Available online at www.sciencedirect.

com

Procedia Computer Science 19 (2013) 1116 – 1120

The 8th International Symposium on Intelligent Systems Techniques for

Ad hoc and Wireless Sensor Networks (IST-AWSN)

M-Health: Skin Disease Analysis System

Camera
Abderrahim Bourouisa ,Ali Zerdazia , Mohammed Fehamb , Abdelhamid
Bouchachiac*
a
Algeria Telecom Research Group ,Cyber Parc Sidi Abdellah 16121 ,Algeria
b
STIC Laboratory ,Tlemcen 13000 ,Algeria
Smart Technology Research Centre , Bournemouth University , UK

Abstract

The increasing use of Smartphone and Tablet , their on-board sensors and increasing computational
power show immense promise for improved mobile health technologies , one of the medical areas which
needs this technologies is skin analysis to identify diseases . In this paper, we describe an innovative
development of a low cost smartphone based intelligent system using integrated camera. The advantage
of this solution is that patients in remote areas, poor and developing countries can scan, analyze and make
regular skin examinations in any place. In this paper, an Artificial Neural Network algorithm is used to
analyze the skin images to identify cancer.

©
© 2011 Published
2013 The byPublished
Authors. Elsevierby
Ltd. Selection
Elsevier B.V. and/or peer-review
Open access under responsibility
under CC BY-NC-ND license. of [name organizer]
Selection and peer-review under responsibility of Elhadi M. Shakshuki

Keywords :M-Health ; Smartphone ; Mobile NN.

*
Corresponding author.Tel :+213661231705.
E-mail addresses :bourouis@algerietelecom.dz,a.zerdazi@algerietelecom.dz,m_feham@mail.univ-tlemcen.dz,
abouchachia@bournemouth.ac.uk

1877-0509 © 2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of Elhadi M. Shakshuki
doi:10.1016/j.procs.2013.06.157
 Abderrahim Bourouis et al. / Procedia Computer Science 19 (2013) 1116 – 1120 1117

1. Introduction

One of the medical areas which needs mobile health technology is skin analysis to
identify diseases. Around 24% of the population in England and Wales (12.9 million people) visited
their general practitioner with a skin problem in 2006, with the most common reasons being skin
infection and eczema (WHO). One of the most preventable types of cancer is the skin cancer
and the best ways to keep the skin healthy and cancer free is checking and examining skin once a
month for suspicious moles or spots .This exam is a visual and clinical skin exam , it costs more
than 50 pounds and it needs hospital scan which is difficult for disabled people . In our paper, we
propose a low cost Smartphone based intelligent scheme that allows people for regular skin
examinations. We use a Smartphone camera and an intelligent learning algorithm to scan skin images.
Currently, there is no mobile health based skin analyser exists that detect skin disease
using mobile neural networks . This inspired us to develop a novel M-Health solution which
distinguishes between normal skin images and abnormal images using mobile neural networks.
An important work in the field of mobile health services and intelligent mobile system is
taking place in recent years by researchers, due to the importance of M-Health to provide efficient
and inexpensive health solution with the advancements in mobile technologies. Few
of them include pervasive health monitoring by wearable devices [1], a mobile health
monitoring system and a ring sensor for blood oxygen saturation level monitoring
are described in [2] and [3]. A Smartphone based health data acquisition system is presented in
[4], remote heart monitoring using cellular wireless networks is discussed in [5]. A design
approach for ECG data compression for a mobile tele-cardiology model involving a significant
compression ratio and reduction in transmission time over GSM network is described in [6].A
prototype system that uses a mobile device with a built-in camera, wireless network connectivity
and intelligent image analysis algorithms that estimates the calorie intake is proposed in [7]. A mobile
personal trainer wearable system that supervises physical activity and fitness
exercises in outdoor activity is described in [8]. A mobile cloud computing for m-health
services is presented in [9][10]. The use of mobile technologies for teledermatology,
telepathology and mobile monitoring the vital bio-signals of patients have been widely reported
and show promising progress [11-12].
Comparing with the above related research, the work presented here is distinctive in the following
aspects: (1) using an mobile inexpensive device to capture high resolution skin images; (2) using artificial
Neural Networks for local abnormal/normal skin image classification, i.e. distinguishing between normal
and disease related skin images; (3) implanting this Neural Network-based skin disease analyzer to an
Android-based mobile platform and performing real-time early detection of skin cancer.
The rest of the paper is organized as follows: Section 2 provides the methodology and
the functions of major components; in Section 3 we describe our research results. Section 4
summarizes and concludes this paper.

2. Methodology

Our system was designed to provide a fully functional feature rich software application
with simple graphical user interface to analyze skin images with mobile neural network.
The mobile device contained a good capacity of processing to conduct image
classification of skin. However, The neural network algorithm is trained and tested in computer
using skin datasets of normal and abnormal state , see Fig.1 . Mobile intelligence for analysis system
allows to have a substantial impact with disabled people to make regular scan in house.
1118 Abderrahim Bourouis et al. / Procedia Computer Science 19 (2013) 1116 – 1120

The design of the image acquisition of the skin was developed and optimized for
the use on a mobile smartphone device running Android Android OS is on three in
every four Smartphones, it is based on the Linux operating system. It provides all the core plumbing, such
as device drivers, memory and process management, network stack, and security.[13]. The android
Application Programming Interface (API) allowed easy access and control of the
integrated camera.

