International Journal of Innovations & Advancement in Computer Science

IJIACS
ISSN 2347 – 8616
Volume 7, Issue 3
March 2018

IoT Based Smart Lab for Diabetes Detection Using Iris Image

Prof. S.U. Deoghare Aboli Kamble Prajal Takalkar Megha Zende

Electronics and Electronics and Electronics and Electronics and
Telecommunication Telecommunication Telecommunication Telecommunication
PCCOE PCCOE PCCOE PCCOE
Pune,India Pune,India Pune,India Pune,India

ABSTRACT— Iris image analysis for clinical diagnosis optic thalami and spinal cord; receive impressions
is one of the most efficient non-invasive diagnosis from every nerve in the body. The nerve filaments,
methods for determining health status of organs. Correct muscle fibers and minute blood vessels in
and timely diagnosis is a critical, yet essential different areas of iris reproduce the changing
requirement of medical science. From the literature, it is
conditions in the corresponding organs. Here, the
found that modern technology also fails in lot of cases to
diagnose disease correctly. The attempt is being made to clinical information is extracted by means of
explore the area of diagnosis from different perspectives. various marks, signs, abnormal colors or
The approach used is a combination of ancestor's discolorations in the iris.
technology Iridodiagnosis with modern technology. Abnormalities in the iris are suggested to
Iridodiagnosis is an alternative branch of medical represent abnormalities in the respective organ.
science, which can be used for diagnostic purposes.To Colors, marks, textures, fibers and pigmentation
begin with, a database is created of eye images with
changes in the iris, as well as in the pupil and sclera,
clinical history of subject’s emphasis on diabetic disease
in pathological laboratory. The various algorithms are
may be studied. These signs are thought to correlate
developed for image quality assessment, segmentation of with disease. Different colors or rings within the iris
iris, iris normalization and clinical feature classification are believed to represent different aspects of health
for clinical diagnosis. A significant improvement is and to play a role in diagnosis.
demonstrated in classification performance over the Modern technology also fails in lot of cases
existing approaches. This approach will be useful in the to diagnose disease correctly. The attempt is being
diagnosis field which is faster, user friendly and less time made to explore the area of diagnosis from different
consuming.
perspectives. The approach used is a combination of
KEYWORDS— Classification, Feature extraction, ancestor’s technology Irido diagnosis with modern
IoT, iridodiagnosis, Iris, Segmentation Irdiology Chart, technology. Irido diagnosis is an alternative branch
Iris, ROI, PCA, ANN,PNN,KNN Diabetes
of medical science, which can be used for
diagnostic purposes.
INTRODUCTION Glaucoma is an optic nerve problem that
According to recent research held by world health reduces the vision. Intraocular pressure mainly
organization (who) india had 69.2 million people creates glaucoma. Open angle glaucoma and closed
living with diabetes (8.7%) as per the 2015 angle glaucoma are the two common types of
data.india is among top threecountries with high glaucoma. Glaucoma type is predicted ibelown the
diabetic population. So it is necessary to find the irdio corneal angle images between the iris and
easiest method to detect the diabetes. cornea. It is a more time consuming process.
After research we come to know Machine learning built the association between the
thatIridology is one the method to detect the focal edge and angle grades. The experimental
diabetes.Iridology is the study of the iris (the result shows 87.3% open angle and 88.4% closed
colored part of the eye). The intention of iridology angle. Iridology technique is used to for the
is to gain information about underlying diseases. identification examination. It not only examines the
The iris of an eye is connected to immense number disease and also found the unhealthy conditions,
of minute nerve filaments through optic nerves, toxin precipitation and other degeneration of

333 Prof. S.U. Deoghare, Aboli Kamble, Prajal Takalkar, Megha Zende
 International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 7, Issue 3
March 2018

organic functions. It is actually an iris test which For recognization of any diseases we refer standard
identifies and predicts the disease. This system iridology chart, From which we come to know that,
contains four processes that include image capture, Which region of Iris connected to which part of
pre-processing, template generation, feature body. So, from this we recognize many types of
extraction and pattern. diseases. For demonstration purpose we are going to
recognize diabetes. From irdiology chart, we find
PROPOSED APPROACH that from liver section.
The block diagram of proposed system is shown
Iridology (also known as iridodiagnosis or
in Fig
iridiagnosis) is an alternative medicine technique
whose proponents claim that patterns, colors, and
other characteristics of the iris can be examined to
determine information about a patient's systemic
health. The basic premise of Iridology is holistic in
nature to consider the relationships within the body
and where imbalances are located.

