Classification of Osteoporosis using Fractal Texture Features

Classification of Osteoporosis using Fractal Texture Features Page 11
International Journal for Modern Trends in Science and Technology
ISSN: 2455-3778 |Volume No: 02 | Issue No: 02 | February 2016
Classification of Osteoporosis using
Fractal Texture Features
V.Srikanth, C.Dineshh Kumar and A.Tobin
Department of Electronics and Communication Engineering
Panimalar Engineering College
Chennai, Tamil Nadu, India
Article History ABSTRACT
Received on: 10-02-2016
Accepted on: 14-02-2016 In our proposed method an automatic Osteoporosis classification
system is developed. The input of the system is Lumbar spine digital
radiograph, which is subjected to pre-processing which includes
conversion of grayscale image to binary image and enhancement using
Contrast Limited Adaptive Histogram Equalization technique(CLAHE).
Further Fractal Texture features(SFTA) are extracted, then the image is
classified as Osteoporosis, Osteopenia and Normal using a Probabilistic
Neural Network(PNN). A total of 158 images have been used, out of
which 86 images are used for training the network and 32 images for
testing and 40 images for validation. The network is evaluated using a
confusion matrix and evaluation parameters like Sensitivity, Specificity,
precision and Accuracy are computed fractal feature extraction
techniques.
Published on: 16-02-2016
Keyword
Osteoporosis
Osteopenia
Lumbar Spine
CLAHE
SFTA
PNN
Copyright © 2015 International Journal for Modern Trends in Science and Technology
All rights reserved.

I. INTRODUCTION
Osteoporosis is a progressive bone disease that
is characterized by a decrease in bone mass and
density which can lead to an increased risk
of fracture. In osteoporosis, the Bone Mineral
Density (BMD) is reduced, bone micro architecture
deteriorates, and the amount and variety of
proteins in bone are altered. The form of
osteoporosis most common in women after
menopause is referred to as primary type 1 or
postmenopausal osteoporosis, which is attributable
to the decrease in oestrogen production after
menopause. Primary type 2 osteoporosis or senile
osteoporosis occurs after age 75 and is seen in
both females and males at a ratio of 2:1. Secondary
osteoporosis may arise at any age and affect men
and women equally, this form results from chronic
predisposing medical problems or disease, or
prolonged use of medications such as
glucocorticoid, when the disease is called steroid or
glucocorticoid-induced osteoporosis. So in order to
detect the presence of osteoporosis we have to use
image processing techniques on Digital
radiographers contribute to the test values used for
training the neural network. Thus Image
processing is used to monitor the traits of the X-
rays for possible infirmities that might occur and
provide the necessary data for further treatment.
II. LITERATURE SURVEY
According to AlceuFerraz Costa[1], a new and
efficient texture feature extraction method called
the Segmentation-based Fractal Texture Analysis,
or SFTA is preferred than Haralick and Gabor filter.
The extraction algorithm consists in decomposing
the input image into a set of binary images from
which the fractal dimensions of the resulting
regions are computed in order to describe
segmented texture patterns. The decomposition of
the input image is achieved by the Two- Threshold
Binary Decomposition (TTBD) algorithm, which we
also propose in this work. The SFTA is evaluated
for the tasks of content-based image retrieval
(CBIR) and image classification, comparing its
performance to that of other widely employed
feature extraction methods such as Haralick and
Gabor filter banks. SFTA achieved higher precision
and accuracy for CBIR and image classification.
Additionally, SFTA was at least 3.7 times faster
than Gabor and 1.6 times faster than Haralick with
respect to feature extraction time.
Histogram equalization, which stretches the
dynamic range of intensity, is the most common
method for enhancing the contrast of an image. An
adaptive method to avoid this drawback is block-
based processing of histogram equalization. In
block-based processing, image is divided into sub-
images or blocks, and histogram equalization is
performed to each sub-images or blocks. Contrast
Limited Adaptive Histogram Equalization (CLAHE),
proposed by K. Zuierveld, has two key parameters:
block size and clip limit. These parameters are
mainly used to control image quality. In this paper,
a new novel method was proposed by ByongSeok
Min [2] to determine two parameters of the CLAHE
using entropy of an image.
An Automatic Brain Tumor Classification using
PNN and Clustering developed by P.Sangeetha [3]
describes the Probabilistic Neural Network (PNN)
will be employed to classify the various stages of
Tumor cut levels such as Benign, Malignant or
Normal. Probabilistic Neural Network with Radial
Basis Function will be applied to implement tumor
cells segmentation and classification. Decision
should be made to classify the input image as
normal or abnormal cells. Prediction of malignant
cells or non-tumor cells can be executed using two
variants: i) Feature extraction and ii) classification
using Probabilistic Neural Network (PNN). The
ability of their proposed Brain Tumor Classification
method is demonstrated on the basis of obtained
results on Brain Tumor image database. In their
proposed method, only 5 classes of Brain tumors
are considered, with respect to an example of 20
test images for instance but this method can be
extended to more classes of Brain tumors.
Since medical X-Ray images are grayscale images
with almost the same texture characteristics,
conventional color or texture features cannot be
used for appropriate categorization in medical X-
Ray image archives. Therefore, a novel feature is
proposed by Seyyed Mohammad Mohammadi [4]
which is the combination of shape and texture
features. The feature extraction process is started
by edge and shape information extraction from
original medical X-Ray image. Finally, Gabor filter
is used to extract spectral texture features from
shape images. In order to study the effect of feature

