EFFICIENT CANCER DETECTION METHOD FOR DIFFERENTCANCER DISEASES: A REVIEW Amrutha Sunil Divya D Department of Computer Science and Engineering Mahatma Gandhi University Kottayam Department of Computer Science and Engineering Mahatma Gandhi University Kottayam Adi Shankara Institute Of Engineering and Technology Adi Shankara Institute Of Engineering and Technology Kalady, Kerala Kalady, Kerala Abstract – : Day by day cancer is grabbingour world. Cancer is a serious and common threat to human life. It is estimatedthat 8.2 million people died due to cancer. Cancer is the result of abnormalcell division and growth with the potential to invade or spread to the otherparts of the body.
There are more than 200 different types of cancer. Some areliver, lungs, skin, blood and breast etc. Lifestyles, environmental facts andgenetics can be the major reasons for cancer. Early detection of cancerincreases the possibility of successful treatment and a healthy life. Cancertreatment is really a fight between cancer attack and chemo soldiers. The paperis dealing with blood, liver, lungs, breast and brain cancer detection methods.Recognising the possible warning sign of cancer and taking proper action leadsto early diagnosis. Index Terms = Image fusion,Image processing, neural network I.
Introduction Cancer detecting is a multistageprocess. Sometimes the cancer is discovered by chance or from screening, thatis checking for cancer in people who have no symptoms. It can help the doctorsto find and treat several types of cancer at its early stage.
Cancer detectioninvolves radiological imaging. The common imaging method used to detect cancerand to monitor its spreading is CT scans (computed tomography) using X-rays. Itprovides cross-sectional imaging by the computer about hard tissues. MRI(magnetic resonance imaging) provides information about soft tissues. MRI usespowerful magnetic fields.The paper contains anoverview of skin, blood, breast, liver, lungs, and brain cancer detectionmethods. Brain cancer detection is enhanced through image fusion which allowscombination of features of different modality images1. To resultin an easy and reliable method to detect cancer tissue through fusion, itshould include the application of Discrete wavelet transform (DWT) and Neuralnetwork2.
Breast cancerdetection can be done by combining thermography and high frequency excitation techniques4. The paper provides a description of RFeffects on human body and simulation results. To validate the method amultilayer 3D breast model is simulated. Detecting lungs7 andliver5 cancer is a difficult task. The work proposes 2D and 3D CTscan method to detect the cancerous cells effectively.
II. METHODOLOGIESBrain Cancer Brain Cancer detectionis done by image fusion technique 1. This process combinesmultimodality images through image processing. For acquiring an enhanced image,several operations are performed on the image. Image processing is defined assignal processing, in which an image is taken as input and output may be animage or features of that image.
Fig 2 : CT image of Brain tumor The overall step inbrain tumor detection is illustrated infig 1.The first step isimage acquisition. It refers to the process of collecting the real world CT andMRI images and storing it in the database. Fig shows the sample CT image ofbrain that have infected cancer cells. Then the number of pixel in the image ischanged to obtain a new version of original image, this process is calledresampling. To increase the brightness, contrast and to reduce the variationsdue to noise are done by image processing. The 2 method applied for enhancements are contrast enhancement and noiseremoval. Contrast enhancement does not change the values that represent theimage instead, it modifies the color mapping and make it more bright.
Noiseremoval is process of extracting the necessary information by removing theunwanted details from an image.Next stage is imagedecomposition using DWT. It is done by passing the images through the filtersat different level of decomposition. By taking the coefficients of thedecomposed images, fusion of the images are performed. DWT is a numerical tool that is used todiscretely sample the wavelets and it produces high pass and low pass waveletseries.
