Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 40 views

    A Proposal For Automatic Diagnosis of Malaria: (Extended Abstract)

    Uploaded by

    Ardian Syah
    case

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    A Proposal For Automatic Diagnosis of Malaria: (Extended Abstract)

    Uploaded by

    Ardian Syah
    40 views2 pages
    case

    Original Title

    p681

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    case

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    40 views2 pages

    A Proposal For Automatic Diagnosis of Malaria: (Extended Abstract)

    Uploaded by

    Ardian Syah
    case

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 2

    A Proposal for Automatic Diagnosis of Malaria

    [Extended Abstract]
    Allisson D. Oliveira Giordano Cabral D. López
    University Federal Rural of University Federal Rural of Universitat Politécnica de
    Pernambuco (UFRPE) Pernambuco (UFRPE) Catalunya (UPC)
    two Brothers, 52171-900 two Brothers, 52171-900 despatx 2508860
    Recife-PE, Brazil Recife-PE, Brazil Barcelona, Spain
    allissondantas@gmail.com giordanorec@gmail.com daniel.lopezcodina@upc.edu
    Caetano. Firmo F. Zarzuela Serrat J. Albuquerque
    University Federal Rural of Institut Català de la Salut University Federal Rural of
    Pernambuco (UFRPE) (ICS) Pernambuco (UFRPE)
    two Brothers, 52171-900 Av Drassanes 08011 two Brothers, 52171-900
    Recife-PE, Brazil Barcelona, Spain Recife-PE, Brazil
    caetanof rmo@gmail.com francesc.bcn.ics@gencat.cat joa@deinfo.ufrpe.br

    ABSTRACT 2. RELATED WORK


    This paper presents a methodology for automatic diagnosis Crowd-sourcing approach has been used to pattern recog-
    of malaria using computer vision techniques combined with nition which in essence seeks to solve complex computational
    artificial intelligence. We had obtained an accuracy rate of problems. In this approach, the same problem is distributed
    74% in the detection system. to several participants who contribute to the resolution of
    the problem. Several gaming platforms have been developed
    Categories and Subject Descriptors to solve the problems as crowd-sourcing in biology and med-
    ical sciences. Thus this approach allows non specialists to
    H.4 [Information Systems Applications]: Miscellaneous; solve such problems. A recent work for recognition of in-
    D.2.8 [Software Engineering]: Metrics—complexity mea- fected cells by malaria uses crowd-sourcing [5]. It is a game
    sures, performance measures; I.2.1 [Artificial Intelligence]: for recognition of cells infected with malaria, where the slides
    Miscellaneous are given to non-specialists in malaria and in the background
    a learning system where artificial is made several compar-
    General Terms isons with issues of human and machine learning. This work
    reports impressive results of the non-specialist players who
    Experimentation, Algorithms.
    reported malaria infection with an accuracy of 1.25% when
    compared to a specialist doctor.
    Keywords
    Malaria, automatic diagnostic, detecting malaria, artificial 3. METHODOLOGY
    intelligence, haar, computer vision techniques. The facial recognition method proposed by Viola and Jones[7]
    is a known worldwide optimal algorithm for this purpose.
    1. INTRODUCTION However other studies like Firmo[1] demonstrate that it is
    Malaria is a worldwide public health problem. In 2010, possible to use this technique for disease detection in labo-
    according to the World Health Organization (WHO) ap- ratory slides.
    proximately 24 million cases were diagnosed with positive The method of Viola and Jones[7] uses rectangles as fea-
    incidence of the disease. A total of 106 endemic areas were tures to locate faces. These are rectangular objects with
    identified and Africa was the most affected continent [6]. It two light and dark regions. The calculation is given by the
    is caused by a protozoan of the genus Plasmodium, where difference between the sum of the intensities of the pixels of
    the vector is the female of the mosquito Anopheles. There clear region andP the sum ofP the intensities of the pixels in
    are four types of Plasmodium that cause malaria in humans: the dark region Rclear − Rdark .
    P.vivax, P.falciparum, P.malariae and P.ovale. The species Rectangles are illustrated in Figure 1.
    P. falciparum and P. vivax are responsible for 95% of cases
    reported worldwide [3].
    Malaria is very strongly related to remote areas such as
    in Africa, a system of automatic diagnosis and low cost had
    become a basic tool if we really want to control it.
    Figure 1: basic characteristics of haar
    Copyright is held by the author/owner(s).
    WWW 2013 Companion, May 13–17, 2013, Rio de Janeiro, Brazil. Rectangles can be processed as an Integral Image that is
    ACM 978-1-4503-2038-2/13/05. the representation of an image in your position (x, y). The

