Delineation of malignant skin tumors by hyperspectral imaging using diffusion maps dimensionality reduction

V. Zheludev, I. Pölönen (Corresponding Author), N. Neittaanmäki-Perttu, A. Averbuch, P. Neittaanmäki, M. Grönroos, Heikki Saari

    Research output: Contribution to journalArticleScientificpeer-review

    15 Citations (Scopus)

    Abstract

    A new non-invasive method for delineation of lentigo maligna and lentigo maligna melanoma is demonstrated. The method is based on the analysis of the hyperspectral images taken in vivo before surgical excision of the lesions. For this, the characteristic features of the spectral signatures of diseased pixels and healthy pixels are extracted, which combine the intensities in a few selected wavebands with the coefficients of the wavelet frame transforms of the spectral curves. To reduce dimensionality and to reveal the internal structure of the datasets, the diffusion maps technique is applied. The averaged Nearest Neighbor and the Classification and Regression Tree (CART) classifiers are utilized as the decision units. To reduce false alarms by the CART classifier, the Aisles procedure is used.
    Original languageEnglish
    Pages (from-to)48 - 60
    JournalBiomedical Signal Processing and Control
    Volume16
    DOIs
    Publication statusPublished - 2015
    MoE publication typeA1 Journal article-refereed

      Fingerprint

    Keywords

    • hyperspectral imaging
    • framelet
    • delineation
    • malignant
    • tumor

    Cite this