Development of a computer vision binocular system for non-contact small animal model skin cancer tumour imaging

摘要

This paper describes the development of a novel gauging computer vision system for murine non-melanoma skin cancer tumours volume imaging. The system utilized binocular stereovision, enhanced through the use of telecentric lenses. These lenses optically compromised for the distortion factors and provided orthographic projection, leading to parallax free image acquisition. In order to improve the resolution of the system, a structured light projector, with 450 nm dominant wavelength, was used to illuminate the target with a custom pattern. Robust image processing algorithms granted accurate segmentation, feature recognition, labeling and correlation between the stereo pairs. Under these premises, the well-known "matching" problem was resolved successfully and geometrical interpolation provided an accurate three-dimensional reconstruction of the tumour volume. Through back-projection of the calibration object the resolution of the system was calculated up to 0.04 mm. The system was applied to measure the induced geometrical alterations of the tumour after PDT by using the Fosgel photosensitizer, excited by a laser diode emitting at 652 nm. The measurement of the volume induced alterations after each PDT treatment and up to the final tumour shrinkage is critical, to compare PDT efficacy between different protocols.