3D Skin Texture Analysis: A Neural Network and Photometric Stereo Perspective - 12.030

S. Anwar et al., "3D Skin Texture Analysis: A Neural Network and Photometric Stereo Perspective", in Proc. of 3rd Int. Conf. on 3D Body Scanning Technologies, Lugano, Switzerland, 2012, pp. 30-40, https://doi.org/10.15221/12.030.

Title:

3D Skin Texture Analysis: A Neural Network and Photometric Stereo Perspective

Authors:

Shahzad ANWAR, Lyndon N. SMITH, Melvyn L. SMITH

Machine Vision Laboratory, University of the West of England, Bristol, UK

Abstract:

Cancerous skin lesions often exhibit irregular and non-axisymmetric morphologies and 3D textures, which are difficult to detect using existing techniques. Consequently, this paper describes the employment of a novel technique incorporating Multilayer perceptron (MLP) Neural Network (NN) and photometric stereo (PS) techniques for the analysis of complex lesions. The analyses of surface normal data (tilt and slant angles), for measurement of the degree of 3D skin surface disruption can provide potentially useful indicators for melanoma. Here this was achieved by replacing the axisymmetric 3D hemispherical profile with NN models of irregular lesion morphologies. PS was used to recover surface data and a NN was used to model the underlying forms of complex morphologies and irregular 3D textures. Initially, application of the method to three types of lesion resulted in an average difference (in slant angle) between the NN output and the actual surface normal of: 1.02, 2.11 and 2.86 degrees for axisymmetric, irregular and complex lesions, respectively. The proposed method performance was significantly better when compared to other methods. The experimental study performed shows the effectiveness of the proposed method, with ROC area under the curve of 84% to 88%.

Keywords:

3D clinical imaging, Photometric Stereo, bump map Neural Network

Details:

Full paper: 12.030.pdf
Proceedings: 3DBST 2012, 16-17 Oct. 2012, Lugano, Switzerland
Pages: 30-40
DOI: https://doi.org/10.15221/12.030

License/Copyright notice:

Proceedings: © Hometrica Consulting - Dr. Nicola D'Apuzzo, Switzerland, hometrica.ch.
Authors retain all rights to individual papers, which are licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
The papers appearing in the proceedings reflect the author's opinions. Their inclusion in the proceedings does not necessary constitute endorsement by the editor or by the publisher.

Note: click the + on the top left of the page to open/close the menu.