Abstract—Skincancer rates have been increasing for the past few decades. The risk factor isthe direct exposure of skin lesions to UV radiation which causes various skindiseases. Skin cancers are most common disease and are deadly to the human.Early detection of skin cancer can be cured. With the latest technologies,early detection is possible.
One of such technique is artificial intelligence.The dermoscopy image is given as input and it is processed for noise filteringand image enhancement. Then the image is segmented using thresholding. A cancerousskin has certain features and such features are extracted using featureextraction. These features are given as input to the neural network. The Neuralnetwork is used to classify whether it is cancerous or non-cancerous.Keywords-skin cancer,artificial intelligence,neuralnetwork,segmentation (key words) I.
IntroductionCancer whichaffects the skin is called skin cancer. Skin cancer is of two types malignantor benign form. Benign Melanoma is the appearance of moles on the skin it isnot a deadly one. Malignant melanoma is the appearance bleeding sores. It isthe deadliest form of all skin cancers. It arises from cancerous growth in pigmentedskin lesion.
If it is diagnosed at the right time, this disease is curable. Butdiagnosis is difficult. It needs sampling and laboratory tests. Throughlymphatic system or blood melanoma can spread to all parts of the body. Soautomatic detection will be useful at these cases. Basically skin diseasediagnosis depends on the different characteristics like color, shape, textureetc. there are no accepted treatment for skin diseases Different physicianswill treat differently for same symptoms. Key factor in skin diseases treatmentis early detection further treatment reliable on the early detection.
In thispaper, Proposed system is used for the diagnosis multiple skin disease using artificialintelligence and neural network.This paper is organized asfollows: Section I gives the introduction about Skin cancer and features ofskin cancers. Section II describes the Automatic Skin cancer Detection systemand various steps involved in the system. Section III gives the explanation ofvarious algorithms used yet. Section IV describes about the proposed system.
Section V concludes the paper followed by references. III. Automatic skin cancer detection A. ImageacquisitionThe first stage of any system is the acquisition of inputimage after the input image is obtained various process can be done on it toobtain the desired output here we do for image pre-processing and segmentation.CMOS camera is used as the medium for the acquisition of input image.B. NoisefilteringNoise filteringis the process of removing noisefrom an image.
Noise can be random or white noise with no coherence, orcoherent noise introduced by the device’s mechanism or processing algorithms. By doing this process we can obtain a good quality ofimage for the further segmentation.C. Segmentation This approach is adisplaying system that takes in a practical mapping from an information pictureto a yield picture.
The information picture is the first picture, and the yieldpicture is a division cover. This empowers the system to show useful residuals,and additionally to supply higher determination data to the yield layers, so asto enhance execution of the system in contrast with systems without the skipassociations. The exactness of the division procedure extraordinarilyinfluences ensuing component extraction and order. Factors Concerning the Segmentation Various factors that affect thesegmentation of skin cancer images are as follows: • The skin lesions have complexstructure, large variations in size as well as complex colours in the skin.
• The lesion is contrast to thesurrounding skin. • The borders of lesions are notalways well defined. • The influence of smallstructures, hairs, bubbles, light reflection, and other artifacts. • The influence of the skinlesions in the surrounding regions.
D. FeatureExtraction The highlights which have been utilized toportray the skin sore pictures are depicted. In this work, we utilize shading,surface, and shading histogram highlights to speak to injury zones. The purposeof picking these sorts of highlights is a result of the way that shading andsurface are the main properties commanding in the sore region.
Featureextraction is the critical device which can be utilized to dissect andinvestigate the picture properly. They include extraction depends on the ABCD manage ofdermatoscopy. The ABCD remains for Asymmetry, Border structure, Color varietyand Diameter of sore. It characterizes the reason for the conclusion of a malady.E. Classification: Injury grouping is the last advance. Soas to arrange a picture grouping strategies like SVM method is used: USING SUPPORT VECTOR MACHINE:BolsterVector Machines depend on the idea of choice planes.
A choice plane isotherwise called a hyper plane that isolates between arrangements of itemshaving distinctive class enrollments. Theisolating line characterizes a limit on the correct side of which all s areGREEN and to one side of which all items are RED. That is all focused on oneside of the hyper plane are named yes, while the others are delegated no. Thealgorithm of SVM classifier is given as 1.
Definition ofClassification Classes -Contingentupon the goal and the qualities of the picture information, the order classesought to be unmistakably characterized. 2. Selection of Features – Highlights to separate between the classesought to be set up utilizing multispectral as well as multi-transientattributes, surfaces and so on. 3.
Sampling of Training Data – Preparing information ought to be inspectkeeping in mind the end goal to decide proper choice tenets. 4. Estimation of UniversalStatistics – Different arrangement procedures will becontrasted and the preparation information, so that a suitable choice lead ischosen for ensuing grouping. 5.
