. The 3 tumors are from an ADC mappng database, 0 of whch have dfferent mage resolutons between the TwCE and ADC mages n all three dmensons and of whch have exactly the same 3D mage resoluton n both modaltes. We calculated the overlap rato between the mapped ROI generated automatcally by the tool and an ROI corrected by a neuro-radologst. The overlap rato (OR) s defned by Equaton 6, where A and B are two tumor ROIs and sze(.) s the number of voxels n that ROI. * sze ( A B) ( sze( A) + sze( B)) The results are shown n Table 3 wth 0 out of 3 ROIs (64.5%) have an overlap rato over 90%. Table 3. Dstrbuton of overlap ratos. Overlap rato 00% 95~00% 90~95% 80~90% 60~80% 0~60% Number of patents (7)

12 Algorthms 009, 36 Fgure 3. (a)-(d) and ()-(l) show four examples of tumor segmentatons on TwCE mages; (e)-(h) and (m)-(p) show the correspondng mapped tumor contours on ADC maps. (a) (b) (c) (d) (e) (f) (g) (h) () (j) (k) (l) (m) (o) (p) (q)

13 Algorthms 009, Classfcaton Performance Usng the conventonal method of mean ADC change (subjects wth a mean ADC ncrease classfed as responders and those wth an ADC decrease as non-responders) [,0], the accuracy s 9.4% (5/85), wth a senstvty of 7.95% and a specfcty of 60.87% (see Table 4). Table 4. Performance of the conventonal mean ADC classfcaton method. Classfer Senstvty Specfcty Accuracy Az Mean ADC change 7.95% 60.87% 9.4% 0.33 The experment wth AdaBoost nvolved 0 learnng teratons. The RF classfer was composed of 0 trees, each of whch s constructed consderng fve random features. The SVM classfer used non-lnear polynomal kernels and normalzed all features. The results for the experment usng only the general hstogram features wthout GMM are shown n Table 5 wth senstvty, specfcty, accuracy and area under the ROC curve (Az). The ROC curves are shown n Fgure 4. The curve usng conventonal mean ADC was plotted by varyng the threshold of the mean ADC change used for the classfcaton, whle the curve usng the three ML technques were plotted by Weka. Weka plots the ROC curves by varyng the threshold on the probablty assgned to the postve class. Table 5. Performance comparson among three classfers wthout GMM features. Classfer Senstvty Specfcty Accuracy Az AdaBoost 45.45% 75% 63.53% * 0.6 Random forest 54.55% 73% 65.88% * 0.66 SVM 7.7% 9.3% 67.06% * 0.60 (*: All p-values <0.000 comparng wth accuracy of Table 4) Fgure 4. ROC curve for three classfers wthout GMM features. Wth GMM features added, the three classfers wth the same parameter setups were appled to the data. The results are shown n Table 6 wth senstvty, specfcty, accuracy and area under the curve (Az) of the ROC curve. The ROC curves are shown n Fgure 5.

14 Algorthms 009, 363 Table 6. Performance comparson among three classfers wth GMM features. Classfer Senstvty Specfcty Accuracy Az AdaBoost 39.39% 80.77% 64.7% * 0.60 Random forest 5.5% 80.77% 69.4% * 0.70 SVM 7.7% 9.3% 67.06% * 0.60 (*: All p-values <0.000 comparng wth accuracy of Table 4) Fgure 5. ROC curve for three classfers wth GMM features. 6. Dscusson Compared to usng only the mean ADC value, the quanttatve statstcal hstogram features and the proposed classfcaton system tremendously mproved the accuracy from 9.4% to 69.4% (Az ncreased from 0.33 to 0.70). The statstcal analyss ndcates that all three classfers are sgnfcantly dfferent from the conventonal mean ADC method wth our dataset. Compared to general statstcal hstogram features, the classfcaton wth GMM features usng random forest technque slghtly mproved the accuracy from 65.88% to 69.4%, whle adaboost and RF classfers generated the same accuracy no matter whether GMM features were ncluded. There s no sgnfcant dfference between the three machne-learned classfers. The conventonal mean ADC method performs worse than a random classfer (Az < 0.5). The reason s that conventonally researchers hypotheszed that mean ADC ncreases because the tumor cell densty decrease after an effectve treatment. Ths assumpton may not be vald for our dataset, because t nvolves n an ant-angogeness drug, whch suppresses the cancer cell growth wthout necessary kllng tumor cells (decreasng ther densty) at an early stage (5-7 weeks). Another possble reason s that n our dataset many of the GBM tumors are recurrent GBM tumors that are usually necrotc. The treatment tends to reduce necross and edema, whch wll dmnsh ADC. Essentally there are two competng processes at work: cell densty, edema and necross [5]. Another state-of-art study ncluded features that capture spatal nformaton n tumor heterogenety features. Functonal dffuson map (fdm) [,3] s a popular technque studyng the ADC value ncrease or decrease voxel-by-voxel. Moffat et al. appled fdm to 0 patents, classfed patents nto

15 Algorthms 009, 364 the three categores: PR, SD and PD, and reported 00% accuracy []. However, the threshold they used for classfcaton was determned from a sngle dataset of 0 patents used for both tranng and testng, whle n our experments, a cross valdaton analyss was performed. In Moffat et al s study, they explored the assessment of fractonated radaton therapy for dfferent types of bran tumors wth 0 patents scanned on the same scanner []. However, n our study, we focused on the GBM bran tumors treated by ant-angogeness drugs, whch suppress the blood supply for the tumor cells and may not drectly decrease the tumor cellularty. The dfference n accuracy may come from the dfferent mechansm of treatment. Addtonally, our dataset s from GBM drug trals across multple stes, thus our prelmnary study s an mportant contrbuton for explorng DWI as an early magng bomarker n a real pharmaceutcal drug tral. In future work, we wll extract texture feature to nclude spatal nformaton, and shape features wll be extracted as well. By ntroducng a new rcher feature set ndcatng more useful tumor nformaton, we am to nclude more nformaton about tumors and further mprove the performance of the classfcaton system. One lmtaton of ths study s that we classfed CR, PR and SD as responders for the ground truth to acheve a bnary classfcaton. Snce SD and PR may have dfferent patterns n terms of ther ADC hstogram change, a mult-category classfcaton system wll be explored n future work. Another lmtaton of the study s that we used the Macdonald crtera at the eghth or tenth week after treatment for determnng treatment response. In future work, tme-to-progresson and survval tme wll be a better endpont to classfy treatment response. Another lmtaton comes from the 3D ROI mappng tool. Ths tool s more computatonally effcent compared to the co-regstraton technques, but t cannot correct for patent moton. Therefore, n our study, a board-certfed radologst s vsually checked and edted all segmentaton results as needed. In the future, a more sophstcated regstraton method wth an mage smlarty measure may mprove the accuracy of the tumor contours on ADC maps, and consequently mprove the accuracy of the extracted features and the classfer performance. ADC values obtaned on pre-operatve MRI scans are reported to be of prognostc value n patents wth globlastoma [5,4]. The term "prognoss" refers to predctng the lkely outcome of treatment. ADC, reported to be nversely proportonal to tumor cellularty, s ganng nterest n predctng GBM tumor prognoss. Our proposed framework now uses changes n DW-MRI for early predcton of treatment response; however, the framework wth feature extracton and machne learnng technque could be generalzed to pre-treatment DW-MRI for prognoss predcton. In ths study, we developed a CADrx framework wth machne learnng technques to automatcally predct tumor treatment response before the sze change usng DW-MRI. In our prelmnary study, our major contrbutons are extractng statstcal ADC hstogram features, applyng GMM to model the ADC hstogram to nterpret the competng effects of cellular densty and edema, and applyng machne learnng technques usng all the extracted features. Cell densty and edema may be reflected n ADC values before sze changes are apparent on standard MRI sequences. Therefore, ADC holds promse as a bomarker, n determnng both whch tumors are more lkely to respond to treatment and whch tumors are actually respondng. In concluson, ths work shows that a CADrx system usng quanttatve ADC hstogram features and a machne-learned classfer has better performance n treatment response assessment over conventonal analyss usng only a mean ADC value. Ths wll have major mplcatons for clncal trals. Ths work has potental clncal sgnfcance for early treatment response assessment n GBM.

