1

CLASSIFICATION AND DETECTION OF

BONE CANCER IN MEDICAL IMAGES
USING NEURAL NETWORK
1
Vivek naidu R, 1yaswanth gupta T, 1Raja sekhar B
1,2Department of Electronics and Communication Engineering
1,2SRM Institute of Science and Technology, Ramapuram, Chennai

Abstract: Modified surrenders area in Mrand CT pictures is outstandingly basic in various expressive and
supportive applications. Since of tall sum data in MR pictures and darkened boundaries, tumor division and
classification is especially troublesome.This work has shown one altered bone tumor disclosure strategy to
intensify the precision and leave and diminish the affirmation time. The objective is classifying the tissues to three
classes of ordinary, start and dangerous. . In MR and CT pictures, the aggregate of information is as well much for
manual elucidation and examination.. In the midst of past few a long time, bone cancer division in alluring
resonation imaging (MRI) has gotten to be an rising examine locale inside the field of restorative imaging system.
Exact disclosure of degree and area of bone cancer plays a pivotal portion inside the assurance of cancer. The

assurance technique comprises of four stages, pre-processing of MR pictures.

KEYWORD: Image Processing, Cancer Detection, Algorithms of Image Processing

1 INTRODUTION
This Disease can start any place in the body. It starts when cells become insane and bunch out customary
cells. This makes it hard for the body to work the manner in which it ought to. Tumor can be managed
incredibly well for certain people. Believe it or not, a greater number of people than some other time in
ongoing memory have full presences after tumor treatment. Threat isn't just a single contamination. There are
such infection. It's not just one contamination. Threat can start in the lungs, the chest, the colon, or even in
the blood. Infections are unclear somely, yet they are assorted in the manners in which they create and spread.
The cells in our bodies all have certain occupations to do. Normal cells confine in an efficient manner. They
pass on when they are depleted or hurt, and new cells have their spot. Infection is where the cells start to turn
out to be wild. The tumor cells keep creating and making new cells. They swarm out customary cells. This
causes issues in the piece of the body where the development started. Development cells can similarly spread
to various pieces of the body. For instance, danger cells in the lung can go to the bones and create there.
Exactly when danger cells spread, it's called metastasis (meh-TAStuh-sister). Right when lung development
spreads to the bones, it's actually called lung harm. To trained professionals, the development cells during the
bones look basically like the ones from the lung. It's not called bone threat except if it started during the
bones. A couple of malignancies create and spread fast. Others build up even more slowly. They moreover
respond to treatment differently. A couple of kinds of threat are best treated with a medical procedure; others
respond better to drugs called chemotherapy (key-mo-THER-uh-pee). Every now and again at any rate 2
drugs are used to get the best results. At the point when someone has tumor, the expert should find what kind
 2

of infection it is. People with harm require treatment that works for such a sickness.

For this process we are using image processing steps as given introduction of image processing.

Picture handling is a technique to play out certain procedure on a picture, to get an improved picture or to
remove some helpful data from it. It is a sort of sign preparing in which information is a picture and yield
might be picture or attributes/highlights related with that picture. These days, picture handling is among
quickly developing advances. It structures center exploration zone inside designing and software engineering
disciplines as well.

