Detection of Brain Tumor Using Self Organizing Map ... - IRD India

Detection of Brain Tumor Using Self Organizing MapWith K-mean Algorithm1 Pradeep Singh Yadav & 2 Chandrahas Sahu1 SSITM (Bhilai) , 2 SSCET(Bhilai)Pradeepyadav_1987@rediffmail.com, Chandrahassscet@gmail.comAbstract – The part of the image that has the tumor hasmore intensity in that portion and we can make ourassumptions about the radius of the tumor in the image,these are the basic things considered in the algorithm. Firstof all some image enhancement and noise reductiontechniques are used to enhance the image quality, afterthat some morphological operations are applied to detectthe tumor in the image. The morphological operations arebasically applied on some assumptions about the size andshape of the tumor and in the end the tumor is mappedonto the original gray scale image with 255 intensity tomake visible the tumor in the image. The algorithm hasbeen tried on a number of different images from differentangles and has always given the correct desired result.Index Terms— object recognition, characteristic vector, edgetracing, feature detectionI. INTRODUCTIONThe purpose of our project is to identify a tumorfrom a given MRI scan of a brain using digital imageprocessing techniques. In this self organizing map isused for dividing the whole image into number ofcluster, than after K-mean algorithm is used forclustering of cluster.II. TUMORA tumor or tumor is the name for a neoplasm or asolid lesion formed by an abnormal growth of cells(termed neoplastic) which looks like a swelling .Tumoris not synonymous with cancer. A tumor can be benignpre-malignant or malignant, whereas cancer is bydefinition malignant.III. TYPES OF TUMOR:BENIGN TUMOR: A benign tumor is a tumor thatlacks all three of the malignant properties of a cancer.Thus, by definition, a benign tumor does not grow in anunlimited, aggressive manner, does not invadesurrounding tissues, and does not spread to non-adjacenttissues (metastasize). Common examples of benigntumors include moles and uterine fibroids.MALIGNAN : Malignancy (from the Latin roots mal- ="bad" and -ignis = "fire") is the tendency of a medicalcondition, especially tumors, to become progressivelyworse and to potentially result in death. It ischaracterized by the properties of anaplasia,invasiveness, and metastasis. Malignant is acorresponding adjectival medical term used to describe asevere and progressively worsening disease. The term ismost familiar as a description of cancer.PREMALIGNANT:A precancerous condition (orpremalignant condition) is a disease, syndrome, orfinding that, if left untreated, may lead to cancer. It is ageneralized state associated with a significantlyincreased risk of cancer.IV. MRIMagnetic resonance imaging (MRI), or nuclearmagnetic resonance imaging (NMRI), is primarily amedical imaging technique used in radiology tovisualize detailed internal structure and limited functionof the body. MRI provides much greater contrastbetween the different soft tissues of the body thancomputed tomography (CT) does, making it especiallyuseful in neurological (brain), musculoskeletal,cardiovascular, and oncological (cancer) imaging.Unlike CT, MRI uses no ionizing radiation. Rather, ituses a powerful magnetic field to align the nuclearmagnetization of (usually) hydrogen atoms in water inthe body. Radio frequency (RF) fields are used tosystematically alter the alignment of this magnetization.This causes the hydrogen nuclei to produce a rotatingmagnetic field detectable by the scanner. This signal canbe manipulated by additional magnetic fields to build upenough information to construct an image of the body.123ISSN (Print) : 2319 – 2526, Volume-1, Issue-2, 2013

International Journal on Advanced Computer Theory and Engineering (IJACTE)V. METHODOLOGYThe part of the image containing the tumornormally has more intensity then the other portion andwe can assume the area, shape and radius of the tumorin the image. We have used these basic conditions todetect tumor in our code and the code goes through thfollowing steps: we present a new efficient edge tracingmethod, which was developed to find characteristicvector for the black-and-white images. The previouslyproposed edge tracing methods have two maindisadvantages. These are: The edge detection of onlyone object in case of more than one object in anumerical image and investigating only outer contoursnot inner contours.VI. FIGURESVII. PREPROCESSINGIn processing the following different steps arefollowed:-THRESHOLD SEGMENTATION:Segmentation isdone on basis of a threshold, due to which whole imageis converted into binary image. Basic matlab commandsfor thresholding are used for this segmentation.WATERSHED SEGMENTATION:It is the best methodto segment an image to separate a tumor but it suffersfrom over and under segmentation, due to which wehave used it as a check to our output. We have not usedwatershed segmentation on our input, rather it is onlyused on our output to check of the result is correct or notand it give the correct answer every time as is shownbelow.Morphological Operators:After that some morphological operations areapplied on the image after converting it into binaryform. The basic purpose of the operations is to showonly that part of the image which has the tumor that isthe part of the image having more intensity and morearea then that specified in the strel command.The basic commands used in this step are strel, imerodeand imdilate.Imerode():It is used to erode an image.Imdilate():It is used to dilate an imageIX. OUTPUTS/RESULTSWe have mapped the resultant tumor image onto theoriginal grayscale image for presentation purposes124ISSN (Print) : 2319 – 2526, Volume-1, Issue-2, 2013

International Journal on Advanced Computer Theory and Engineering (IJACTE)[3] Metin Akay, Time Frequency and Wavelets inBiomedical Signal Processing, IEEE Press Serieson Biomedical Engineering.[4] Starčuk Z, Starčuk jr, Horký J., ‘Baseline’problems in very short echo-time proton MRspectroscopy of low molecular weightmetabolites in the brain, Measurment ScienceReview, Vol.1, No.1, 2001, pp.17-20.[5] Weber-Fahr W, Ende G, Braus D, Bachert P,Soher J, Henn F, Buchel C., A Fully AutomatedMethod for Tissue Segmentation and CSF-Correction of Proton MRSI MetabolitesCorroborates Abnormal Hippocampal NAA inSchizophrenia, NeuroImage, Vol.16, 2002,pp.49–60.[6] Axelson D, Bakken I, Gribbestad I, Ehrnholm B,Nilsen G, Aasly J., Applications of NeuralNetwork Analyses to In Vivo 1H MagneticResonance Spectroscopy of Parkinson DiseasePatients, Journal of Magnetic ResonanceImaging, Vol.16, 2002, pp.13–20.X. REFERENCES[1] Bonavita S, Di Salle F, Tedeschi D., ProtonMRS in neurological disorders, European Journalof Radiology, No.31, 1999, pp.30-125.[2] Baik H, Choe B, Lee H, Suh T, Son B, Lee J.,Metabolic Alterations in Parkinson’s Diseaseafter halamotomy, as Revealed by 1H MRSpectroscopy, Korean J Radiol, Vol.3, No.3,2002, pp.180-9.125ISSN (Print) : 2319 – 2526, Volume-1, Issue-2, 2013