Hyperspectral wide gap second derivative analysis for in vivo detection of cervical intraepithelial neoplasia.
J Biomed Opt
; 20(12): 121303, 2015 Dec.
Article
em En
| MEDLINE
| ID: mdl-26220210
Hyperspectral reflectance imaging technique has been used for in vivo detection of cervical intraepithelial neoplasia. However, the clinical outcome of this technique is suboptimal owing to multiple limitations such as nonuniform illumination, high-cost and bulky setup, and time-consuming data acquisition and processing. To overcome these limitations, we acquired the hyperspectral data cube in a wavelength ranging from 600 to 800 nm and processed it by a wide gap second derivative analysis method. This method effectively reduced the image artifacts caused by nonuniform illumination and background absorption. Furthermore, with second derivative analysis, only three specific wavelengths (620, 696, and 772 nm) are needed for tissue classification with optimal separability. Clinical feasibility of the proposed image analysis and classification method was tested in a clinical trial where cervical hyperspectral images from three patients were used for classification analysis. Our proposed method successfully classified the cervix tissue into three categories of normal, inflammation and high-grade lesion. These classification results were coincident with those by an experienced gynecology oncologist after applying acetic acid. Our preliminary clinical study has demonstrated the technical feasibility for in vivo and noninvasive detection of cervical neoplasia without acetic acid. Further clinical research is needed in order to establish a large-scale diagnostic database and optimize the tissue classification technique.
Texto completo:
1
Coleções:
01-internacional
Base de dados:
MEDLINE
Assunto principal:
Análise Espectral
/
Interpretação de Imagem Assistida por Computador
/
Aumento da Imagem
/
Displasia do Colo do Útero
/
Neoplasias do Colo do Útero
/
Microscopia
Tipo de estudo:
Diagnostic_studies
Limite:
Female
/
Humans
Idioma:
En
Ano de publicação:
2015
Tipo de documento:
Article