FLOWCHART
Step 1:
Initially we take iris images which is captured with
the help of certain cameras, or casia and stored in
the database which contains normal as well as
abnormal results of iris.
Step 2:
Segmentation:
The captured iris image is converted into gray
Fig No.1.Block diagram of proposed system level so that further image processing will be easy.
Iris segmentation is one of the most important steps
IRDIOLOGY CHART in iris image analysis in order to avoid errors in the
segmentation the active counter approach is used.
The basic concept of our proposed system is based
The basic idea of active contour is to use a
on the iridology chart. The iridology chart is as
deformable model and let it evolve in each iteration
shown in figure below,
to minimize a given energy function.
There are two methods for active contours:
Edge-based: - Edge-based method uses an edge-
detector which is dependent on the image gradient
to stop the evolving curve on the boundary of the
desired object.
Region-based: - In Region-based method, instead
of finding the edges, the foreground and
background regions are detected. It can detect
objects whose boundaries are not necessarily
defined by the gradient.
We have used region based methods due to
its various advantages over edge based methods
such as insensitivity to noise and robustness.
Intensity information is extracted in local regions at
a controllable scale, to find the pupil and iris
Fig No.2.Irdiology Chart boundaries accurately. One advantage of our

334 Prof. S.U. Deoghare, Aboli Kamble, Prajal Takalkar, Megha Zende
 International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 7, Issue 3
March 2018

method is that it accurately localizes the pupillary processing stage. Where, we can find new range
and limbic boundary, once the priori estimate is set from an existing one range.
and also overcomes the difficulties caused by Step 4:
intensity in homogeneities.
ROI extraction:
After normalization, the next step which
comes into is ROI extraction. ROI extraction is
nothing but cropping particular portion of
normalized iris image according to “iridology
chart”. We are interested in Liver section which is
connected to iris somewhere .So, we are extracting
that part only.

Step 5:
Feature extraction using PCA algorithm:
Feature extraction includes the features of
the extracted region such as mean, variance,
standard deviation etc. According to the iridology-
chart, the region of interest can be identified.
Principal Component Analysis (PCA) is used for
extracting the features of region of interest.Principal
component analysis (PCA) is one of the statistical
techniques frequently used in signal processing to
the data dimension reduction or to the data
decorrelation. The region of interest is identified by
visual inspection as per chart of Iridology. The
dimension of the processing image
reduced to 100 × 100 which is region of interest.
The features for region of interest are extracted
using Principal Component Analysis (PCA).For
applying the PCA to the region of interest, matrix
manipulation is performed on the ROI. The PCA
works on whole data base rather than single, it is
Fig No.3.Flowchart required to store all features of images into one 2D
matrix. Finally 2D matrix of image features is
Step 3: converted into single dimension vector and
concatenate each image features to form matrix.
Normalization: PCA basically does the job of providing weightage
Once the iris region is segmented from an to particular feature; for that purpose we need to
eye image, the next step is to transform the iris subtract the mean of pixel value of all images of
region into fixed dimensions. The dimensional data base from each image pixel value.
inconsistencies between iris images are mainly due
Step 6 :
to the stretching of the iris caused by pupil dilation
from varying levels of illumination.Normalization is Feature Classification :We use kNN for feature
a process that changes the range of pixel intensity classification .kNN is one of the simplest of
values.Normalization is sometimes called contrast classification algorithms available for supervised
stretching or histogram stretching. Normalization is learning. The idea is to search for closest match of
scaling technique or a mapping technique or a pre the test data in feature space. In pattern recognition,
the k-nearest neighbors algorithm (k-NN) is a non-
parametric method used for classification and