2 PROPOSED SYSTEM
PROPOSED METHODOLOGY:
a)Input image
The input picture is in DICOM organize this picture can be change over into JPEG organize and resize the picture,
since the picture is having more estimate, it requires more time for division prepare and less picture quality. So the
measure ought to be resized into 256*256. The input pictures for this work utilizing Bone (MRI) and Lung (CT)
pictures gotten from determination healing centers.
b) Preprocessing
Preprocessing is utilized to move forward the quality of an picture. Each picture has contained a few salt and
pepper clamor having a few blurriness. To expel the clamor and blurriness’ utilizing Middle channel.
c)Median filter
The middle channel may be a slidinsg windowsspatials channel, it replaces the center esteems of an windows the
centre of all the pixels .Due to changing the middle of all the values through the center esteem it evacuate the
commotion and protect the edges of an picture, its one sort of smoothening method. It improves the quality of an
picture. There's no lessening in differentiate, it doesn’t move boundaries and unlikely values are not made close
edges.
d) Segmentation process
DWT utilized to identify the cancer in bone (MRI) and lung (CT) pictures. The both calculations is utilized to
fragment the cancer from bone and lung pictures. The picture can be fragmented altogether and at last gotten the
picture into portions.
Classification
Ordinarily the classification is utilized to classify that the picture is typical or unusual. NN is one sort of classifier,
the highlights and values of the cancer influenced picture and non cancer picture is as of now put in database, the
concentrated is additionally having in cancer influenced picture, the classifier compares the given picture inside the
database in the event that the cancer is distinguished whereas comparing the each pixels, it show the message box
the cancer is influenced, after completing the NN preparing.
f) Feature extraction
The include extraction may be a major handle in acknowledgment applications and classifications, the surface
based highlight extraction is going on in this work, regularly a few surface based highlight extraction
classifications are there those are GLCM, LBP,SLBP... The gray scale invariant surface is measured and inferred
from definition of surface in neighborhood locale. It is an proficient surface administrator , it names picture pixels
by the edge handle from the neighborhood of each pixel and speaks to in double number. In this the cancer portion
 3

is extracted from the lung and bone pictures, usually based on the surface and differentiate of an picture. Input
cancer picture (MRI/CT) Preprocessing Division Utilizing Classification with KNN Highlight extraction
Measurable values
Here we are affirming that the organize of the cancer with respect to the cancer portion. By clustering the cancer
portion ready to classify and detect the arrange of the caner, whether it is within the dangerous arrange or kind
arrange. Here ready to consider that kind is the beginning organize of the caner ready to by recognizing this able to
diminish the fondness of cancer. But threatening is the ultimate organize of the caner we cant decrease the
condition.
f) Statistical values
The factual values are decided for both bone and lung pictures utilizing PSO and CSO strategies. The MSE, PSNR
values are calculated and the MSE is less than the PSNR esteem. The other parameters are affectability, specificity,

precision and handling time.

 4

MODULES

Create a Neural Network Using Keras

Kerasisasimple-to-usebutpowerfuldeeplearninglibraryforPython.Inthispost,we’llbuilda simple
Convolutional Neural Network (CNN) and train it to solve a real problem
withKeras.Kerascontainsnumerousimplementationsofcommonlyusedneural-
networkbuildingblockssuchaslayers,objectives,activationfunctions,optimizers,andahostoftoolstomakeworking
with image and text data easier to simplify the coding necessary for writing deepneuralnetwork code.

It is made user-friendly, extensible, and modular for facilitating faster experimentation withdeep neural
networks. It not only supports Convolutional Networks and Recurrent Networksindividuallybut also
theircombination.

Itcannothandlelow-levelcomputations,soitmakesuseoftheBackendlibrarytoresolveit.The backend library

act as a high-level API wrapper for the low-level API, which lets it runonTensorFlow,CNTK,orTheano.

Kerasencompassesthefollowingadvantages,whichareasfollows:

• Itisveryeasytounderstandandincorporatethefasterdeploymentofnetworkmodels.
• It has huge community support in the market as most of the AI companies are keen onusingit.
• Itsupportsmultibackend,whichmeansyoucanuseanyoneofthemamongTensorFlow,CNTK,andTheanowit
hKerasasabackendaccordingtoyourrequirement.
• Sinceithasaneasydeployment,italsoholdssupportforcross-platform.Followingarethedeviceson
whichKeras can bedeployed:
1. iOSwith CoreML
2. AndroidwithTensorFlowAndroid
3. Webbrowserwith.jssupport
4. Cloudengine
5. Raspberrypi

Figure 2. OVERALL VIEW of the Keras model:

 5

CHANGING MATRICE
EnteringSMatrice
Easy way for to urge start MATLAB is to memorize how to work frameworks. Begin MATLAB
take after long side illustration.
list of its components. You've got got because it were to require after a few crucial conventions: Enter an
unequivocal list of elements.
Start by entering Metrics framework as a
A = [161 3112 113; 51 1110 1111 18; 91 61 71 121; 4 1115 114 11]
MATLAB displays the matrix you just entered.
A=
161 31 12 113
51 1110 1111 18
91 61 71 121
41 115 114 11
This accurately matches the numbers inside the etching. Once you have got got entered the lattice, it is naturally
recollected within the MATLAB workspace. You will be able imply to it basically as A.
Expressions
programm lingos, MATLAB gives logical expressions, but not at all like most programming tongues, these
expressions incorporate entirety systems. The building squares of expressions are.
Variables
MATLAB does not require any sort assertions or estimation enunciations. When MATLAB encounters a advanced
variable title, it actually makes the variable and allocates the fitting entirety of capacity. On the off chance that the
variable as of presently exists, MATLAB changes its substance and, within the occasion that imperative, allocates
cutting edge capacity. For outline, num_students = 25 Creates a 1-by-1 framework named num_students and stores
the esteem 25 in its single component. Variable names include of a letter, taken after by any number of
letters, digits, or underscores. MATLAB livelihoods because it were the essential 31 characters of a variable title.
MATLAB is case sensitive; it recognizes between capitalized and lowercase letters. A and a are not the same
variable. To see the system doled out to any variable, essentially enter the variable title
6 Numpy
NumPy which stands for Numerical Python, is a library consisting of multidimensional array objects
and a collection of routines for processing those arrays. Using NumPy, mathematical and logical operations on
arrays can be performed. This tutorial explains the basics of NumPy such as its architecture and environment. It
also discusses the various array functions, types of indexing, etc. An introduction to Matplotlib is also provided.
All this is explained with the help of examples for better understanding.NumPy is a Python package. It stands
for 'Numerical Python'. It is a library consisting of multidimensional array objects and a collection of routines
for processing of array.
Numeric, the ancestor of NumPy, was developed by Jim Hugunin. Another package Num array was
also developed, having some additional functionalities. In 2005, Travis Oliphant created NumPy package by
incorporating the features of Num array into Numeric package. There are many contributors to this open source
project.
Operations using NumPy, a developer can perform the following operations −
• Mathematical and logical operations on arrays.
• Fourier transforms and routines for shape manipulation.
• Operations related to linear algebra. NumPy has in-built functions for linear algebra and
 6

random number generation.

Operators
Expression utilize recognizable number-crunching administrators and priority rules..

+ Additions
- Subtractions
* Multiplications
/ Divisions
\ Left divisions (described in "Matrices and Linear Algebra" in Using MATLAB)
^ Powers
' Complex conjugate transposes
() Specify evaluation orders

NumPy – A Replacement for MatLab

NumPy is often used along with packages like SciPy (Scientific Python) and
Mat−plotlib (plotting library). This combination is widely used as a replacement for MatLab, a popular platform
for technical computing. However, Python alternative to MatLab is now seen as a more modern and complete
programming language.
The most important object defined in NumPy is an N-dimensional array type called ndarray. It
describes the collection of items of the same type. Items in the collection can be accessed using a zero-based
index.
Every item in an ndarray takes the same size of block in the memory. Each element in ndarray is an
object of data-type object (called dtype).
Any item extracted from ndarray object (by slicing) is represented by a Python object of one of array
scalar types. The following diagram shows a relationship between ndarray, data type object (dtype) and array
scalar type −