fusion on the classification performance, different
effective features like local binary pattern. It
provides low computation complexity and
straightforward implementation. Due to following
advantages we can strongly claim that these
features are the most powerful and reliable features
for medical image X-Ray classification.
III. PROPOSED METHOD
The proposed approach starts first from
preprocessing. It is then followed by Grayscale
Conversion and the image is enhanced using a
CLAHE filter. Then the features namely fractal
features are extracted from the enhanced image
and the fractal features are used to determine the
category in which they fall determined by the PNN
classifier for one sample. This is repeated for the
stock of samples in which 2/3 of the samples are
taken as Database samples.
Methodologies
A. Pre-processing
B. Grayscale conversion
C. Image Enhancement.
D. Feature Extraction
E. PNN Classifier.
Figure 1.1 Block Diagram of proposed work
IV. RESEARCH METHODOLOGY
A. Preprocessing
Steps which are done prior to processing of an
image are called preprocessing. It includes image
enhancement and resizing. These are done in order
to make the image more suitable than an original
image for specific applications.
B. Grayscale Conversion
If the image selected is in three dimensions, it is
converted to a grayscale image using ‘rgb2gray’
conversion command for feature extraction. Then
the intensity variation of gray level image is shown
in the graph. Its value varies from 0 to 255.Resizing
of image is done for accurate processing of image.
Image can be resized to any size of our interest.
C, Image Enhancement
Image enhancement is the process of adjusting
digital images so that the results are more suitable
for display or further image analysis. After resizing
the image, we go for image enhancement. We use
CLAHE (Contrast Limited Adaptive Histogram
Equalization) technique. This method separates the
image into a number of tiles, and then adjusts the
contrast such that the tile histogram has the
desired shape. The tiles are then stitched together
using bilinear interpolation. The transformation
function modifies the pixels based on the gray level
content of an image. These techniques are used to
enhance details over small areas in an image.
Figure 1.2 Image after Enhancement
D. Feature Extraction
Feature is a parameter of interest to describe an
image. Transforming the input data into the set of
features is called feature extraction. If the features
extracted are carefully chosen it is expected that
the features set will extract the relevant
information from the input data in order to perform
the desired task. The Segmentation-based Fractal
Texture Analysis or SFTA method is a feature
extraction algorithm that decomposes a given
image into a set of binary images through the
application of what the authors call the Two
Threshold Binary Decomposition (TTBD). For each
resulting binary image, fractal dimensions of its
Training
Classification
(PNN Classifier)
Osteopenia Osteoporosis
Pre-
Processing(Image
Enhancement)
Feature Extraction(
SFTA Features)
Input X-Ray image Input X-Ray image
Pre-
Processing(Image
Enhancement )
Feature Extraction(SFTA
Features)
Features Database
Test Image
Normal