Classification -In light of the choice administer, all pixels are ordered ina solitary class. There are two techniques for pixel by pixel arrangement andper – field grouping, regarding divided zones. III.comparision ofvarious algorithmsA.Fuzzy logicIn fuzzy logic algorithm, a combination of both ABCD(Asymmetry, Boarder factor, Color factor, Diameter) rules and Waveletcoefficients has been used to improve the image feature classification accuracyIn this, the percentage of red, blue,green is calculated using,Red% = R÷ (B+G) ×100 Blue% = B÷(R+G) ×100 Green% = G÷ (B+R) ×100C1-RED, C1-BLUE,C1-GREEN is calculated here inorder to determine if R/B/G isdominant over the other, Wavelet transform, Deconstruction, Reconstruction: The wavelet is repeated as,W (j) = W (j+1) + U (j+1)Fuzzy interference decision systemwill give us quantitative information about ABCD factors which is used withfuzzy interference system further. Accuracy is 60% only.B.
K-Nearest NeighbourKNN remains for k-closest neighbour calculation; it is one ofthe easiest yet generally utilized machines learning calculation. A protest isordered by the distance from its neighbours with the question being doled outto the class most basic among its k separate closest neighbours. On the offchance that k = 1, the calculation just turns out to be closest neighbor calculation,what’s more the protest is characterized to the class of its closest neighbour. The downside of kclosest neighbours classifier is, it is influenced by the quantities of features.The result might be because of the solver whose undertaking in little componentspace is harder than in bigger ones. Truth be told, as the dimensionalityexpands then the arrangement issue turns out to be all the more directlydetachable, which tends to facilitate the assignment of finding a legitimateisolating hyper plane. Hence, the preparation time will be longer when comparedto SVM.
C.Artificial Neural Network An Artificial NeuralNetwork (ANN) is a data handling that is roused incidentally organic sensorysystems, for example, the mind, process the data. An ANN is arranged for aparticular application, for example, design acknowledgement or informationorder, through a learning procedure.
A prepared neural system can be thought ofas a “specialist” in the class of data it has been given to breakdown.Incase of a medical field, error rates of ANN were high when compared to SVM inwhich 82.7% test set correctness has been achieved.
D.Support Vector MachineSVMs are presently a hotly debated issue in the machinelearning group, making a comparative eagerness at the minute as ArtificialNeural Networks used to do some time recently. Far being, SVMs yet speak to aneffective method for general (nonlinear) grouping, relapse and anomaly discoverywith a natural model portrayal.
Bolster vector machines are an arrangement ofrelated regulated learning strategies utilized for grouping and relapse. Givenan arrangement of preparing cases, each set apart as having a place with oneof two classifications, a SVM preparingcalculation assembles a model that predicts whether another illustration fallsinto one classification or then again the other. So, when compared to all abovemethods, SVM is good to go. IV.PRORPOSEDMETHODA.
Proposed block In this project we have designed a diagnosis system basedon the techniques of image processing. This work is done on different skinpatterns and tones of images and it is analyzed to obtain the result whetherthe person is suffering from skin cancer or not. This system helps in the earlydetection and cure of skin cancer .this is cost effective and feasible testmethod for the detection of skin cancer. The below mentioned is the block ofthe early detection skin cancer analyzer.A).Colour image to gray scale As the skin tone of people may differ,based on their region of living this may affect the efficiency in the output.
so in our project we convert the image to gray scale image to increase theefficiency of the output.B).Image restorationImageRestoration is the process of recovering the degraded image from a blurred andnoisy one. The degraded images can be stored in different ways. Such asimperfection of imaging system, bad focusing, motion and etc are the variousdefects which cause image degradation. The corrupted images lead to faultdetection, therefore, to select the most appropriate denoising algorithm it isessential to know about noises present in an image.
The image noises can bedivided into four groups of Gaussian, Salt and Pepper, Poisson and Speckle. Thesample of such noises has been shown below, a) Image without noise b) Gaussiannoise c) Poison noise d) Salt and Pepper noise e) Speckle noise Fig:1 Mean filters: It works bestwith Gaussian noise and for salt and pepper noise. Although this filter reducesthe noise, blur the image and reduce sharp edges. – Arithmetic mean filter: Itis the simplest of mean filter.
It can uniform the noise and works well withGaussian noise. – Geometric mean filter: Itcan preserve the detail information of an image better than the arithmetic meanfilter – Harmonic mean filter: Itworks well with salt noise, and other types of noise such as Gaussian noise,but doesn’tWork well with pepper noise. – Contra harmonic mean filter: It canpreserve the edge and remove noise much better than arithmetic mean filter. Because of preserving the edgescharacter we use harmonic and contra harmonic filters in this system.