Convolutional neural networks

Sounds like a weird combination of biology and math with a little CS sprinkled in, but these networks have been
some of the most influential innovations in the field of computer vision. 2012 was the first year that neural nets
grew to prominence as Alex Krizhevsky used them to win that year’s ImageNet competition (basically, the annual
Olympics of computer vision), dropping the classification error record from 26% to 15%, an astounding
improvement at the time.Ever since then, a host of companies have been using deep learning at the core of their
services. Facebook uses neural nets for their automatic tagging algorithms, Google for their photo search, Amazon
for their product recommendations, Pinterest for their home feed personalization, and Instagram for their search
infrastructure
MATLAB executes GUIs as figure windows containing different sorts of uicontrol objects. you want to program
each protest to perform the design activity when enacted by the client of the GUI. In expansion, you want to be
ready to spare and dispatch your GUI. All of those errands are streamlined by Direct, MATLAB's graphical client
interface advancement environment.

GUI Development Environment

the method of implementing a GUI involves two basic tasks:

 7

Laying out the GUI components

Programming the GUI components

GUIDE basically might be a group of format devices. Be that because it may, Direct moreover creates an M-file
that contains code to handle the initialization and propelling of the GUI. This M-file gives a system for the
execution of the callbacks - the functions that execute when clients enact components within the GUI.
Features of the GUIDE-Generated Application M-File
GUIDE disentangles the creation of GUI applications by consequently producing an M-file system
straightforwardly from your format. you will be able at that time utilize this technique to code your application M-
file. This approach gives variety of points of interest: The M-file contains code to actualize variety of valuable
highlights (see Arranging Application Alternatives for data on these highlights). The M-file receives an viable
approach to overseeing protest handles and executing callback schedules (see Making and Putting away the
Question Handle Structure for more data). The M-files gives how to oversee worldwide information (see
Overseeing GUI Information for more data).

TensorFlow
TensorFlow is an open-source library developed by Google primarily for deep learning applications. It
also supports traditional machine learning. TensorFlow was originallydevelopedforlargenumerical
computationswithoutkeepingdeeplearningin mind.However, it proved to be very useful for deep learning
development as well, andtherefore Google open-sourced it. TensorFlow
accepts data in the form of multi-dimensionalarrays of higher dimensions called tensors. Multi-
dimensional arrays are very handy inhandlinglargeamountsofdata.

TensorFlow works on the basis of data flow graphs that have nodes and edges. As theexecution
mechanism is in the form of graphs, it is much easier to execute TensorFlow codein a distributed manner across
a cluster of computers while using GPUs. TensorFlowprogramsworkon two basicconcepts:

1. Buildingacomputationalgraph

2. Executingacomputationalgraph

First, you need to start by writing the code for preparing the graph. Following this,
youcreateasessionwhereyou executethis graph.

Tensorflow programming is slightly different from regular programming. Even if you'refamiliar with
Python programming or machine learning programming in sci-kit-learn, thismaybeanew concept to you.

The way data is handled inside of the program itself is a little different from how it normallyis with the
regular programming language. For anything that keeps changing in regularprogramming,avariableneeds to
becreated.

In TensorFlow, however, data can be stored and manipulated using three different
programmingelements:

1. Constants

2. Variables

3. Placeholders
 8

First Layer – High Level Perspective

However, let’s talk about what this convolution is actually doing from a high level. Each of these filters
can be thought of as feature identifiers. When I say features, I’m talking about things like straight edges, simple
colors, and curves. Think about the simplest characteristics that all images have in common with each other.
Let’s say our first filter is 7 x 7 x 3 and is going to be a curve detector. (In this section, let’s ignore the fact that
the filter is 3 units deep and only consider the top depth slice of the filter and the image, for simplicity.)As a
curve detector, the filter will have a pixel structure in which there will be higher numerical values along the area
that is a shape of a curve (Remember, these filters that we’re talking about as just numbers!).
Neural Network

Neural networks are predictive models loosely based on the action of biological neurons.