region boundaries are calculated that describe the
texture patterns.TTBD takes an input greyscale
image and returns a set of binary images by first
computing a set of T threshold values from the gray
level distribution information in an input image.
This is accomplished by recursively applying to
each image region the multilevel Otsu algorithm,
an algorithm that quickly finds the threshold that
minimizes the input image intra-class variance
until the desired number of thresholds is obtained.
The input image is decomposed into a set of binary
images by selecting pairs of thresholds from T and
applying two-threshold segmentation.Fractal
measurements are used to describe the boundary
complexity of objects, with each region boundaries
of a binary image represented as a border image.
The fractal dimension is computed from each
boarder image using a box counting algorithm.
A. Haussdorf fractal dimension
Haussdorf dimension serves as a measure of the
local size of a set of numbers, taking into account
the distance between each of its members. The
Haussdorf dimension of an n-dimensional inner
product space equals n. This underlies the earlier
statement that the Haussdorf dimension of a point
is zero, of a line is one, etc., and that irregular sets
can have non integer Haussdorf dimensions.
Haussdorf fractal dimension of an object
represented by the binary image. Non-zero pixel
belongs to an object and zero pixel constitute the
background.
𝑑𝑖𝑚 𝐻 𝑋 = inf⁡{𝑑 ≥ 0: 𝐶 𝐻
𝑑
𝑋 − 0} (1)
Where 〖 dim〗 _H(X) is the infimum of the set of
d∈ [0, ∞) such that the d-dimensional Haussdorf
measure of X is zero.
B. OTSU’S Segmentation
Otsu's method is used to automatically perform
clustering-based image thresholding i.e., the
reduction of a graylevel image to a binary image.
The algorithm assumes that the image contains
two classes of pixels following bi-modal histogram,
it then calculates the optimum threshold
separating the two classes so that their combined
spread is minimal.In Otsu's method, we
exhaustively search for the threshold that
minimizes the intra-class variance (the variance
within the class), defined as a weighted sum of
variances of the two classes.
𝜎 𝑤
2
𝑡 = 𝜔1 𝑡 𝜎1
2
𝑡 + 𝜔2 𝑡 𝜎2
2
𝑡 (2)
Weights ω_i are the probabilities of the two
classes separated by a threshold t and
σ_i^2variances of these classes. Otsu thresholding
returns a set of thresholds for the input image
employing the multilevel Otsu algorithm. The
multilevel Otsu algorithm consist in finding the
threshold that minimizes the input image intra-
class variants. Then, recursively, the Otsu
algorithm is applied to each image region until total
thresholds are found.
Figure 1.3 Image after OTSU Segmentation
C. Edge detection
Edge detection is an image processing technique
for finding the boundaries of objects within images.
It works by detecting discontinuities in brightness.
Edge detection is used for image segmentation and
data extraction in areas such as image processing,
computer vision, and machine vision.Edge
detection returns a binary image with the regions
boundaries of the input image. The input image
must be a binary image. The returned image takes
the value 1 if the corresponding pixel in the image
has the value 1 and at least one neighbouring pixel
with value 0. Otherwise takes value 0.
E. Artificial Neural Network
Artificial neural networks (ANNs) are a family of
statistical learning algorithms inspired by biological
neural networks. They are used to estimate or
approximate functions that can depend on a large
number of inputs and are generally unknown.
Artificial neural networks are generally presented
as systems of interconnected "neurons" which can
compute values from inputs, and are capable of
machine learning, as well as pattern recognition
thanks to their adaptive nature. After being
weighted and transformed by a function
(determined by the network's designer), the
activations of these neurons are then passed on to
other neurons. This process is repeated until
finally, an output neuron is activated.
A.PNN classifier
PNN is a useful neural network architecture with