16 Algorthms 009, 365 References. Gger, M.L. Computer-aded dagnoss n medcal magng A new era n mage nterpretaton; Techncal Report; World Markets Research Centre: London, UK, 000; pp Ross, B.D.; Moffat, B.A.; Lawrence, T.S.; Mukherj, S.K.; Gebarsk, S.S.; Qunt, D.J.; Johnson, T.D.; Junck, L.; Robertson, P.L.; Muraszko, K.M.; Dong, Q.; Meyer, C.R.; Bland, P.H.; McConvlle, P.; Geng, H.; Rehemtulla, A.; Chenevert, T.L. Evaluaton of cancer therapy usng dffuson magnetc resonance magng. Mol. Cancer Ther. 003,, Padhan, A.R.; Lu, G.; Mu-Koh, D.; Chenevert, T.L.; Thoeny, H.C.; Takahara, T.; Dzk-Jurasz, A.; Ross, B.D.; Cauteren, M.V.; Collns, D.; Hammoud, D.A.; Rustn, G.J.S.; Taoul, B.; Choyke, P.L. Dffuson-weghted magnetc magng as a cancer bomarker: consensus and recommendatons. Neoplasa 009,, Phllps, W.E.; Velthuzen, R.P.; Phupanch, S.; Hall, L.O.; Clarke, L.P.; Slbger, M.L. Applcatons of fuzzy C-means segmentaton technque for tssue dfferentaton n MR mages of a hemorrhagc globlastoma multforme. J. Magn. Reson. Imagng 995, 3, Clark, M.C.; Hall, L.O.; Goldgof, D.B.; Velthuzen, R.; Murtagh, R.; Slbger, M.S. Automatc tumor segmentaton usng knowledge-based technques. IEEE Trans. Med. Imagng 998, 7, Fletcher-Heath, L.M.; Hall, L.O.; Goldgof, D.B.; Murtagh, R.F. Automatc segmentaton of non-enhancng bran tumors n magnetc resonance mages. Artf. Intell. Med. 00,, Prastawa, M.; Bulltt, E.; Moon, N.; Leemput, K.V.; Gerg, G. Automatc bran tumor segmentaton by subject specfc modfcaton of atlas prors. Acad. Radol. 003, 0, Kaus, M.; Warfeld, S.; Nabav, A.; Black, P.M.; Jolesz, F.A.; Kkns, R. Automated segmentaton of mr mages of bran tumors. Radology 00, 8, Ho, S.; Bulltt, E.; Gerg, G. Level set evoluton wth regon competton: Automatc 3-d segmentaton of bran tumors. In Proceedngs of Internatonal Conference on Pattern Recognton, Quebec, Canada, August, 00; pp Vntsk, S.; Gonzalez, C.F.; Knobler, R.; Andrews, D.; Iwanaga, T.; Curts, M. Fast tssue segmentaton based on a 4D feature map n characterzaton of ntracranal lesons fast tssue segmentaton based on a 4D feature map n characterzaton of ntracranal lesons. J. Magn. Reson. Imagng 999, 9, Lee, C.H.; Schmdt, M.; Murtha, A.; Bstrtz, A.; Sander, J.; Grener, R. Segmentng bran tumor wth condtonal random felds and support vector machnes. In Proceedngs of Workshop on Computer Vson for Bomedcal Image Applcatons at Internatonal Conference on Computer Vson, Bejng, Chna, October, 005; Vol. 3765, pp Zhu, Y.; Yan, H. Computerzed tumor boundary detecton usng a hopfeld neural network. LEEE Trans. Med. Imagng 997, 6, Zhang, J.; Ma, K.; Er, M.H.; Chong, V. Tumor segmentaton from magnetc resonance magng by learnng va one-class support vector machne. In Proceedngs of Internatonal Workshop on Advanced Image Technology, Sngapore, January, 004; pp. 07.

17 Algorthms 009, Nea, J.; Xue,.; Lu, T.; Young, G.S.; Setayesh, K.; Guo, L.; Wong, S.T.C. Automated bran tumor segmentaton usng spatal accuracy-weghted hdden Markov Random Feld. Comput. Med. Imagng Graph. 009, 33, Corso, J.J.; Sharon, E.; Dube, S.; El-Saden, S.; Snha, U.; Yulle, A. Effcent Multlevel Bran Tumor Segmentaton wth Integrated Bayesan Model Classfcaton. IEEE Trans. Med. Imagng 008, 7, Lu, J.; Udupa, J.; Odhner, D.; Hackney, D.; Moons, G. A system for bran tumor volume estmaton va mr magng and fuzzy connectedness. Comput. Med. Imagng Graph. 005, 9, Dube, S.; Corso, J.J.; Yulle, A.; Cloughesy, T.F.; El-Saden, S.; Snha, U. Herarchcal Segmentaton of Malgnant Glomas va Integrated Contextual Flter Response. Proc. SPIE 008, 694, 6943Y. 8. Dube, S.; Corso, J.J.; Cloughesy, T.F. ; El-Saden, S.; Yulle, A.; Snha, U. Automated MR mage processng and analyss of malgnant bran tumors: enablng technology for data mnng. In Data Mnng Systems Analyss and Optmzaton n Bomedcne; Amercan Insttute of Physcs Proceedngs: New York, NY, USA, 007; Vol. 953, pp US Food and Drug Admnstraton, Gudance for ndustry: clncal tral endponts for the approval of cancer drugs and bologcs. Federal Regster 007, 7, No Chenevert, T.L.; Stegman, L.D.; Taylor, J.M.; Robertson, P.L.; Greenberg, H.S.; Rehemtulla, A.; Ross, B.D. Dffuson magnetc resonance magng: an early surrogate marker of therapeutc effcacy n bran tumors. J. Natl. Cancer Inst. 000, 9, Mardor, Y.; Pfeffer, R.; Spegelmann, R.; Roth, Y.; Maer, S.E.; Nssm, O.; Berger, R.; Glcksman, A.; Baram, J.; Orensten, A.; Cohen, J.S.; Tchler, T. Early detecton of response to radaton therapy n patents wth bran malgnances usng conventonal and hgh b-value dffuson-weghted magnetc resonance magng. J. Cln. Oncol. 003,, Moffat, B.A.; Chenevert, T.L.; Meyer, C.R.; Mckeever, P.E.; Hall, D.E.; Hoff, B.A.; Johnson, T.D.; Rehemtulla, A.; Ross, B.D. The functonal dffuson map: a nonnvasve MRI bomarker for early stratfcaton of clncal bran tumor response. PANS 005, 0, Hamstra, D.A.; Chenevert, T.L.; Moffat, B.A.; Johnson, T.D.; Meyer, C.R.; Mukherj, S.K.; Qunt, D.J.; Gebarsk, S.S.; Fan, X.; Tsen, C.I.; Lawrence, T.S.; Junck, L.; Rehemtulla, A.; Ross, B.D. Evaluaton of the functonal dffuson map as an early bomarker of tme-to-progresson and overall survval n hgh-grade gloma. PNAS 005, 0, Huo, J.; Km, H.J.; Pope, W.B.; Okada, K.; Alger, J.R.; Wang, Y.; Goldn, J.G.; Brown, W.S. Hstogram-based classfcaton wth Gaussan mxture modelng for GBM tumor treatment response usng ADC map. Proc. SPIE 009, 760, 760Y. 5. Pope, W.B.; Km, H.J.; Huo, J.; Alger, J.R.; Brown, W.S.; Gjertson, D.; Sa, V.; Young, J.R.; Tekchandan, L.; Cloughesy, T.; Mschel, P.S.; La, A.; Nghemphu, P.; Rahmanuddn, S.; Goldn, J.G. Recurrent globlastoma multforme: ADC hstogram analyss predcts response to bevaczumab treatment. Radology 009, 5, Sajda, P. Machne learnng for detecton and dagnoss of dsease. Annu. Rev. Bomed. Eng. 006, 8,