The selection of the name “neural network” was one of the great PR successes of the Twentieth Century.
It certainly sounds more exciting than a technical description such as “A network of weighted, additive values with
nonlinear transfer functions”. However, despite the name, neural networks are far from “thinking machines” or
“artificial brains”. A typical artificial neural network might have a hundred neurons. In comparison, the human
nervous system is believed to have about 3x1010 neurons. We are still light years from “Data”.

The original “Perceptron” model was developed by Frank Rosenblatt in 1958. Rosenblatt’s model
consisted of three layers, (1) a “retina” that distributed inputs to the second layer, (2) “association units” that
combine the inputs with weights and trigger a threshold step function which feeds to the output layer, (3) the
output layer which combines the values. Unfortunately, the use of a step function in the neurons made the
perceptions difficult or impossible to train. A critical analysis of perceptrons published in 1969 by Marvin Minsky
and Seymore Paper pointed out a number of critical weaknesses of perceptrons, and, for a period of time, interest
in perceptrons waned.

Interest in neural networks was revived in 1986 when David Rumelhart, Geoffrey Hinton and Ronald
Williams published “Learning Internal Representations by Error Propagation”. They proposed a multilayer neural
network with nonlinear but differentiable transfer functions that avoided the pitfalls of the original perceptron’s
step functions. They also provided a reasonably effective training algorithm for neural networks.

Types of Neural Networks:

1. Artificial Neural Network

2. Probabilistic Neural Networks
3. General Regression Neural Networks

DTREG

DTREG implements the most widely used types of neural networks:

1. Multilayer Perceptron Networks (also known as multilayer feed-forward network),

2. CascadeCorrelationNeuralNetworks,

3. ProbabilisticNeuralNetworks(NN)

4. GeneralRegressionNeuralNetworks(GRNN).

RadialBasisFunctionNetworks:

a) FunctionalLinkNetworks,

b) Kohonennetworks,
c) Gram-Charliernetworks,
 9

d) Hebbnetworks,
e) Adalinenetworks,
f) HybridNetworks.

TheMultilayerPerceptronNeuralNetworkModel

Thefollowingdiagramillustratesaperceptronnetworkwiththreelayers:

This network has an input layer (on the left) with three neurons, one hiddenlayer (in the middle) with
three neurons and an output layer (on the right) withthreeneurons.There is one neuron in the input layer for each
predictor variable. In the case ofcategorical variables, N-1 neurons are used to represent the N categories of
thevariable.

InputLayer
Avectorofpredictorvariablevalues(x1...xp)ispresentedtothe input layer. The input layer (or processing
before the input layer) standardizesthesevaluessothattherangeofeachvariableis-1to1.Theinputlayerdistributesthe
values to each of the neurons in the hidden layer. In addition to the predictorvariables, there is a constant input of
1.0, called the bias that is fed to each of thehidden layers; the bias is multiplied by a weight and added to the sum
going intotheneuron.

Hidden Layer
Arriving at a neuron in the hidden layer, the value from eachinputneuronismultiplied byaweight
(wji),andthe resultingweightedvaluesare addedtogetherproducingacombinedvalueuj.Theweightedsum(uj)isfedinto
atransferfunction,σ,whichoutputsavaluehj.Theoutputsfromthehiddenlayer aredistributedtotheoutput layer.

Output Layer
Arriving at a neuron in the output layer, the value from
eachhiddenlayerneuronismultipliedbyaweight(wkj),andtheresultingweighted
valuesareaddedtogetherproducingacombinedvaluevj.Theweightedsum(vj)
isfedintoatransferfunction,σ,whichoutputsavalueyk.Theyvaluesarethe outputsofthenetwork.

Ifaregressionanalysisisbeingperformedwithacontinuoustargetvariable,thenthere is a single neuron in the

output layer, and it generates a single y value. Forclassification problems with categorical target variables, there
are N neurons intheoutputlayerproducing Nvalues,oneforeachoftheNcategoriesofthetargetvariable.