The low frequency coefficient shows the gross approximations of thesource images and high frequency coefficients correspond to sharper brightnessin the image. Apply certain fusion rules on the obtained coefficients to mergeand then apply inverse DWT on the fused coefficient to obtain a fused image.Then the output is given as input to the segmentation stage. The activity ofsegmentation is the extraction of affected regions from the image, from whichinformation can easily be understood. The segmented image is used fot thedetection of cancer by using Feed forward neural network2classifier to extract the tumor cells from non tumor cells. Blood Cancer Fig3: Normal and leukemia affected Blood cells Cancer which affectsthe white blood cells are called Leukemia. Human blood consists of White bloodcells, Red blood cells, plasma and platelets. A person’s body who is sufferingfrom leukemia produces too many blood cells of certain type than another.
Theydo not function properly from normal cells. Leukemia is grouped in to 2 waysAcute and Chronic. Lymphoid cells and myeloid cells are the 2 type of abnormalwhite blood cells that can turn into leukemia. Blood cancer can beidentified by Fig 4: Blood cancer detection process 3 Microscopic imageacquisition is the process of collecting and storing microscopic image of bloodcells with significant magnification 3. Next step is imagepre-processing. Due to excessive staining the image may contain noise. Thisnoise has to be removed to improve the quality of image using proper techniqueand also remove the background of the image because our focus isonly the white blood cells. In image segmentation the image is partitioned intomultiple segments.
An automatic image segmentation will gives the accurateresults and will be able to differentiate the blast cell from the normal cells.Blast cells are abnormal immature white blood cells. Then thefeature of each blood cell is extracted to differentiate each cells withothers. The feature extraction is dealing with texture, color, geometrical andstatistical.
The feature of blast cells include scanty cytoplasm, round, allblast cells are uniform and usually contains single nucleoli inside nucleus. Nextis Image classification, here the K-nearest neighbour classifier classifies acell as normal cell or blast cell. It is done by comparing the extractedfeatures. Breast Cancer Now a day this is oneof the common reasons for the death of women.
Breast cancer can be diagnosedeffectively if it is detected at its early stage. Mammography, ultra sound,MRI, thermography are the common methods for detecting breast cancer. But theysuffer from lack of accuracy, reliability and high cost. In order to provideaccuracy and to lower the cost magneto-thermal approach is used for the breastcancer detection.
This method is based on the combination ofElectro-Magnetic and thermal analysis4. The breast tissue isexcited firstly with the radio frequency source and then measures thetemperature distribution of breast tissue. RF excitation causes more accurate result and provides additionalsignatures such as SAR for breast cancer detection. SAR is one of the importantmeasurement factors for energy absorption. Breast model comprises of 4 layersSkin, fat, gland and a malignant layer.
Dielectric properties of the layers areassumed to be frequency dependent. While the conductivity increases with thefrequency, the permittivity decreases. The frequency dependency is modelledusing linear model. The RF excitation of the breast tissue provides reliabledata due to higher temperature difference between malignant and normal tissue.
After exciting the tissue EM analysis isperformed. The electric field distribution near the tumor cells is higher thanthe rest of the breast tissue.it is due to the dielectric properties of tumorare different from the rest of the breast and it absorbs more energy. Liver Cancer Liver cancer rates areincreasing day by day. The disease can only be identified at the final stagesince it doesn’t show any symptoms.
The chance of defeating the disease isgreater if it is identified at its early stage. To make the task of detectingthe disease an effectively and efficiently less time consuming, simpler andtest are adopted. Fig 5:CT image of liver cancer MRI and CT scan imagesare taken as input to the system. First step is noise removal and then regionbased segmentation is performed. There are 4 types of segmentation5techniques; they are thresholding, boundary based, hybrid and region based.
CTimage is sufficient and low cost method compared to MRI. Tissues can be clearlyvisible in CT scan and it identifies normal and abnormal structures in thebody. To improve the quality of CT images, noise removal is done. Next step isthe image segmentation: it is a process of pixel classification which is doneby partitioning into the new subsets by assigning individual pixels to classes.Here region based segmentation is used which assume that neighbouring pixelswithin the same region should have similar values. Threshold value is computedat multiple times until a final value is derived using wavelet transformation6.The final value justifies whether the clustered image carrying the affectedpart is cancer cells or not. If all the values are within the same range thenthey are cancer affected cells and if variations are out of boundary then it isconsidered to be no cancer cells.