Fig:2 C)Removing Thick Hairs Though the andskin lines such as rashes, moles will be smoothed using restoration filters,the image may include the hairs. Thick hairs in automated analysis of smallskin lesions are considered to mislead the segmentation process. To remove thethick hairs in skin cancer images, methods such as mathematical morphology methods,curvilinear structure detection, and automated software called Dull Razor andTop Hat transform combined with a bicubic interpolation approach are preferred.The hair-free images are acquired using these operations. a) Filtered image b) Segmented image Fig:3At the end of pre-processing step of skin cancerdetection system, the resulting images are distinguishable from those initialimages. D). Image enhancement Histogramequalization the technique of adjusting image intensities for enhancing thecontrast. It is one of the non-linear contrast enhancement technique.
Letf be a given image represented as mr/ mc matrix of integer pixel intensities ranging from 0 to L ? 1. L isthe number of possible intensity values. Often it will be 256. p is the normalized histogram of f with a binfor each possible intensity. So , pn = number of pixels with intensity n total number of pixels where, n = 0, 1, …
,L ? 1. Fig: 4E).Edge detection Canny edge detection uses multi stage algorithm for detectingwide range of edges in the image.
The generalcriteria for edge detection include:1. Detection of edge with low error rate, which means that thedetection should accurately catch as many edges as possible2. The edge point detected from the operator should accuratelylocalize on the center of the edge.3. A given edge in the image should only be marked once, and wherepossible, image noise should not create false edges.The Process of Canny edge detectionalgorithm can be broken down to 5 different steps:1.
In order to removethe noise apply Gaussian filter to smooth the imageThe intensity gradients of the image hasbeen found2. Apply non-maximumsuppression to get rid of spurious response to edge detection3. Apply double thresholdto determine potential edges4.
Track edge by hysteresis: Finalize the detection of edges bysuppressing all the other edges that are weak and not connected to strongedges. Fig:5a) It is the input image , b) It is the gray scale converted and enhancedimage,c) It is the canny edge detected imageF).Feature extraction Featureextraction is done using the properties called ABCDE in automated diagnosis ofskin cancer.
ABCDE represents Asymmetry, Border, Colour variation, Diameter andEvolution.Asymmetry: Asymmetric nature ofmelanoma is property in which the imaginary line passing through middle oflesion, either up or down or side to side gives two unequal or twonon-symmetric parts. Degree of asymmetry can be calculated by using asymmetricIndex which is calculated by using the formula,AI = (?A/A) × 100, where A isthe total area of the image and ?A is the difference in area between totalimage and lesion area.Border irregularity: Theborder or edge of the skin cancer affected area will be usually blurred orragged or irregular or notched. Border irregularity is usually calculated bycompact index in medical image processing. Compact index is used to estimateunanimous 2D objects. The measure is sensitive to noise along the boundary.
Compact index is calculated using the formula,CI= (pl*pl)/4?Alwhere Pl is Perimeter of theLesion and Al is area of the Lesion.Colour variation: Emergencein colour variation can be detected if lesion is melanoma. The colours can bevariations in black, brown and red depending on the production of melaninpigment in the affected area.
Colour variation can be detected statically andby plotting histograms of the segmented image. The intensity variation is highif there are colour variations.Diameter: Skin cancer(melanoma) usually have diameter more than 6mm.
Since diameter is irregular, itis calculated by drawing from edge pixels to pixels in the midpoint andaveraged. G). Classifier based on neural networks Fig:6A neural network consists of units(neurons), arranged in layers, which convert an input vector into someoutput.
Each unit takes an input, applies a (often nonlinear) function toit and then passes the output on to the next layer. Generally thenetworks are defined to be feed-forward: a unit feeds its output to all theunits on the next layer, but there is no feedback to the previous layer. Weightings are applied to the signals passing from one unit to another, and itis these weightings which are tuned in the training phase to adapt a neuralnetwork to the particular problem at hand. This is the learning phase.Neuralnetworks have found application in a wide variety of problems.
Theserange from function representation to pattern recognition, which is what wewill be consider here. Thus as the earlier mentioned theSVM classifier is used here to classify whether it is a cancerous or non-cancerousi.e., benign or malignant cancer.
The layered architecture of neural network isbeing used here for the classification purpose. The efficiency and the accuracyare expected up to 98 %. But the efficiency may vary according to the type ofsegmentation and classifiers used in it. V.
CONCLUSION A computerbased early detection of skin cancer analyzer system is being proposed. It hasbeen found to be a better diagnosis method than the artificial and k-nearestneighbor methods. This methodology uses image processing and support vectormachine for classification of malignant melanoma from other skin diseases.Dermoscopic image were collected and processed by various image processingtechniques. The cancerous region is separated from the healthy skin by themethod of segmentation. Based the features the images are classified ascancerous or non- cancerous. It has got good accuracy and efficiency of 98%also.
By further varying the image processing techniques and classifiersaccuracy and efficiency can be improved for this system.Acknowledgment We would like to thank our guide and professorsElectronics and communication Engineering Department, M.Kumarasamy College ofEngineering management for their guidance and support and facilities extendedto us.References