Share Email Print

Proceedings Paper

A binocular machine vision system for non-melanoma skin cancer 3D reconstruction
Author(s): Dimitris S. Gorpas; Kostas Politopoulos; Eleni Alexandratou; Dido M. Yova
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Computer vision advancements have not till now achieved the accurate 3D reconstruction of objects smaller than 1cm diameter. Although this problem is of great importance in dermatology for Non Melanoma Skin Cancer diagnosis and therapy, has not yet been solved. This paper describes the development of a novel volumetric method for NMSC animal model tumors, using a binocular vision system. Monitoring NMSC tumors volume changes during PDT will grant important information for the assessment of the therapeutic progress and the efficiency of the applied drug. The vision system was designed taking into account the targets size and the flexibility. By using high resolution cameras with telecentric lenses most distortion factors were reduced significantly. Furthermore, z-axis movement was possible without requiring calibration, in contrary to wide angle lenses. The calibration was achieved by means of adapted photogrammetric technique. The required time for calibrating both cameras was less than a minute. For accuracy expansion, a structured light projector was used. The captured stereo-pair images were processed with modified morphological filters to improve background contrast and minimize noise. The determination of conjugate points was achieved via maximum correlation values and region properties, thus decreasing significantly the computational cost. The 3D reconstruction algorithm has been assessed with objects of known volumes and applied to animal model tumors with less than 0.6cm diameter. The achieved precision was at very high levels providing a standard deviation of 0.0313mm. The robustness of our system is based on the overall approach and on the size of the targets.

Paper Details

Date Published: 13 February 2006
PDF: 12 pages
Proc. SPIE 6081, Multimodal Biomedical Imaging, 60810D (13 February 2006); doi: 10.1117/12.643780
Show Author Affiliations
Dimitris S. Gorpas, National Technical Univ. of Athens (Greece)
Kostas Politopoulos, National Technical Univ. of Athens (Greece)
Eleni Alexandratou, National Technical Univ. of Athens (Greece)
Dido M. Yova, National Technical Univ. of Athens (Greece)

Published in SPIE Proceedings Vol. 6081:
Multimodal Biomedical Imaging
Fred S. Azar; Dimitris N. Metaxas, Editor(s)

© SPIE. Terms of Use
Back to Top