Lung Cancer It is also called aslung carcinoma. The symptoms of lung cancer are chest pain, coughing, breathingproblem etc. There is a chance for the cancer cells to move towards to bloodand lymph fluid surrounding the lung tissue.
Statistical studies say that about25% of cancer death is due to lung cancer. Early stage detection can save manylives. Fig 6: CT image of lungcancer The first step isimage acquisition, where the CT images are collected and taken as input to thesystem.
After acquisition the image is passed to image preprocessing systemwhere several conversion and removal take place. In gray scale conversiontechnique the RGB image is converted to gray scale. Then the image isnormalized using the MATLAB function imresize. The result will contain noise,they are removed by median filter.
Median filtering7 is an imageprocessing salt and pepper noise removal system. This noise free gry scaleimage is then transformed into pixels of 0’s and 1’s. From the original imageparts that are unwanted is to be removed from the binary image. Next step isbinarization, here the black and white images are generated. Segmentation isdone using threshold method. It uses 3 threshold values Thresh1, Thresh2,Thresh3.
According to this threshold values it is determined that whether thelungs is affected or not. If the white pixel percentage is greater than Thresh1then it is said that the whole lung is cancer affected. Right lung is affectedif the white pixel is greater than Thresh2, and left lung is affected if it isgreater than Thresh3. III CONCLUSION Any diseases can becured only if it is detected at the early stage. Cancer is now a wide rangespreading disease and its detection is a difficult task since its symptoms aresimilar to ordinary disease. From the paper it is sure that enhanced imageprocessing is one of the methods to detect the cancerearly. There are 100s of cancers other that blood, breast, brain, liver andlungs.
Early detection and regular screening is the technique to acquire goodhealth and better treatment References1R. K. Atyali and S. R. Khot, “An enhancement indetection of brain cancer throughimage fusion,” 2016 IEEE International Conference on Advances in Electronics,Communication and Computer Technology (ICAECCT), Pune, 2016, pp. 438-442.2Ambily P.K.
, Shine P.James,Remya R.Mohan”Brain tumor detection using image fusion and neuralnetwork” International Journal of Engineering Research andGeneral Science Volume 3, Issue 2, March-April, 2015 ISSN 2091-2730 13833M. Saritha, B. B. Prakash, K. Sukesh and B.Shrinivas, “Detection ofblood cancer in microscopic images of human blood samples: Areview,” 2016 International Conference on Electrical, Electronics, and Optimization Techniques(ICEEOT), Chennai, 2016, pp.
596- 600.4S. Rahmatinia and B. Fahimi, “Magneto-ThermalModeling of Biological Tissues: AStep Toward Breast Cancer Detection,” in IEEE Transactionson Magnetics, vol. 53, no.
6, pp. 1-4,June 2017.5 P. R. Anisha, C.
K. K. Reddy and L. V. N.
Prasad, “Apragmatic approach for detectingliver cancer using image processing and data miningtechniques,” 2015 International Conference on Signal Processing and CommunicationEngineering Systems, Guntur, 2015, pp. 352-357..6 Priyanka Kumar1 , Shailesh Bhalerao2 Detection of tumor in liver using image segmentation IOSR Journal of Electronics and Communication Engineering (IOSR-JECE)e-ISSN: 2278-2834,p- ISSN: 2278-8735.Volume 9,Issue 2, Ver. VIII (Mar – Apr.
2014), PP 110-1157 M. B. A. Miah andM. A. Yousuf, “Detection of lung cancer from CT image usingimage processing and neural network,” 2015International Conference on Electrical Engineering and Information CommunicationTechnology (ICEEICT),Dhaka, 2015, pp.