Tissue Characterization and Detection of Dysplasia Using Scattered Light

Abstract

In this paper, the structural parameters of dysplasia formation in the epithelial tissue are estimated using a stochastic decomposition algorithm (SDM) by means of scattered light. We extract texture parameters obtained from the decomposition that capture the signature of dysplasia formation. These parameters include the number and mean energy of coherent scatterers; deviation from Rayleigh scattering; average energy of diffuse scatterers; and normalized correlation coefficient. The tests are performed on simulations, and tissue-mimicking phantom data. The simulations are based on the light scattered from the cells with varying parameters such as, index of refraction, number of cells, and size of cells. The obtained results demonstrate the proof-of-concept in being able to differentiate between tissue structures that give rise to changes in cell morphology as well as other physical properties such as change in index of refraction. Fusing all the estimated parameter set together results in the differentiation performance (Az value) up to 1(perfect detection) for simulated data, and Az>0.927 for the phantom data.

On‐the‐fly detection of changes on and below the surface in epithelium mucosal tissue architecture from scattered light

“…The applicability of the model for differentiating different tissue characteristics using simulations, phantom data and on a limited preliminary in vitro animal experiment for tracking mucosal tissue inflammation over time has been verified and tested in our previous reported work with efficacy given by the area A z under the Receiver Operating Characteristics (ROC) curve by fusing all the estimated parameter set together. Very high A z values were reported (A z value of 1 for simulated data (perfect detector), A z > 0.927 for the phantom data, and A z values of 0.859, 0.983, and 0.999 for differentiation between pairs of various levels of inflammation for the tissue data) [26][27][28]. For our onthe-fly segmentation scheme, tests are performed on both tissue mimicking phantoms and real tissue data (in vitro).…”

“…In this paper, our goal is to capture on-the-fly where changes in the tissue mucosal structures occur with no prior knowledge of their structure(s) by introducing a segmentation algorithm based on the textural model parameters of the reflected data obtained from the Stochastic Decomposition Method (SDM) [26][27][28]. This manuscript describes a variation of light scattering spectroscopy similar to the work originally introduced in [29], but employs a somewhat different data processing scheme.…”

“…The SDM model [26][27][28] uses reflected light at one specific wavelength l. SDM method is consistent Figure 1 Layered structure and principal histological components of normal, dysplastic, and adenomatous colon tissue [36]. Figure 2 Schematic representation of colon architecture [16].…”

“…This is done by taking the normalized PSD aŝ S Sðf ; lÞ ¼ Sðf ; lÞ s 2 ðlÞ ð5Þ s(l) conveys information about the energy of the coherent scatterers for the diffuse component, whereasŜ Sðf ; lÞ conveys information about the spatial structural distribution of the tissue (morphological texture). In addition to s 2 ðlÞ andŜ Sðf ; lÞ, we also consider the Kolmogorov-Smirnov (KS) distance D(l) [26][27][28]. KS distance is the distance between two cumulative probability distribution functions; it is the distance used to test the hypothesis that the given data follows a certain nominal distribution [56].…”

“…Hence by definition, the lower the value of D(l) he closer the distribution of diffuse component to Gaussian distribution which means higher density of diffuse scatterers. The connection between the changes in tissue structure and the changes in the SDM parameters are discussed in detail in [26][27][28], where every parameter relationship to the tissue structure is laid down.…”