18 Algorthms 009, Freund, Y.; Schapre, R.E. A short ntroducton to boostng. J. Jpn. Soc. For. Artf. Intell. 999, 4, Duda, R.O.; Hart, P.E.; Stork, D.H. Pattern classfcaton; Wley Interscence: Malden, MA, USA, Ho, T.K. Random decson forest. In Proceedngs of the 3rd Internatonal Conference on Document Analyss and Recognton Montreal, Canada, August, 995; pp Breman, L.: Random decson forest. Mach. Learn. 00, 45, Vapnk, V. Estmaton of Dependences Based on Emprcal Data; Nauka: Moscow, Russa, Bshop, C. Neural Networks for Pattern Recognton; Clarendon Press: Oxford, UK, (accessed November 0, 009). 34. Burbdge, R.; Trotter, M.; Buxton, B.; Holden, S. Drug desgn by machne learnng: support vector machnes for pharmaceutcal data analyss. Comput. And. Chem 00, 6, Huo, J.; Alger, J.R.; Km, H.J.; Pope, W.B.; Okada, K.; Goldn, J.G.; Brown, M.S. Betweenscanner varaton n normal whte matter ADC n the settng of a mult-center clncal tral. Ismrm 009, (n press). 36. Otsu N. A threshold selecton method from gray level hstograms. IEEE Trans. Syst. Man. Cybern. 979, 9, Adams, R.; Bschof, L. Seeded regon growng. IEEE Trans. Syst. Man. Cybern. Int. 994, 6, Rubner, Y.; Tomas, C.; Gubas, L.J. A metrc for dstrbutons wth applcatons to mage databases. In Proceedngs of ICCV, Bombay, Inda, January, 998; pp Lng, H.; Okada, K. An effcent Earth mover's dstance algorthm for robust hstogram comparson. IEEE Trans. Patt. Anal. Mach. Intell. 007, 9, Wtten, I.H.; Frank, E. Data Mnng: Practcal Machne Learnng Tools and Technques; Morgan Kaufmann: San Francsco, CA, USA, Yamasak, F.; Sugyama, K.; Ohtak, M.; Takeshma, Y.; Abed, N.; Akyamad, Y.; Takabad, J.; Amatyac, V.J.; Satoa, T.; Kajwaraa, Y.; Hanayaa, R.; Kursua, K. Globlastoma treated wth postoperatve rado-chemotherapy: Prognostc value of apparent dffuson coeffcent at MR magng. Eur.J. Aol. 009, (n press). 43. Marzban, C. The ROC curve and the area under t as a performance measure. Weather Forecast. 004, 9, Huhn, S.L.; Mohapatra, G.; Bollen, A.; Lamborn, K.; Prados, M.D.; Feuersten, B.G. Chromosomal abnormaltes n globlastoma multforme by comparatve genomc hybrdzaton: correlaton wth radaton treatment outcome. Cln. Cancer Res. 999, 5, by the authors; lcensee Molecular Dversty Preservaton Internatonal, Basel, Swtzerland. Ths artcle s an open-access artcle dstrbuted under the terms and condtons of the Creatve Commons Attrbuton lcense (

Study and Comparison of Various Techniques of Image Edge Detection

Study and Comparison of Various Techniques of Image Edge Detection Gureet Sngh et al Int. Journal of Engneerng Research Applcatons RESEARCH ARTICLE OPEN ACCESS Study Comparson of Varous Technques of Image Edge Detecton Gureet Sngh*, Er. Harnder sngh** *(Department of

More information

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures Usng the Perpendcular Dstance to the Nearest Fracture as a Proxy for Conventonal Fracture Spacng Measures Erc B. Nven and Clayton V. Deutsch Dscrete fracture network smulaton ams to reproduce dstrbutons

More information

*VALLIAPPAN Raman 1, PUTRA Sumari 2 and MANDAVA Rajeswari 3. George town, Penang 11800, Malaysia. George town, Penang 11800, Malaysia

*VALLIAPPAN Raman 1, PUTRA Sumari 2 and MANDAVA Rajeswari 3. George town, Penang 11800, Malaysia. George town, Penang 11800, Malaysia 38 A Theoretcal Methodology and Prototype Implementaton for Detecton Segmentaton Classfcaton of Dgtal Mammogram Tumor by Machne Learnng and Problem Solvng *VALLIAPPA Raman, PUTRA Sumar 2 and MADAVA Rajeswar

More information

AUTOMATED DETECTION OF HARD EXUDATES IN FUNDUS IMAGES USING IMPROVED OTSU THRESHOLDING AND SVM

AUTOMATED DETECTION OF HARD EXUDATES IN FUNDUS IMAGES USING IMPROVED OTSU THRESHOLDING AND SVM AUTOMATED DETECTION OF HARD EXUDATES IN FUNDUS IMAGES USING IMPROVED OTSU THRESHOLDING AND SVM Wewe Gao 1 and Jng Zuo 2 1 College of Mechancal Engneerng, Shangha Unversty of Engneerng Scence, Shangha,

More information

A New Machine Learning Algorithm for Breast and Pectoral Muscle Segmentation

A New Machine Learning Algorithm for Breast and Pectoral Muscle Segmentation Avalable onlne www.ejaet.com European Journal of Advances n Engneerng and Technology, 2015, 2(1): 21-29 Research Artcle ISSN: 2394-658X A New Machne Learnng Algorthm for Breast and Pectoral Muscle Segmentaton

More information

Research Article Statistical Analysis of Haralick Texture Features to Discriminate Lung Abnormalities

Research Article Statistical Analysis of Haralick Texture Features to Discriminate Lung Abnormalities Hndaw Publshng Corporaton Internatonal Journal of Bomedcal Imagng Volume 2015, Artcle ID 267807, 7 pages http://dx.do.org/10.1155/2015/267807 Research Artcle Statstcal Analyss of Haralck Texture Features

More information

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores Parameter Estmates of a Random Regresson Test Day Model for Frst Three actaton Somatc Cell Scores Z. u, F. Renhardt and R. Reents Unted Datasystems for Anmal Producton (VIT), Hedeweg 1, D-27280 Verden,

More information

Copy Number Variation Methods and Data

Copy Number Variation Methods and Data Copy Number Varaton Methods and Data Copy number varaton (CNV) Reference Sequence ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA Populaton ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA ACCTGCAATGAT TTGCAACGTTAGGCA

More information

FAST DETECTION OF MASSES IN MAMMOGRAMS WITH DIFFICULT CASE EXCLUSION

FAST DETECTION OF MASSES IN MAMMOGRAMS WITH DIFFICULT CASE EXCLUSION computng@tanet.edu.te.ua www.tanet.edu.te.ua/computng ISSN 727-6209 Internatonal Scentfc Journal of Computng FAST DETECTION OF MASSES IN MAMMOGRAMS WITH DIFFICULT CASE EXCLUSION Gábor Takács ), Béla Patak

More information

Available online at ScienceDirect. Procedia Computer Science 46 (2015 )

Available online at   ScienceDirect. Procedia Computer Science 46 (2015 ) Avalable onlne at www.scencedrect.com ScenceDrect Proceda Computer Scence 46 (215 ) 1762 1769 Internatonal Conference on Informaton and Communcaton Technologes (ICICT 214) Automatc Characterzaton of Bengn