    681
    position contains the intensity value of all pixels above and 4. RESULTS
    to the left of point, including it its own location (x, y). For The training detection system consists of sequenced clas-
    such is given in equation 1. sifiers based on the methodology proposed by Viola and
    X Jones[7] in order to form a single strong classifier with the
    ii(x, y) = i(xl , y l ), (1)
    purpose of identifying malaria infection.
    xl ≤x,y l ≤y
    In his best performance reached It had obtained an accu-
    The evaluation of any feature rectangle can be made with 4 racy rate of 74% in classification of true positives in clas-
    reference to the image. The sum of the pixels of the rectan- sifiers with 10 stages, the set of tests was composed of 35
    gles of arbitrary size can be calculated in constant time. positive examples. We also carried out experiments with 10,
    To extract the features most effective algorithm is used 15, 22 and 25 classifiers.
    Adaboost, one of the features this algorithm is weight dis-
    tribution on sets of examples and a modification this dis- 5. CONCLUSIONS
    tribution during the course iterations of the algorithm, this
    This approach enables to train a technician with an auto-
    algorithm is based in bosting, where strong hypotheses are
    mated system and develop a malaria diagnosis on site and in
    formed by means of a linear combination of weak hypothe-
    real time. This is very important for remote areas. This is
    ses. This proposes then a cascade of classifiers[1][7].
    accomplished through a mobile device already informing the
    For the diagnosis of malaria were considered some steps
    result and medicating the patient or even refer to a treat-
    that are respectively image acquisition,preprocessing image,
    ment center or hospital data would be stored. The results
    training and classification.
    can be send to a central epidemiological once the device to
    In this first phase of the project image acquisition, we use
    get internet connection.
    a set of 13 images of malaria slides extracted by the labo-
    ratory group research MOSIMBIO of University Politecnica
    de Catalunya, identified 112 positive cases of malaria in this 6. ACKNOWLEDGMENTS
    set of 13 images, using the method for preparing blood smear This work was partially supported by FACEPE and UFRPE
    slides and a camera coupled to a microscope. and DEINFO.
    In the preprocessing image, in first step is used the image
    is converted in grayscale justified by the fact minimize the 7. REFERENCES
    amount of calculation per pixel and noise reduction.In the
    [1] AC.Firmo. Classificação adaboost com treinamento por
    second step histogram equalization that is a technique to
    enxame de partı́culas para diagnóstico da
    distribute the grayscale values of pixels in an image, so as
    esquistossomose mansônica no litoral de pernambuco.
    to obtain a uniform image[4][2].
    Master’s thesis, Universidade de Pernambuco, Recife,
    The training is given a set of positive images containing
    2010.
    face and a set of negative images that are not faces[7], in
    the case for malaria sets are defined infected with malaria [2] G. Bradski and A. Kaehler. Learning OpenCV:
    (positive) and not infected with malaria (negative) as Figure Computer Vision with the OpenCV Library. O’Reilly
    2 illustrates. Media, 2008.
    [3] L. S. Garcia. Malaria. Clinics in Laboratory Medicine,
    30(1):93 – 129, 2010. Emerging Pathogens.
    [4] R. González and R. Woods. Digital Image Processing.
    Prentice Hall, 2008.
    [5] S. Mavandadi, S. Dimitrov, S. Feng, F. Yu, U. Sikora,
    O. Yaglidere, S. Padmanabhan, K. Nielsen, and
    A. Ozcan. Distributed medical image analysis and
    diagnosis through crowd-sourced games: A malaria case
    Figure 2: (a)positive sample grayscale(b)negative
    study. PLoS ONE, 7(5):e37245, 05 2012.
    sample grayscale
    [6] W. H. Organization. Malaria, 2012.
    [7] P. Viola and M. J. Jones. Robust real-time face
    The capture of image of training is overseen per human,that
    detection. Int. J. Comput. Vision, 57(2):137–154, May
    save objects of interest , a total of 147 sample positive and
    2004.
    520 negative sample was used to create the set of examples.
    This samples are partitioned into set of training and valida-
    tion(76%) and test set(24%), the classifiers are formed using
    the conventional haartraining process[7][2].

    682

    You might also like