Multi-scale hyperspectral imaging of cervical neoplasia

This preliminary study aimed at investigating the feasibility and effective of multi-scale hyperspectral imaging in detecting cervical neoplasia at both tissue and cellular levels.

In this paper, we describe a noninvasive diagnosis method with a hyperspectral imager for detection and location of cervical intraepithelial neoplasia (CIN) at multiple scales. At the macroscopic level, the hyperspectral imager was applied to capture the reflectance images of the entire cervix in vivo at a series of wavelengths. At the microscopic level, the hyperspectral imager was coupled with a microscope to collect the transmittance images of the pathological slide. The collected image data were calibrated. A wide-gap second derivative analysis was applied to differentiate CIN from other types of tissue.

At both macroscopic and microscopic levels, hyperspectral imaging analysis results were consistent with those of histopathological analysis, indicating the technical feasibility of multi-scale hyperspectral imaging for cervical neoplasia detection with accuracy and efficacy.

We propose a multi-scale hyperspectral imaging method for noninvasive detection of cervical neoplasia. Comparison of the imaging results with those of gold standard histologic measurements demonstrates that the hyperspectral diagnostic imaging system can distinguish CIN at both tissue and cellular levels.