More information

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS)

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS) Internatonal Assocaton of Scentfc Innovaton and Research (IASIR (An Assocaton Unfyng the Scences, Engneerng, and Appled Research Internatonal Journal of Emergng Technologes n Computatonal and Appled Scences

More information

Detection of Lung Cancer at Early Stage using Neural Network Techniques for Preventing Health Care

Detection of Lung Cancer at Early Stage using Neural Network Techniques for Preventing Health Care IJSRD - Internatonal Journal for Scentfc Research & Development Vol. 3, Issue 4, 15 ISSN (onlne): 31-613 Detecton of Lung Cancer at Early Stage usng Neural Network echnques for Preventng Health Care Megha

More information

Joint Modelling Approaches in diabetes research. Francisco Gude Clinical Epidemiology Unit, Hospital Clínico Universitario de Santiago

Joint Modelling Approaches in diabetes research. Francisco Gude Clinical Epidemiology Unit, Hospital Clínico Universitario de Santiago Jont Modellng Approaches n dabetes research Clncal Epdemology Unt, Hosptal Clínco Unverstaro de Santago Outlne 1 Dabetes 2 Our research 3 Some applcatons Dabetes melltus Is a serous lfe-long health condton

More information

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA Journal of Theoretcal and Appled Informaton Technology 2005 ongong JATIT & LLS ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA 1 SUNGMIN

More information

Semantics and image content integration for pulmonary nodule interpretation in thoracic computed tomography

Semantics and image content integration for pulmonary nodule interpretation in thoracic computed tomography Semantcs and mage content ntegraton for pulmonary nodule nterpretaton n thoracc computed tomography Danela S. Racu a, Ekarn Varutbangkul a, Jane G. Csneros a, Jacob D. Furst a, Davd S. Channn b, Samuel

More information

ARTICLE IN PRESS. computer methods and programs in biomedicine xxx (2007) xxx xxx. journal homepage:

ARTICLE IN PRESS. computer methods and programs in biomedicine xxx (2007) xxx xxx. journal homepage: computer methods and programs n bomedcne xxx (2007) xxx xxx journal homepage: www.ntl.elseverhealth.com/journals/cmpb Improvng bran tumor characterzaton on MRI by probablstc neural networks and non-lnear

More information

A Geometric Approach To Fully Automatic Chromosome Segmentation

A Geometric Approach To Fully Automatic Chromosome Segmentation A Geometrc Approach To Fully Automatc Chromosome Segmentaton Shervn Mnaee ECE Department New York Unversty Brooklyn, New York, USA shervn.mnaee@nyu.edu Mehran Fotouh Computer Engneerng Department Sharf

More information

Early Detection of Treatment Response for GBM Brain Tumor using ADC Map of DW-MRI

Early Detection of Treatment Response for GBM Brain Tumor using ADC Map of DW-MRI Early Detection of Treatment Response for GBM Brain Tumor using ADC Map of DW-MRI Jing Huo 1, Whitney Pope 1, Kazunori Okada 2, Jeffery Alger 3, Hyun Jung Kim 1, Yang Wang 1, Jonathan Goldin 1, and Matthew

More information

Journal of Engineering Science and Technology Review 11 (2) (2018) Research Article

Journal of Engineering Science and Technology Review 11 (2) (2018) Research Article Jestr Journal of Engneerng Scence and Technology Revew 11 (2) (2018) 8-12 Research Artcle Detecton Lung Cancer Usng Gray Level Co-Occurrence Matrx (GLCM) and Back Propagaton Neural Network Classfcaton

More information

Prognosis and Diagnosis of Breast Cancer Using Interactive Dashboard Through Big Data Analytics

Prognosis and Diagnosis of Breast Cancer Using Interactive Dashboard Through Big Data Analytics Prognoss and Dagnoss of Breast Cancer Usng Interactve Dashboard Through Bg Data Analytcs Gomath N, and Sandhya P 2 * Department of Computer Scence and Engneerng, Veltech Dr. RR & Dr. SR Unversty, Avad,

More information

Journal of Engineering Science and Technology Review 11 (2) (2018) Research Article

Journal of Engineering Science and Technology Review 11 (2) (2018) Research Article Jestr Journal of Engneerng Scence and Technology Revew () (08) 5 - Research Artcle Prognoss Evaluaton of Ovaran Granulosa Cell Tumor Based on Co-forest ntellgence Model Xn Lao Xn Zheng Juan Zou Mn Feng

More information

CLUSTERING is always popular in modern technology

CLUSTERING is always popular in modern technology Max-Entropy Feed-Forward Clusterng Neural Network Han Xao, Xaoyan Zhu arxv:1506.03623v1 [cs.lg] 11 Jun 2015 Abstract The outputs of non-lnear feed-forward neural network are postve, whch could be treated

More information

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of Appled Mathematcal Scences, Vol. 7, 2013, no. 41, 2047-2053 HIKARI Ltd, www.m-hkar.com Modelng Mult Layer Feed-forward Neural Network Model on the Influence of Hypertenson and Dabetes Melltus on Famly

More information

Prediction of Total Pressure Drop in Stenotic Coronary Arteries with Their Geometric Parameters

Prediction of Total Pressure Drop in Stenotic Coronary Arteries with Their Geometric Parameters Tenth Internatonal Conference on Computatonal Flud Dynamcs (ICCFD10), Barcelona, Span, July 9-13, 2018 ICCFD10-227 Predcton of Total Pressure Drop n Stenotc Coronary Arteres wth Ther Geometrc Parameters

More information

Optimal Planning of Charging Station for Phased Electric Vehicle *

Optimal Planning of Charging Station for Phased Electric Vehicle * Energy and Power Engneerng, 2013, 5, 1393-1397 do:10.4236/epe.2013.54b264 Publshed Onlne July 2013 (http://www.scrp.org/ournal/epe) Optmal Plannng of Chargng Staton for Phased Electrc Vehcle * Yang Gao,

More information

Functional and Molecular Imaging for Radiation Therapy Guidance

Functional and Molecular Imaging for Radiation Therapy Guidance Functonal and Molecular Imagng for Radaton Therapy Gudance Imagng 3D modelng Treatment plannng Pt setup and treatment delvery L Xng, T L, Y Yang, E. Schrebmann, B Thorndyke, D. Spelman 3D/4D CBCT Department

More information

Classification of Breast Tumor in Mammogram Images Using Unsupervised Feature Learning

Classification of Breast Tumor in Mammogram Images Using Unsupervised Feature Learning Amercan Journal of Appled Scences Orgnal Research Paper Classfcaton of Breast Tumor n Mammogram Images Usng Unsupervsed Feature Learnng 1 Adarus M. Ibrahm, 1 Baharum Baharudn, 1 Abas Md Sad and 2 P.N.

More information

INTEGRATIVE NETWORK ANALYSIS TO IDENTIFY ABERRANT PATHWAY NETWORKS IN OVARIAN CANCER

INTEGRATIVE NETWORK ANALYSIS TO IDENTIFY ABERRANT PATHWAY NETWORKS IN OVARIAN CANCER INTEGRATIVE NETWORK ANALYSIS TO IDENTIFY ABERRANT PATHWAY NETWORKS IN OVARIAN CANCER LI CHEN 1,2, JIANHUA XUAN 1,*, JINGHUA GU 1, YUE WANG 1, ZHEN ZHANG 2, TIAN LI WANG 2, IE MING SHIH 2 1The Bradley Department

More information

THE NATURAL HISTORY AND THE EFFECT OF PIVMECILLINAM IN LOWER URINARY TRACT INFECTION.