NeuralNetworks(NN)

Neural Network (NN) and General Regression Neural Networks (GRNN) havesimilar architectures,
but there is a fundamental difference:,networks
performclassificationwherethetargetvariableiscategorical,whereasgeneralregressionneural networks perform
regression where the target variable is continuous. Ifyou select a NN/GRNN network, DTREG will automatically
 10

select the correcttypeofnetworkbased on thetypeoftargetvariable.

ArchitectureofaNN

AllNNnetworkshavefourlayers:

Input layer
There is one neuron in the input layer for each predictorvariable. In the case of categorical variables, N-
1 neurons are used whereN is the number of categories. The input neurons (or processing before theinput layer)
standardizes the range of the values by subtracting the medianand dividing by the interquartile range. The input
neurons then feed thevaluesto eachoftheneurons inthehiddenlayer.

Hidden layer
This layer has one neuron for each case in the
trainingdataset.Theneuronstoresthevaluesofthepredictorvariablesforthecasealongwith the target value. When
presented with the x vector of input values from
theinputlayer,ahiddenneuroncomputestheEuclideandistanceofthetestcasefromthe neuron’s center point and then
applies the RBF kernel function using thesigmavalue(s).Theresultingvalueispassedtotheneuronsinthepatternlayer.

Pattern layer / Summation layer

The next layer in the network is different forNN networks and for GRNN networks. For NN networks
there is one patternneuron for each category of the target variable. The actual target category of
eachtrainingcaseisstoredwitheachhiddenneuron;theweightedvaluecomingoutofahiddenneuronisfedonlytothepattern
neuronthatcorrespondstothehiddenneuron’s category. The pattern neurons add the values for the class they
represent(hence,it is aweighted voteforthat category).

For GRNN networks, there are only two neurons in the pattern layer. One neuron isthe denominator
summation unit the other is the numerator summation unit. Thedenominator summation unit adds up the weight
values coming from each of
thehiddenneurons.Thenumeratorsummationunitaddsuptheweightvaluesmultipliedbytheactualtargetvalueforeachhi
ddenneuron.

Decision layer
The decision layer is different for NN and GRNNnetworks. For NN networks, the decision layer
compares the weighted votes foreach target category accumulated in the pattern layer and uses the largest vote
topredictthetarget category.

For GRNN networks, the decision layer divides the value accumulated inthe numerator summation unit
by the value in the denominator summationunitand uses theresultas thepredicted targetvalue.
 11

The following diagram is actual diagram or propose network used in ourproject

InputLayer
The input vector, denoted as p, ispresented as the black vertical bar.ItsdimensionisR ×1.In this paper,R
=3.
RadialBasisLayer
In Radial Basis Layer, the vector distances between input vector p and the
weightvectormadeofeachrowofweightmatrixWarecalculated.Here,thevectordistanceisdefinedasthedotproductbetw
eentwovectors[8].AssumethedimensionofWisQ×R. The dot product between p and the i-th row of W produces
the i-th element ofthedistancevector||W-p||,whosedimensionisQ×1.Theminussymbol,“-”,indicatesthat it is the
distance between vectors. Then, the bias vector b is combined with ||W-p||byanelement-by-
elementmultiplication,.Theresultisdenotedasn=||W-
p||..p.ThetransferfunctioninNNhasbuiltintoadistancecriterionwithrespecttoacenter.Inthispaper,itisdefinedasradbas(
n)=2ne-(1)Eachelementofnissubstituted layer.