Fig. 1.Global architecture of mobile skin analysis

A Multi-Layer Perceptron (MLP) was employed to analyse images of 100 abnormal skin and 100
normal case. MLP algorithm is a feedforward neural network with one or more layers between input
and output layer. Feedforward means that data flows in one direction from input to output layer .This
type of network is trained with the backpropagation learning algorithm. Images were scanned by
Sumsung Galaxy S1 and resized into 32x32 before transferred to the computer to be trained , . The
NN program is developed by Netbeans JAVA modules using input layer ,one hidden with 20 neurons
and output layer with two classes : normal and abnormal . 20 neurons are selected for the hidden layer
because it leads to the most optimal results comparing with other type of neuron settings in the hidden
layer.
Input image is presented in a RGB one dimensional vector which contains all red , green and blue
values. The input of neural is presented by n (the dimension of the color skin image ) : n = (imageHeight
* imageWidth * 3) , see Fig 2 .
 Abderrahim Bourouis et al. / Procedia Computer Science 19 (2013) 1116 – 1120 1119

Fig .2 The architecture of the Neural Networks

The mobile device Galaxy S GT-I9000 is provided with an unmodified version of Android
OS 2.3.4, we tested the same application with rooted version 2.3.4.
The technical s p e c i f i c a t i o n s for the mobile device are:
1 GHz Cortex-A8 processor.
16GB storage, 512 MB RAM, 2GB ROM.
Super AMOLED capacitive touchscreen, 16M colours , 480 x 800 pixels, 4.0 inches
(~233 ppi pixel densit y) .
-3.2 MP camera with auto-focus and video recording capabilities, MMS-ready.
Tri-band UMTS/HSDPA Quad-band GSM world phone capabilit ies.

3. Results :

We train the system iteratively using computer, we started with 100 normal and 100 abnormal images
from skin images databases [14]. The total error rate in this step was 0.00996279170 .The android
application is an APK file , after installation and using our mobile neural networks algorithm , the
detection rate for the normal image from abnormal was 96,50 % when the image scanned is resized to
32x32 .
Energy consumption is very important factor for the monitoring system. We optimized it by using the
root method .We can root the Android OS by different ways, in our case we used one click root
software [15]. Figure 3 shows the difference between using root and unroot in battery life.
The initial evaluation indicates that the system can work more than 12 hours, which can be used for
tests of 50 patients, with Wifi network enabled.

Time (H)
14

12

10

8
Time (H)
6

0
With root Without root

Fig 3. Energy consumption in our free fall detector system

1120 Abderrahim Bourouis et al. / Procedia Computer Science 19 (2013) 1116 – 1120

4. Conclusion

In conclusion, a mobile health application is presented for skin cancer detection using commercially
available Smartpho ne. The application was designed and tested using the Android OS.
The features of this system are :
Low Cost Mobile Solution.
Used in Any Location.
Faster & Low Energy Consumption .
Easy to Use & No Operating Cost Intelligent skin Diseases Detection.

References

[1] Abderrahim Bourouis, Mohamed Feham, Abdelhamid Bouchachia .Ubiquitous Mobile Health Monitoring System for Elderly
(UMHMSE). International Journal of Computer Science & Information Technology ; June 2011 ,Vol 3, No 3.
[2] Qing Pan, Pan Yang, Rui Zhang, Chengyu Lin, Shijin Gong, Li Li, Jing Yan, Gangmin Ning . A mobile health system design
for home and community use, Biomedical and Health Informatics (BHI), 2012 IEEE-EMBS International Conference ;2012 , p.
116 - 119 .
[3] Asada H, Shaltis P, Reisner A, Rhee S, Hutchinson R. Mobile monitoring with wearable photoplethysmograhic biosensors.
IEEE Eng Med Biology Mag ;May-June 2003,p. 28 - 40 .
[4] Al-Taee M.A , Jaradat N.A , Ali, D.M.A. Mobile phone- based health data acquisition system using Bluetooth technology
.Applied Electrical Engineering and Computing Technologies (AEECT); 2011 ,p.1-6.
[5] Ross P. Managing care through air. IEEE Spectrum; Dec 2004, p.26- 31.
[6] Istepanian R, Petrosian A.Optimal Zonal wavelet-based ECG data compression for a mobile telecardiology system. IEEE
Trans on IT in Biomedicine ; Sept 2000 ,4(3): 200-211.
[7] Fengqing Zhu , Bosch M , Insoo Woo , SungYe Kim , Boushey C.J. , Ebert D.S. , Delp E.J.
The Use of Mobile Devices in Aiding Dietary Assessment and Evaluation. IEEE ,Journal of Selected Topics in Signal
Processing; Aug 2010,vol. 4, p. 756-766.
[8] F Buttussi, L Chittaro.MOPET.A Context-Aware and User Adaptive Wearable System for Fitness Training. Artificial
Intellegence in Medicine; February 2008, vol. 42, pp. 153-163.
Health Monitoring
International Journal of Computer Science & Information
Technology (IJCSIT); June 2011, Vol 3, No 3.
[10] Nkosi M.T and Mekuria, F. Cloud Computing for Enhanced Mobile Health Applications .The second IEEE International
Conference on Cloud Computing Technology and Science (CloudCom) ; 2010, p. 629 633.
[11] Zhang P, Kogure Y, Masuoka H.A remote patient monitoring system using a Java-enabled 3G mobile phone. Conference
IEEE Engineering in Medicine and Biology Society ;2007,3713-6.
[12] National Institutes of Health National Heart. National High Blood Pressure Education Program Working Group Report on
Hypertension in Diabetes. Hypertension 1994; 23(2):145-58.
[13] Paul Ferrill .Pro Android Python with SL4A Apress edition; 2011,p.16-30.
[14 ] http://dermatlas.med.jhmi.edu , Dec 2012.
[15] http://www.oneclickroot.com, Feb 2013.