Cervical cancer is the leading cause of cancer death for women in developing countries. Colposcopy plays an important role in early screening and detection of cervical intraepithelial neoplasia (CIN). In this paper, we developed a multimodal colposcopy system that combines multispectral reflectance, autofluorescence, and RGB imaging for in vivo detection of CIN, which is capable of dynamically recording multimodal data of the same region of interest (ROI). We studied the optical properties of cervical tissue to determine multi-wavelengths for different imaging modalities. Advanced algorithms based on the second derivative spectrum and the fluorescence intensity were developed to differentiate cervical tissue into two categories: squamous normal (SN) and high grade (HG) dysplasia. In the results, the kinetics of cervical reflectance and autofluorescence characteristics pre and post acetic acid application were observed and analyzed, and the image segmentation revealed good consistency with the gold standard of histopathology. Our pilot study demonstrated the clinical potential of this multimodal colposcopic system for in vivo detection of cervical cancer.
|