Thei-thelementofacanberepresentedasai=radbas(||Wi-p||..bi) whereWiisthevectormadeofthei-
throwofWandbiisthei-thelementofbiasvector b.
SomecharacteristicsofRadialBasisLayer:
The i-thelement of a equals to 1 if the input p is identical to the ithrow of
inputweightmatrixW.Aradialbasisneuronwithaweightvectorclosetotheinputvectorpproducesavaluenear1andthenits
outputweightsinthecompetitivelayerwill pass their values to the competitive function. Itis also possible that
severalelementsofaarecloseto1sincetheinputpatternisclosetoseveraltrainingpatterns.
CompetitiveLayer
There is no bias in Competitive Layer. In Competitive Layer, the vector a is
firstlymultipliedwithlayerweightmatrixM,producinganoutputvectord.Thecompetitivefunction,denotedasCin
Fig.2,producesa1correspondingtothelargestelementofd, and 0’s elsewhere. The output vector of competitive
function is denoted as c.
Theindexof1incisthenumberoftumorthatthesystemcanclassify.Thedimensionofoutputvector,K,is 5inthis paper.

HowNNnetworkwork

Although the implementation is very different, neural networks are conceptuallysimilartoK-

NearestNeighbor(k-
NN)models.Thebasicideaisthatapredictedtargetvalueofanitemislikelytobeaboutthesameasotheritemsthathaveclosev
aluesofthepredictorvariables.Considerthis figure:
 12

Assumethateachcaseinthetrainingsethastwopredictorvariables,xandy.Thecasesare plotted using their

x,ycoordinates as shown in the figure. Also assume that
thetargetvariablehastwocategories,positivewhichisdenotedbyasquareandnegative
whichisdenotedbyadash.Now,supposewearetryingtopredictthevalueofanewcase represented by the triangle with
predictor values x=6, y=5.1. Should we predictthetarget aspositiveornegative?

Notice that the triangle is position almost exactly on top of a dash representing anegative value. But
that dash is in a fairly unusual position compared to the otherdashes which are clustered below the squares and
left of center. So it could be thattheunderlying negativevalueis anodd case.

Thenearestneighborclassificationperformedforthisexampledependsonhowmanyneighboring points are

considered. If 1-NN is used and only the closest point isconsidered, then clearly the new point should be classified
as negative since it is ontop of a known negative point. On the other hand, if 9-NN classification is used
andtheclosest9pointsareconsidered,thentheeffectofthesurrounding8positivepointsmayoverbalancetheclosenegative
point.

A neural network builds on this foundation and generalizes it to consider all of

theotherpoints.Thedistanceiscomputedfromthepointbeingevaluatedtoeachoftheother points, and a radial basis
function(RBF) (also called a kernel function)isappliedto thedistancetocomputetheweight(influence)foreach
point.Theradialbasisfunctionissonamedbecausetheradiusdistanceistheargumenttothefunction.

Weight=RBF(distance)
Thefurthersomeotherpointisfromthenewpoint,thelessinfluence

RadialBasisFunction
 13

Different types of radial basis functions could be used, but the most common istheGaussian function:

Advantages and disadvantages of NN networks

• ItisusuallymuchfastertotrainaNN/GRNNnetworkthanamultilayerperceptronnetwork.

• NN/GRNNnetworksoftenaremoreaccuratethanmultilayerperceptronnetworks.

• NN/GRNNnetworksarerelativelyinsensitivetooutliers(wildpoints).

• NNnetworksgenerateaccuratepredictedtargetprobabilityscores.

• NNnetworksapproachBayesoptimalclassification.

• NN/GRNNnetworksareslowerthanmultilayerperceptronnetworksatclassifyingnewcases.

• NN/GRNNnetworksrequiremorememoryspacetostorethemodel.
 14

3 IMPLENTATON

We have taken dataset of fingerprint images from kaggle. The finger images have beencaptured from
ten print inked fingerprint cards that were randomly selected to insure arepresentative mix of print with varying
quality, ranging from those of extremely poorquality to those of excellent quality. The fingerprint verification
competition databasescontainfourdisjointdatabases,eachcollectedwithadifferentsensor/technology.

Theimportanceofhavingpublicdatasetsisrecognizedinthebiometricscommunity,bothto evaluate the

performance of existing biometric systems and to encourage the developmentof new algorithms. The Images are
in uneven size we changed to desired size of 200*200with 3 bytes colour. By importing Keras Library which is
simple-to-use but powerful deeplearning library for Python it wraps the efficient numerical computation library
TensorFlowand allows you to define and train neural network models in just a few lines of code.