THE NATURAL HISTORY AND THE EFFECT OF PIVMECILLINAM IN LOWER URINARY TRACT INFECTION. MET9401 SE 10May 2000 Page 13 of 154 2 SYNOPSS MET9401 SE THE NATURAL HSTORY AND THE EFFECT OF PVMECLLNAM N LOWER URNARY TRACT NFECTON. L A study of the natural hstory and the treatment effect wth pvmecllnam

More information

Statistically Weighted Voting Analysis of Microarrays for Molecular Pattern Selection and Discovery Cancer Genotypes

Statistically Weighted Voting Analysis of Microarrays for Molecular Pattern Selection and Discovery Cancer Genotypes IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.6 No.2, December 26 73 Statstcally Weghted Votng Analyss of Mcroarrays for Molecular Pattern Selecton and Dscovery Cancer Genotypes

More information

Automatic Labelling and BI-RADS Characterisation of Mammogram Densities

Automatic Labelling and BI-RADS Characterisation of Mammogram Densities Lmted crculaton. For revew only. Automatc Labellng and BI-RADS Charactersaton of ammogram Denstes K. aras,. G. Lnguraru 3,. G. Ballester 4, S. Petroud,. Tsknaks and Sr. Brady Insttute of Computer Scence,

More information

Gene Selection Based on Mutual Information for the Classification of Multi-class Cancer

Gene Selection Based on Mutual Information for the Classification of Multi-class Cancer Gene Selecton Based on Mutual Informaton for the Classfcaton of Mult-class Cancer Sheng-Bo Guo,, Mchael R. Lyu 3, and Tat-Mng Lok 4 Department of Automaton, Unversty of Scence and Technology of Chna, Hefe,

More information

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and Ths artcle appeared n a journal publshed by Elsever. The attached copy s furnshed to the author for nternal non-commercal research and educaton use, ncludng for nstructon at the authors nsttuton and sharng

More information

IMPROVING THE EFFICIENCY OF BIOMARKER IDENTIFICATION USING BIOLOGICAL KNOWLEDGE

IMPROVING THE EFFICIENCY OF BIOMARKER IDENTIFICATION USING BIOLOGICAL KNOWLEDGE IMPROVING THE EFFICIENCY OF BIOMARKER IDENTIFICATION USING BIOLOGICAL KNOWLEDGE JOHN H. PHAN The Wallace H. Coulter Department of Bomedcal Engneerng, Georga Insttute of Technology, 313 Ferst Drve Atlanta,

More information

Survival Rate of Patients of Ovarian Cancer: Rough Set Approach

Survival Rate of Patients of Ovarian Cancer: Rough Set Approach Internatonal OEN ACCESS Journal Of Modern Engneerng esearch (IJME) Survval ate of atents of Ovaran Cancer: ough Set Approach Kamn Agrawal 1, ragat Jan 1 Department of Appled Mathematcs, IET, Indore, Inda

More information

Proceedings of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lisbon, Portugal, June 16-18, 2005 (pp )

Proceedings of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lisbon, Portugal, June 16-18, 2005 (pp ) Proceedngs of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lsbon, Portugal, June 6-8, 2005 (pp285-20) Novel Intellgent Edge Detector for Sonographcal Images Al Rafee *, Mohammad Hasan Morad **,

More information

Introduction ORIGINAL RESEARCH

Introduction ORIGINAL RESEARCH ORIGINAL RESEARCH Assessng the Statstcal Sgnfcance of the Acheved Classfcaton Error of Classfers Constructed usng Serum Peptde Profles, and a Prescrpton for Random Samplng Repeated Studes for Massve Hgh-Throughput

More information

Using Past Queries for Resource Selection in Distributed Information Retrieval

Using Past Queries for Resource Selection in Distributed Information Retrieval Purdue Unversty Purdue e-pubs Department of Computer Scence Techncal Reports Department of Computer Scence 2011 Usng Past Queres for Resource Selecton n Dstrbuted Informaton Retreval Sulleyman Cetntas

More information

AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION

AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION www.arpapress.com/volumes/vol8issue2/ijrras_8_2_02.pdf AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION I. Jule 1 & E. Krubakaran 2 1 Department

More information

A New Diagnosis Loseless Compression Method for Digital Mammography Based on Multiple Arbitrary Shape ROIs Coding Framework

A New Diagnosis Loseless Compression Method for Digital Mammography Based on Multiple Arbitrary Shape ROIs Coding Framework I.J.Modern Educaton and Computer Scence, 2011, 5, 33-39 Publshed Onlne August 2011 n MECS (http://www.mecs-press.org/) A New Dagnoss Loseless Compresson Method for Dgtal Mammography Based on Multple Arbtrary

More information

Towards Prediction of Radiation Pneumonitis Arising from Lung Cancer Patients Using Machine Learning Approaches

Towards Prediction of Radiation Pneumonitis Arising from Lung Cancer Patients Using Machine Learning Approaches Towards Predcton of Radaton Pneumonts Arsng from Lung Cancer Patents Usng Machne Learnng Approaches Jung Hun Oh, Adtya Apte, Rawan Al-Loz, Jeffrey Bradley, Issam El Naqa * Dvson of Bonformatcs and Outcomes

More information

National Polyp Study data: evidence for regression of adenomas

National Polyp Study data: evidence for regression of adenomas 5 Natonal Polyp Study data: evdence for regresson of adenomas 78 Chapter 5 Abstract Objectves The data of the Natonal Polyp Study, a large longtudnal study on survellance of adenoma patents, s used for

More information

Modeling the Survival of Retrospective Clinical Data from Prostate Cancer Patients in Komfo Anokye Teaching Hospital, Ghana

Modeling the Survival of Retrospective Clinical Data from Prostate Cancer Patients in Komfo Anokye Teaching Hospital, Ghana Internatonal Journal of Appled Scence and Technology Vol. 5, No. 6; December 2015 Modelng the Survval of Retrospectve Clncal Data from Prostate Cancer Patents n Komfo Anokye Teachng Hosptal, Ghana Asedu-Addo,

More information

Physical Model for the Evolution of the Genetic Code

Physical Model for the Evolution of the Genetic Code Physcal Model for the Evoluton of the Genetc Code Tatsuro Yamashta Osamu Narkyo Department of Physcs, Kyushu Unversty, Fukuoka 8-856, Japan Abstract We propose a physcal model to descrbe the mechansms

More information

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex 1 Sparse Representaton of HCP Grayordnate Data Reveals Novel Functonal Archtecture of Cerebral Cortex X Jang 1, Xang L 1, Jngle Lv 2,1, Tuo Zhang 2,1, Shu Zhang 1, Le Guo 2, Tanmng Lu 1* 1 Cortcal Archtecture

More information

Balanced Query Methods for Improving OCR-Based Retrieval

Balanced Query Methods for Improving OCR-Based Retrieval Balanced Query Methods for Improvng OCR-Based Retreval Kareem Darwsh Electrcal and Computer Engneerng Dept. Unversty of Maryland, College Park College Park, MD 20742 kareem@glue.umd.edu Douglas W. Oard

More information

Evaluation of the generalized gamma as a tool for treatment planning optimization

Evaluation of the generalized gamma as a tool for treatment planning optimization Internatonal Journal of Cancer Therapy and Oncology www.jcto.org Evaluaton of the generalzed gamma as a tool for treatment plannng optmzaton Emmanoul I Petrou 1,, Ganesh Narayanasamy 3, Eleftheros Lavdas

More information

Fast Algorithm for Vectorcardiogram and Interbeat Intervals Analysis: Application for Premature Ventricular Contractions Classification

Fast Algorithm for Vectorcardiogram and Interbeat Intervals Analysis: Application for Premature Ventricular Contractions Classification Fast Algorthm for Vectorcardogram and Interbeat Intervals Analyss: Applcaton for Premature Ventrcular Contractons Classfcaton Irena Jekova, Vessela Krasteva Centre of Bomedcal Engneerng Prof. Ivan Daskalov