335 Prof. S.U. Deoghare, Aboli Kamble, Prajal Takalkar, Megha Zende
 International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 7, Issue 3
March 2018

regression.[1] In both cases, the input consists of the from iris image, So we train our data base to find
k closest training examples in the feature space. The out exact location of ROI using Knn,Pnn and Ann
output depends on whether k-NN is used for classifier. After that the system will be able to
classification or regression: In k-NN classification, detect the presence of diabetes in human body and
the output is a class membership. An object is result will be displayed on the screen as shown in
classified by a majority vote of its neighbors, with above figure whether the iris detected is normal iris
the object being assigned to the class most common or diabetic iris.
among its k nearest neighbors (k is a positive
integer, typically small). If k = 1, then the object is
CONCLUSION
simply assigned to the class of that single nearest
This proposed method is robust and
neighbor.
effective to perform the task of suggesting diagnosis
The best choice of k depends upon the data; of iris and authentication. This will enhance and
generally, larger values of k reduce the effect of bring more confidence to the diagnosing process.
noise on the classification, but make boundaries The iris is very suitable for the verification and
between classes less distinct. A good k can be identification of humans due to its distinctive and
selected by various heuristic techniques (see stable spatial patterns. This effective iris
hyperparameter optimization). The special case recognition is used recognition for person
where the class is predicted to be the class of the identification and check whether iris is affected or
closest training sample (i.e. when k = 1) is called the not and to identify the disease affected part of the
nearest neighbor algorithm. human eye.
Since, each & every part of iris represent
the various parts of human body, in future it will be
used to detect the various diseases without more
harm to human body on the basis of iris images.

ACKNOWLEDGMENT
It is a great pleasure for us to present a
project “IoT based Smart Lab for Diabetes
Detection using Iris Image” where guidance plays
an invaluable key and provides concrete platform
for completion of the project.
The hard work and perseverance of our
mentor will always be embedded in our memory.
Project execution would not have been possible for
us without the continued assistance of certain
people. We take this opportunity to express our
deepest gratitude for all the heartfelt assistance
rendered.
We thank our project guide
Prof.S.U.Deoghare who was responsible for
coordinating all efforts and sincerely grateful to her
Fig No.4 Result
for helping to achieve high standards of
performance.
We are proposing a new framework for
Detection of Diabetic from Iris image. For feature
JOURNAL /ARTICLE /PAPER
analysis, enhancement is essential to extraction of
[1] Piyush Samant, Ravinder Agarwal , “Diagnosis of
deep layer features. Diabetes Using Computer Methods: Soft
In this proposed system we acquired one Computing Methods for Diabetes Detection Using
eye image, stored into database. As there is no Iris”, International Scholarly and Scientific Research
standard co-ordinates to find out only liver section

336 Prof. S.U. Deoghare, Aboli Kamble, Prajal Takalkar, Megha Zende
 International Journal of Innovations & Advancement in Computer Science
IJIACS
ISSN 2347 – 8616
Volume 7, Issue 3
March 2018

& Innovation 11(2) 2017, [4] On Networking of Internet of Things: Explorations

scholar.waset.org/1999.9/10006736 and Challenges”, June 2012
[2] G. DurgaDevi D.M.D Preethi, “Disease Identification Book:
in Iris Using Gabor Filter”, International Journal Of [1] “Digital Image Processing” by S. Shridhar
Engineering And Computer Science ISSN:2319- Publication 3rd edition
7242 Volume 3 Issue 4 April, 2014 Page No. 5396-
5399 Websites:
[3] U. M Chaskar, M. S. Sutaone, “On a Methodology [1] http:// www.irisdiagnosis.org (Accesses May 2010).
for Detecting Diabetic from Iris Image Analysis.” [2] B. Jensen, “Iridology Chart.” (Online) Available:
2012 IEEE, 978-1-4673-0449-8/12/$31.00 http://www.bernardjensen.com/(Accessed: 01-Nov-2016)

337 Prof. S.U. Deoghare, Aboli Kamble, Prajal Takalkar, Megha Zende