Wehave built 5 convolnutional layers and we built a dense layer in the nural network. We
usedactivationfunctioncalledReluwhichisthefunctionisanothernon-linearactivationfunction that has gained
popularity in the deep learning domain. ReLU stands for RectifiedLinear Unit. The main advantage of using the
ReLU function over other activation functionsis that it does not activate all the neurons at the same time. and the
dense layer is aboutsoftmax.So softmax which takes the input values can be positive, negative, zero, or
greaterthanone,butthesoftmax transformstheminto values between 0 and 1.UsingRmsprop(Root Mean Squared
Propagation) we categorized the images as authenticated andunauthenticated from tensorflow. We used matplotlib
to plot the graph in training foraccuracy and loss. Matplotlib is one of the most popular Python packages used for
datavisualization.
15
 16

4 CONCLUSION
By implementing the proposed design, we can save the people who are suffering from the cancer kind of diseases
from the beginning it may not be the cure but it can detect the disease tells us the accurate amount of bone affected
areas and the soul purpose of the project is to detect the The main purpose of the project is detect the bone effect
areas using image processing technologies with the better algorithm is in neural network families with the best
technologies of segmentation, feature extraction, pre-processing and clustering. The theme of the project is

detection of the affected area using with identification with the percentage based .

5 REFERENCES
1.John E Hall, Guyton and Hall textbook of medical physiology e-Book, Elsevier Health Sciences, 2015.
2.Multiple Myeloma Research Foundation (MMRF), “Multiple myeloma research foundation,” 20–. [3] Jean-Luc
HarousseauRgisBataille, “Multiple myeloma,” The New England Journal of Medicine, Massachusetts Medical
Society, Jun 5, 1997.
4.Samer Z Al-Quran, Lijun Yang, James M Magill, Raul C Braylan, and Vonda K Douglas-Nikitin, “Assessment
of bone marrow plasma cell infiltrates in multiple myeloma: the added value of cd138 immunohistochemistry,”
Human pathology, vol. 38, no. 12, pp. 1779– 1787, 2007.
5.Alastair Smith, Finn Wisloff, Diana Samson, Nordic Myeloma Study Group UK Myeloma Forum, and British
Committee for Standards in Haematology, “Guidelines on the diagnosis and management of multiple myeloma
2005,” British journal of haematology, vol. 132, no. 4, pp. 410–451, 2006.
6.Omid Sarrafzadeh, Hossein Rabbani, ArdeshirTalebi, and Hossein UsefiBanaem, “Selection of the best features
for leukocytes classification in blood smear microscopic images,” in Medical Imaging 2014: Digital Pathology.
International Society for Optics and Photonics, 2014, vol. 9041, p. 90410P.
7. Hamaoka T, Madewell JE, Podoloff DA, Hortobagyi GN, Ueno NT. Bone imaging in metastatic breast
cancer. Journal of Clinical Oncology. 2004;
8. Even-Sapir E, Metser U, Mishani E, Lievshitz G, Lerman H, Leibovitch I. The detection of bone metastases in
patients with high-risk prostate cancer: 99mTc-MDP Planar bone scintigraphy, single- and multifield-of-view
SPECT, 18F-fluoride PET, and 18F-fluoride PET/CT. Journal of Nuclear Medicine. 2006
Feb; 47(2):287–97. PMID: 16455635
9. Gold L, Lee C, Devine B. Imaging techniques for treatment evaluation for metastatic breast cancer.
Rockville (MD): Agency for Healthcare Research and Quality (US). 2014 Oct; https://doi.org/10.1007/
s12282-014-0560-0
10. Fogelman I, Cook G, Israel O, Van Der Wall H. Positron emission tomography and bone metastases.
Seminars in Nuclear Medicine Elsevier. 2005; 35:135–42.