More information

A Computer-aided System for Discriminating Normal from Cancerous Regions in IHC Liver Cancer Tissue Images Using K-means Clustering*

A Computer-aided System for Discriminating Normal from Cancerous Regions in IHC Liver Cancer Tissue Images Using K-means Clustering* A Computer-aded System for Dscrmnatng Normal from Cancerous Regons n IHC Lver Cancer Tssue Images Usng K-means Clusterng* R. M. CHEN 1, Y. J. WU, S. R. JHUANG, M. H. HSIEH, C. L. KUO, Y. L. MA Department

More information

Price linkages in value chains: methodology

Price linkages in value chains: methodology Prce lnkages n value chans: methodology Prof. Trond Bjorndal, CEMARE. Unversty of Portsmouth, UK. and Prof. José Fernández-Polanco Unversty of Cantabra, Span. FAO INFOSAMAK Tangers, Morocco 14 March 2012

More information

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16 310 Int'l Conf. Par. and Dst. Proc. Tech. and Appl. PDPTA'16 Akra Sasatan and Hrosh Ish Graduate School of Informaton and Telecommuncaton Engneerng, Toka Unversty, Mnato, Tokyo, Japan Abstract The end-to-end

More information

EXAMINATION OF THE DENSITY OF SEMEN AND ANALYSIS OF SPERM CELL MOVEMENT. 1. INTRODUCTION

EXAMINATION OF THE DENSITY OF SEMEN AND ANALYSIS OF SPERM CELL MOVEMENT. 1. INTRODUCTION JOURNAL OF MEDICAL INFORMATICS & TECHNOLOGIES Vol.3/00, ISSN 64-6037 Łukasz WITKOWSKI * mage enhancement, mage analyss, semen, sperm cell, cell moblty EXAMINATION OF THE DENSITY OF SEMEN AND ANALYSIS OF

More information

Estimation for Pavement Performance Curve based on Kyoto Model : A Case Study for Highway in the State of Sao Paulo

Estimation for Pavement Performance Curve based on Kyoto Model : A Case Study for Highway in the State of Sao Paulo Estmaton for Pavement Performance Curve based on Kyoto Model : A Case Study for Kazuya AOKI, PASCO CORPORATION, Yokohama, JAPAN, Emal : kakzo603@pasco.co.jp Octávo de Souza Campos, Publc Servces Regulatory

More information

Investigation of zinc oxide thin film by spectroscopic ellipsometry

Investigation of zinc oxide thin film by spectroscopic ellipsometry VNU Journal of Scence, Mathematcs - Physcs 24 (2008) 16-23 Investgaton of znc oxde thn flm by spectroscopc ellpsometry Nguyen Nang Dnh 1, Tran Quang Trung 2, Le Khac Bnh 2, Nguyen Dang Khoa 2, Vo Th Ma

More information

THIS IS AN OFFICIAL NH DHHS HEALTH ALERT

THIS IS AN OFFICIAL NH DHHS HEALTH ALERT THIS IS AN OFFICIAL NH DHHS HEALTH ALERT Dstrbuted by the NH Health Alert Network Health.Alert@dhhs.nh.gov August 26, 2016 1430 EDT (2:30 PM EDT) NH-HAN 20160826 Recommendatons for Accurate Dagnoss of

More information

Lymphoma Cancer Classification Using Genetic Programming with SNR Features

Lymphoma Cancer Classification Using Genetic Programming with SNR Features Lymphoma Cancer Classfcaton Usng Genetc Programmng wth SNR Features Jn-Hyuk Hong and Sung-Bae Cho Dept. of Computer Scence, Yonse Unversty, 134 Shnchon-dong, Sudaemoon-ku, Seoul 120-749, Korea hjnh@candy.yonse.ac.kr,

More information

Nonlinear Modeling Method Based on RBF Neural Network Trained by AFSA with Adaptive Adjustment

Nonlinear Modeling Method Based on RBF Neural Network Trained by AFSA with Adaptive Adjustment Advances n Engneerng Research (AER), volue 48 3rd Workshop on Advanced Research and Technology n Industry Applcatons (WARTIA 27) Nonlnear Modelng Method Based on RBF Neural Network Traned by AFSA wth Adaptve

More information

The effect of salvage therapy on survival in a longitudinal study with treatment by indication

The effect of salvage therapy on survival in a longitudinal study with treatment by indication Research Artcle Receved 28 October 2009, Accepted 8 June 2010 Publshed onlne 30 August 2010 n Wley Onlne Lbrary (wleyonlnelbrary.com) DOI: 10.1002/sm.4017 The effect of salvage therapy on survval n a longtudnal

More information

Biomarker Selection from Gene Expression Data for Tumour Categorization Using Bat Algorithm

Biomarker Selection from Gene Expression Data for Tumour Categorization Using Bat Algorithm Receved: March 20, 2017 401 Bomarker Selecton from Gene Expresson Data for Tumour Categorzaton Usng Bat Algorthm Gunavath Chellamuthu 1 *, Premalatha Kandasamy 2, Svasubramanan Kanagaraj 3 1 School of

More information

Research Article Statistical Segmentation of Regions of Interest on a Mammographic Image

Research Article Statistical Segmentation of Regions of Interest on a Mammographic Image Hndaw Publshng Corporaton EURASIP Journal on Advances n Sgnal Processng Volume 2007, Artcle ID 49482, 8 pages do:10.1155/2007/49482 Research Artcle Statstcal Segmentaton of Regons of Interest on a Mammographc

More information

Towards Automated Pose Invariant 3D Dental Biometrics

Towards Automated Pose Invariant 3D Dental Biometrics Towards Automated Pose Invarant 3D Dental Bometrcs Xn ZHONG 1, Depng YU 1, Kelvn W C FOONG, Terence SIM 3, Yoke San WONG 1 and Ho-lun CHENG 3 1. Mechancal Engneerng, Natonal Unversty of Sngapore, 117576,

More information

Subject-Adaptive Real-Time Sleep Stage Classification Based on Conditional Random Field

Subject-Adaptive Real-Time Sleep Stage Classification Based on Conditional Random Field Subject-Adaptve Real-Tme Sleep Stage Classfcaton Based on Condtonal Random Feld Gang Luo, PhD, Wanl Mn, PhD IBM TJ Watson Research Center, Hawthorne, NY {luog, wanlmn}@usbmcom Abstract Sleep stagng s the

More information

PERFORMANCE EVALUATION OF DIVERSIFIED SVM KERNEL FUNCTIONS FOR BREAST TUMOR EARLY PROGNOSIS

PERFORMANCE EVALUATION OF DIVERSIFIED SVM KERNEL FUNCTIONS FOR BREAST TUMOR EARLY PROGNOSIS AR Journal of Engneerng and Appled Scences 2006-2014 Asan Research ublshng etwork (AR). All rghts reserved. ERFORMACE EVALUAIO OF DIVERSIFIED SVM KEREL FUCIOS FOR BREAS UMOR EARLY ROGOSIS Khondker Jahd

More information

Recognition of ASL for Human-robot Interaction

Recognition of ASL for Human-robot Interaction 66 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.17 No.7, July 2017 Recognton of ASL for Human-robot Interacton Md. Al-Amn Bhuyan College of Computer Scences & Informaton Technology,

More information

The Effect of Fish Farmers Association on Technical Efficiency: An Application of Propensity Score Matching Analysis

The Effect of Fish Farmers Association on Technical Efficiency: An Application of Propensity Score Matching Analysis The Effect of Fsh Farmers Assocaton on Techncal Effcency: An Applcaton of Propensty Score Matchng Analyss Onumah E. E, Esslfe F. L, and Asumng-Brempong, S 15 th July, 2016 Background and Motvaton Outlne