h
1
h
2
h
3
h
4
x1
x2
x3 c2
c1
z
slightly different in fundamentals from the back
propagation. The architecture is feed forward in
nature which is similar to back propagation, but
differs in the way that learning occurs. PNN is
supervised learning algorithm but includes no
weights in its hidden layer.
Input nodes Hidden nodes Class nodes
Decision node
Figure 1.4 Block Diagram of PNN Classifier.
Basically, PNN consists of an input layer, which
represents the input pattern or feature vector. The
input layer is fully interconnected with the hidden
layer, which consists of the example vectors (the
training set for the PNN). The actual example vector
serves as the weights as applied to the input layer.
Finally, an output layer represents each of the
possible classes for which the input data can be
classified. The output class node with the largest
activation represents the winning class. In PNN
algorithm, calculating the class-node activations is
a simple process. For each class node, the example
vector activations are summed, which are the sum
of the products of the example vector and the input
vector. The hidden node activation, shown in the
following equation, is simply the product of the two
vectors (E is the example vector, and F is the input
feature vector).
ℎ𝑖 = 𝐸𝑖 𝐹 (3)
The class output activations are then defined as:
𝑐𝑗 =
𝑒
(ℎ 𝑖−1)
𝑦2𝑁
𝑖=1
𝑁
(4)
Where N is the total number of example vectors
for this class, ℎ𝑖 is the hidden-node activation, and
γ is a smoothing factor. The smoothing factor is
chosen through experimentation. If the smoothing
factor is too large, details can be lost, but if the
smoothing factor is too small, the classifier may not
generalize well . It's also very easy to add new
examples to the network by simply add the new
hidden node, and its output is used by the
particular class node. This can be done
dynamically as new classified examples are found.
The PNN also generalizes very well, even in the
context of noisy data.
V. RESULTS AND DISCUSSION
Here, totally 158 images have been used out of
which 86 are taken for training and remaining have
been used for testing and validation. Calculated
feature values of various input X-Ray images are
tabulated. Results show that images are normal,
are Osteopenia and osteoporosis. Comparison of
various features for normal, Osteopenia and
osteoporosis images can be seen.
Table 1.1 Table Showing Result Set
Class Number Of Images
Normal 55
Osteopenia 46
Osteoporosis 57
Figure 1.5 GUI Result shows Normal
Figure 1.6 GUI Result shows Osteopenia

Figure 1.7 GUI Result shows Osteoporosis
A. CONFUSION MATRIX
Confusion matrix is a specific table layout that
allows visualization of the performance of an
algorithm, typically a supervised learning one.
Each column of the matrix represents the
instances in a predicted class, while each row
represents the instances in an actual class.
Evaluation parameters such as Sensitivity,
Specificity, Precision and Accuracy are calculated
for the confusion matrix.
Table 1.2 Three Class Confusion Matrix of SFTA
45 2 0
4 65 4
0 2 34
Table 1.3 Table Showing Validation result of
SFTA
Parameter Formula SFTA
Sensitivity TP/ (TP+FN) 88.2%
Specificity TN/ (FP+TN) 94.2%
Precision TP/ (TP+FP) 97%
Accuracy (TP+TN)/ (TP+TN+FP+FN) 94.3%
VI. CONCLUSION AND FUTURE WORK
In this work, the suitability of texture features in
classification of Lumbar spine digital radiographs is
analyzed. The Experimental results during testing
and theoretical analysis prove in perspective of
time, accuracy of system being a main concern the
fractal features are more precise and more
accurate. In medical imaging or in diagnosis, the
important factor is accuracy rather than speed,
hence present fractal features as a more suitable
technique for feature extraction. The results on
classification have a combined accuracy of 93%.
This value is obtained by validating our practical
results with DEXA results. Hence our system can
assist in diagnosis of osteoporosis. In future the
work can be implemented for distal and thoracic
digital radiograph, the system can also be
embedded in a Digital radiography machine to
diagnose osteoporosis on the fly.
REFERENCES
[1] AlceuFerraz Costa, Gabriel Humpire-Mamani
andAgmaJuci Machado Traina, (2012) ‘An Efficient
Algorithm for Fractal Analysis of Textures’,
Conference on Graphics, Patterns and Images
(SIBGRAPI), pp. 39-46.
[2] ByongSeok Min, Dong Kyun Lim, Seung Jong Kim
and Joo Heung Lee, (2013)‘A Novel Method of
Determining Parameters of CLAHE Based on Image
Entropy’, International Journal of Software
Engineering and Its Applications, Vol.7, No.5,
pp.113-120.
[3] Sangeetha. P, (2014) ‘Brain Tumor Classification
using PNN and Clustering’, International Journal of
Innovative Research in Science, Engineering and
Technology, Vol. 3, No.3, Mar 2014.
[4] Seyyed Mohammad Mohammadi, Mohammad
SadeghHelfroush and Kamran kazemi, (2012) ‘Novel
shape-texture feature extraction for medical X-Ray
image classification’, International Journal of
Innovation Computing, Information and Control
(ICIC), Vol.8, No.1, ISSN 1349-4198.

Classification of Osteoporosis using Fractal Texture Features

More Related Content

What's hot (20)

Similar to Classification of Osteoporosis using Fractal Texture Features (20)

Recently uploaded (20)

Classification of Osteoporosis using Fractal Texture Features