Characterizing optical properties and spatial heterogeneity of human ovarian tissue using spatial frequency domain imaging

Abstract

A spatial frequency domain imaging (SFDI) system was developed for characterizing ex vivo human ovarian tissue using wide-field absorption and scattering properties and their spatial heterogeneities. Based on the observed differences between absorption and scattering images of different ovarian tissue groups, six parameters were quantitatively extracted. These are the mean absorption and scattering, spatial heterogeneities of both absorption and scattering maps measured by a standard deviation, and a fitting error of a Gaussian model fitted to normalized mean Radon transform of the absorption and scattering maps. A logistic regression model was used for classification of malignant and normal ovarian tissues. A sensitivity of 95%, specificity of 100%, and area under the curve of 0.98 were obtained using six parameters extracted from the SFDI images. The preliminary results demonstrate the diagnostic potential of the SFDI method for quantitative characterization of wide-field optical properties and the spatial distribution heterogeneity of human ovarian tissue. SFDI could be an extremely robust and valuable tool for evaluation of the ovary and detection of neoplastic changes of ovarian cancer.

Fig. 1

(a) Experimental configuration of the SFDI system and (b) a photograph of the system.

Fig. 2

Scatter plot showing the comparison of the measured and expected (a) absorption and (b) scattering values of the phantoms.

Fig. 3

DC reflectance image of (a) postmenopausal normal and (d) postmenopausal malignant ovary; (b) and (e) boxed area showing the region from which corresponding H&E and (c) and (f) SR-stained histology images are acquired; arrows in (e) point to micro vessels; arrows in (f) show the degenerated collagen structure of the malignant ovarian stroma, as compared to the dense, homogeneous stroma of the normal ovary (c).

Fig. 4

Photographs and absorption and scattering maps of (a)–(c) premenopausal normal ovary; (d)–(f) postmenopausal normal ovary; and (g)–(i) and (j)–(l) postmenopausal malignant ovaries. The black dashed areas represent where the Radon transform was computed.

Fig. 5

Normalized mean Radon transform profile and corresponding (a) and (b) Gaussian fitting of absorption corresponding to Figs. 3(e) and 3(k); (c) and (d) Gaussian fitting of scattering corresponding to Figs. 3(f) and 3(l).

Fig. 6

Boxplots of (a) mean absorption; (b) absorption Gaussian fitting SD; (c) absorption Gaussian fitting error; (d) mean scattering; (e) scatter Gaussian fitting SD; and (f) scatter Gaussian fitting error.

Fig. 7

One example of testing ROC curves using two parameters and six parameters.