Spectral Band Subset Selection for Discrimination of Healthy Skin and Cutaneous Leishmanial Ulcers

Ricardo Franco-Ceballos, Maria C. Torres-Madronero, July Galeano-Zea, Javier Murillo, Artur Zarzycki, Johnson Garzon, Sara M. Robledo

    Research output: Chapter in Book/Report/Conference proceedingConference and proceedingspeer-review

    Abstract

    Leishmaniasis is a parasitic disease, transmitted by the bite of an insect that has previously fed on an infected host. One of its clinical forms is Cutaneous Leishmaniasis - CL and due to its increasing incidence, it is necessary to create effective and easy-use diagnostic methods. In this paper, we assess two unsupervised band-selection algorithms that allow the dimensional reduction of hyperspectral data taken from CL ulcers, maintaining a high classification accuracy. This is an important task for the development of an non-invasive system based on multispectral imaging, that support the diagnosis and treatment follow-up of cutaneous ulcer caused by Leishmaniasis. Spectral data was obtained in golden hamsters subjected to varying conditions of infection. Two algorithms, one based on similarity and the other based on singular values decomposition, are implemented using MATLAB functions and are applied to the spectral data. The selected subsets of bands are used to classify the spectra into healthy skin, border and ulcer centers using support vector machines - SVM and neural networks - NN. The obtained results are represented in precision tables and allow to observe that both methods achieve an appropriate dimensional reduction of multispectral data without losing key information for their subsequent classification. At the end, we show that it is possible to obtain a subset of spectral bands to discriminate between healthy skin and cutaneous ulcers caused by Leishmaniasis.

    Original languageEnglish
    Title of host publicationPattern Recognition and Image Analysis - 9th Iberian Conference, IbPRIA 2019, Proceedings
    EditorsAythami Morales, Julian Fierrez, José Salvador Sánchez, Bernardete Ribeiro
    PublisherSpringer
    Pages398-408
    Number of pages11
    ISBN (Print)9783030313319
    DOIs
    StatePublished - 2019
    Event9th Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2019 - Madrid, Spain
    Duration: 1 Jul 20194 Jul 2019

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume11867 LNCS
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Conference

    Conference9th Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2019
    Country/TerritorySpain
    CityMadrid
    Period1/07/194/07/19

    Bibliographical note

    Publisher Copyright:
    © 2019, Springer Nature Switzerland AG.

    Keywords

    • Classification
    • Hyperspectral data
    • Leishmaniasis
    • Spectral reduction
    • Unsupervised band selection

    Types Minciencias

    • Artículos de investigación con calidad A2 / Q2

    Fingerprint

    Dive into the research topics of 'Spectral Band Subset Selection for Discrimination of Healthy Skin and Cutaneous Leishmanial Ulcers'. Together they form a unique fingerprint.

    Cite this