More information

Feature Selection for Predicting Tumor Metastases in Microarray Experiments using Paired Design

Feature Selection for Predicting Tumor Metastases in Microarray Experiments using Paired Design Feature Selecton for Predctng Tumor Metastases n Mcroarray Experments usng Pared Desgn Qhua Tan 1,2, Mads Thomassen 1 and Torben A. Kruse 1 ORIGINAL RESEARCH 1 Department of Bochemstry, Pharmacology and

More information

econstor Make Your Publications Visible.

econstor Make Your Publications Visible. econstor Make Your Publcatons Vsble. A Servce of Wrtschaft Centre zbwlebnz-informatonszentrum Economcs Chang, Huan-Cheng; Chang, Pn-Hsang; Tseng, Sung-Chn; Chang, Ch- Chang; Lu, Yen-Chao Artcle A comparatve

More information

RENAL FUNCTION AND ACE INHIBITORS IN RENAL ARTERY STENOSISA/adbon et al. 651

RENAL FUNCTION AND ACE INHIBITORS IN RENAL ARTERY STENOSISA/adbon et al. 651 Downloaded from http://ahajournals.org by on January, 209 RENAL FUNCTION AND INHIBITORS IN RENAL ARTERY STENOSISA/adbon et al. 65 Downloaded from http://ahajournals.org by on January, 209 Patents and Methods

More information

Boosting for tumor classification with gene expression data. Seminar für Statistik, ETH Zürich, CH-8092, Switzerland

Boosting for tumor classification with gene expression data. Seminar für Statistik, ETH Zürich, CH-8092, Switzerland BIOINFORMATICS Vol. 19 no. 9 2003, pages 1061 1069 DOI: 10.1093/bonformatcs/btf867 Boostng for tumor classfcaton wth gene expresson data Marcel Dettlng and Peter Bühlmann Semnar für Statstk, ETH Zürch,

More information

An Approach to Discover Dependencies between Service Operations*

An Approach to Discover Dependencies between Service Operations* 36 JOURNAL OF SOFTWARE VOL. 3 NO. 9 DECEMBER 2008 An Approach to Dscover Dependences between Servce Operatons* Shuyng Yan Research Center for Grd and Servce Computng Insttute of Computng Technology Chnese

More information

INITIAL ANALYSIS OF AWS-OBSERVED TEMPERATURE

INITIAL ANALYSIS OF AWS-OBSERVED TEMPERATURE INITIAL ANALYSIS OF AWS-OBSERVED TEMPERATURE Wang Yng, Lu Xaonng, Ren Zhhua, Natonal Meteorologcal Informaton Center, Bejng, Chna Tel.:+86 684755, E-mal:cdcsjk@cma.gov.cn Abstract From, n Chna meteorologcal

More information

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 by TIANHONG ZHOU B.S., Chna Agrcultural Unversty, 2003 M.S., Chna Agrcultural Unversty, 2006 A THESIS submtted n partal fulfllment of the requrements

More information

Pattern Recognition for Robotic Fish Swimming Gaits Based on Artificial Lateral Line System and Subtractive Clustering Algorithms

Pattern Recognition for Robotic Fish Swimming Gaits Based on Artificial Lateral Line System and Subtractive Clustering Algorithms Sensors & Transducers, Vol. 18, Issue 11, November 14, pp. 7-16 Sensors & Transducers 14 by IFSA Publshng, S. L. http://www.sensorsportal.com Pattern Recognton for Robotc Fsh Swmmng Gats Based on Artfcal

More information

Reconstruction of gene regulatory network of colon cancer using information theoretic approach

Reconstruction of gene regulatory network of colon cancer using information theoretic approach Reconstructon of gene regulatory network of colon cancer usng nformaton theoretc approach Khald Raza #1, Rafat Parveen * # Department of Computer Scence Jama Mlla Islama (Central Unverst, New Delh-11005,

More information

The Influence of the Isomerization Reactions on the Soybean Oil Hydrogenation Process

The Influence of the Isomerization Reactions on the Soybean Oil Hydrogenation Process Unversty of Belgrade From the SelectedWorks of Zeljko D Cupc 2000 The Influence of the Isomerzaton Reactons on the Soybean Ol Hydrogenaton Process Zeljko D Cupc, Insttute of Chemstry, Technology and Metallurgy

More information

Research Article Segmentation of Bone with Region Based Active Contour Model in PD Weighted MR Images of Shoulder

Research Article Segmentation of Bone with Region Based Active Contour Model in PD Weighted MR Images of Shoulder Computatonal and Mathematcal Methods n Medcne Volume 2015, Artcle ID 754894, 13 pages http://dx.do.org/10.1155/2015/754894 Research Artcle Segmentaton of Bone wth Regon Based Actve Contour Model n PD Weghted

More information

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx Neuropsychologa xxx (200) xxx xxx Contents lsts avalable at ScenceDrect Neuropsychologa journal homepage: www.elsever.com/locate/neuropsychologa Storage and bndng of object features n vsual workng memory

More information

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis The Lmts of Indvdual Identfcaton from Sample Allele Frequences: Theory and Statstcal Analyss Peter M. Vsscher 1 *, Wllam G. Hll 2 1 Queensland Insttute of Medcal Research, Brsbane, Australa, 2 Insttute

More information

Improvement of Automatic Hemorrhages Detection Methods using Brightness Correction on Fundus Images

Improvement of Automatic Hemorrhages Detection Methods using Brightness Correction on Fundus Images Improvement of Automatc Hemorrhages Detecton Methods usng Brghtness Correcton on Fundus Images Yuj Hatanaka *a, Toshak Nakagawa *b, Yoshnor Hayash *c, Masakatsu Kakogawa *c, Akra Sawada *d, Kazuhde Kawase

More information

Maize Varieties Combination Model of Multi-factor. and Implement

Maize Varieties Combination Model of Multi-factor. and Implement Maze Varetes Combnaton Model of Mult-factor and Implement LIN YANG,XIAODONG ZHANG,SHAOMING LI Department of Geographc Informaton Scence Chna Agrcultural Unversty No. 17 Tsnghua East Road, Bejng 100083

More information

DETECTION AND CLASSIFICATION OF BRAIN TUMOR USING ML

DETECTION AND CLASSIFICATION OF BRAIN TUMOR USING ML DOI: http://dx.do.org/0.26483/arcs.v92.5807 Volume 9, No. 2, March-Aprl 208 Internatonal Journal of Advanced Research n Computer Scence RESEARCH PAPER Avalable Onlne at www.arcs.nfo ISSN No. 0976-5697

More information

Economic crisis and follow-up of the conditions that define metabolic syndrome in a cohort of Catalonia,

Economic crisis and follow-up of the conditions that define metabolic syndrome in a cohort of Catalonia, Economc crss and follow-up of the condtons that defne metabolc syndrome n a cohort of Catalona, 2005-2012 Laa Maynou 1,2,3, Joan Gl 4, Gabrel Coll-de-Tuero 5,2, Ton Mora 6, Carme Saurna 1,2, Anton Scras

More information

A GEOGRAPHICAL AND STATISTICAL ANALYSIS OF LEUKEMIA DEATHS RELATING TO NUCLEAR POWER PLANTS. Whitney Thompson, Sarah McGinnis, Darius McDaniel,

A GEOGRAPHICAL AND STATISTICAL ANALYSIS OF LEUKEMIA DEATHS RELATING TO NUCLEAR POWER PLANTS. Whitney Thompson, Sarah McGinnis, Darius McDaniel, A GEOGRAPHICAL AD STATISTICAL AALYSIS OF LEUKEMIA DEATHS RELATIG TO UCLEAR POWER PLATS Whtney Thompson, Sarah McGnns, Darus McDanel, Jean Sexton, Rebecca Pettt, Sarah Anderson, Monca Jackson ABSTRACT:

More information

Appendix F: The Grant Impact for SBIR Mills

Appendix F: The Grant Impact for SBIR Mills Appendx F: The Grant Impact for SBIR Mlls Asmallsubsetofthefrmsnmydataapplymorethanonce.Ofthe7,436applcant frms, 71% appled only once, and a further 14% appled twce. Wthn my data, seven companes each submtted

More information

Project title: Mathematical Models of Fish Populations in Marine Reserves

Project title: Mathematical Models of Fish Populations in Marine Reserves Applcaton for Fundng (Malaspna Research Fund) Date: November 0, 2005 Project ttle: Mathematcal Models of Fsh Populatons n Marne Reserves Dr. Lev V. Idels Unversty College Professor Mathematcs Department

More information

IDENTIFICATION AND DELINEATION OF QRS COMPLEXES IN ELECTROCARDIOGRAM USING FUZZY C-MEANS ALGORITHM

IDENTIFICATION AND DELINEATION OF QRS COMPLEXES IN ELECTROCARDIOGRAM USING FUZZY C-MEANS ALGORITHM IDENTIFICATION AND DELINEATION OF QRS COMPLEXES IN ELECTROCARDIOGRAM USING FUZZY C-MEANS ALGORITHM S.S. MEHTA 1, C.R.TRIVEDI 2, N.S. LINGAYAT 3 1 Electrcal Engneerng Department, J.N.V, Unversty, Jodhpur.

More information

Comparison of support vector machine based on genetic algorithm with logistic regression to diagnose obstructive sleep apnea

Comparison of support vector machine based on genetic algorithm with logistic regression to diagnose obstructive sleep apnea Orgnal Artcle Comparson of support vector machne based on genetc algorthm wth logstc regresson to dagnose obstructve sleep apnea Zohreh Manoochehr, Nader Salar 1, Mansour Rezae 1, Habbolah Khazae 2, Sara

More information

NeuroImage. Multimodal classification of Alzheimer's disease and mild cognitive impairment

NeuroImage. Multimodal classification of Alzheimer's disease and mild cognitive impairment NeuroImage 55 (2011) 856 867 Contents lsts avalable at ScenceDrect NeuroImage journal homepage: www.elsever.com/locate/ynmg Multmodal classfcaton of Alzhemer's dsease and mld cogntve mparment Daoqang Zhang

More information

Saeed Ghanbari, Seyyed Mohammad Taghi Ayatollahi*, Najaf Zare

Saeed Ghanbari, Seyyed Mohammad Taghi Ayatollahi*, Najaf Zare DOI:http://dx.do.org/10.7314/APJCP.2015.16.14.5655 and Anthracyclne- Breast Cancer Treatment and Survval n the Eastern Medterranean and Asa: a Meta-analyss RESEARCH ARTICLE Comparng Role of Two Chemotherapy

More information

Nonstandard Machine Learning Algorithms for Microarray Data Mining. Byoung-Tak Zhang

Nonstandard Machine Learning Algorithms for Microarray Data Mining. Byoung-Tak Zhang Nonstandard Machne Learnng Algorthms for Mcroarray Data Mnng Byoung-Tak Zhang Center for Bonformaton Technology (CBIT) & Bontellgence Laboratory School of Computer Scence and Engneerng Seoul Natonal Unversty

More information

INTRAUTERINE GROWTH RESTRICTION (IUGR) RISK DECISION BASED ON SUPPORT VECTOR MACHINES

INTRAUTERINE GROWTH RESTRICTION (IUGR) RISK DECISION BASED ON SUPPORT VECTOR MACHINES Mathematcal and Computatonal Applcatons, Vol. 15, No. 3, pp. 472-480, 2010. Assocaton for Scentfc Research INTRAUTERINE GROWTH RESTRICTION (IUGR) RISK DECISION BASED ON SUPPORT VECTOR MACHINES Zeynep Zengn

More information

An Introduction to Modern Measurement Theory

An Introduction to Modern Measurement Theory An Introducton to Modern Measurement Theory Ths tutoral was wrtten as an ntroducton to the bascs of tem response theory (IRT) modelng and ts applcatons to health outcomes measurement for the Natonal Cancer

More information

Computing and Using Reputations for Internet Ratings

Computing and Using Reputations for Internet Ratings Computng and Usng Reputatons for Internet Ratngs Mao Chen Department of Computer Scence Prnceton Unversty Prnceton, J 8 (69)-8-797 maoch@cs.prnceton.edu Jaswnder Pal Sngh Department of Computer Scence

More information

Integration of sensory information within touch and across modalities

Integration of sensory information within touch and across modalities Integraton of sensory nformaton wthn touch and across modaltes Marc O. Ernst, Jean-Perre Brescan, Knut Drewng & Henrch H. Bülthoff Max Planck Insttute for Bologcal Cybernetcs 72076 Tübngen, Germany marc.ernst@tuebngen.mpg.de

More information

Detection of Cancer Metastasis Using a Novel Macroscopic Hyperspectral Method

Detection of Cancer Metastasis Using a Novel Macroscopic Hyperspectral Method Detecton of Cancer Metastass Usng a Novel Macroscopc Hyperspectral Method Hamed Akbar a, Luma V. Halg a, Hongzheng Zhang b, Dongsheng Wang b, Zhuo Georga Chen b, Baowe Fe a,c,d,* a Department of Radology

More information

Influence of concentration of sugar on mass transfer of pineapple slices during osmotic dehydration

Influence of concentration of sugar on mass transfer of pineapple slices during osmotic dehydration J. Bangladesh Agrl. Unv. 2(: 22 226, 24 ISSN 8-33 Influence of concentraton of sugar on mass transfer of pneapple slces durng osmotc dehydraton S. A. A. Khanom, M. M. Rahman 2 and M. B. Uddn 3* Vegetable

More information

An expressive three-mode principal components model for gender recognition

An expressive three-mode principal components model for gender recognition Journal of Vson (4) 4, 36-377 http://journalofvson.org/4/5// 36 An expressve three-mode prncpal components model for gender recognton James W. Davs Hu Gao Department of Computer and Informaton Scence,

More information

Insights in Genetics and Genomics

Insights in Genetics and Genomics Insghts n Genetcs and Genomcs Research Artcle Open Access New Score Tests for Equalty of Varances n the Applcaton of DNA Methylaton Data Analyss [Verson ] Welang Qu Xuan L Jarrett Morrow Dawn L DeMeo Scott

More information

CONSTRUCTION OF STOCHASTIC MODEL FOR TIME TO DENGUE VIRUS TRANSMISSION WITH EXPONENTIAL DISTRIBUTION

CONSTRUCTION OF STOCHASTIC MODEL FOR TIME TO DENGUE VIRUS TRANSMISSION WITH EXPONENTIAL DISTRIBUTION Internatonal Journal of Pure and Appled Mathematcal Scences. ISSN 97-988 Volume, Number (7), pp. 3- Research Inda Publcatons http://www.rpublcaton.com ONSTRUTION OF STOHASTI MODEL FOR TIME TO DENGUE VIRUS

More information

AUTOMATED CHARACTERIZATION OF ESOPHAGEAL AND SEVERELY INJURED VOICES BY MEANS OF ACOUSTIC PARAMETERS

AUTOMATED CHARACTERIZATION OF ESOPHAGEAL AND SEVERELY INJURED VOICES BY MEANS OF ACOUSTIC PARAMETERS AUTOMATED CHARACTERIZATIO OF ESOPHAGEAL AD SEVERELY IJURED VOICES BY MEAS OF ACOUSTIC PARAMETERS B. García, I. Ruz, A. Méndez, J. Vcente, and M. Mendezona Department of Telecommuncaton, Unversty of Deusto

More information