Advances in Biomedical Informatics by Dawn E. Holmes & Lakhmi C. Jain

Advances in Biomedical Informatics by Dawn E. Holmes & Lakhmi C. Jain

Author:Dawn E. Holmes & Lakhmi C. Jain
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham


7.3.2.2 Feature Computation

In order to evaluate the hyperemia level, specialists analyze the bulbar conjunctiva searching for certain characteristics. With the aim of automating this stage of the process, we implemented the 25 image features that are depicted in Table 7.3. Some of the formulas were obtained by researching the literature [55], while others were suggested by the optometrists.Table 7.3Implemented hyperemia features

In order to compute the features, our input is only the conjunctiva region of the image. We use n and m to refer to the size of this area, and i and j to represent a position within it.

The features can be divided in two main groups: those that use only information of the vessel quantity and disposition, and those that need information regarding color. The former are named by a capital letter and the subscript v, while the latter are named by one capital letter among and an integer subscript. The features that use color information can be computed in the whole conjunctiva (labeled with I), only in the vessels (labeled with V) and only in the background of the conjunctiva (labeled with B).

In order to distinguish the vessels from the background, we used the Canny edge detection algorithm to obtain an edge image E. In the formulas, VE represents the pixels that belong to a vessel in the conjunctiva and represents the pixels that belong to the background of the conjunctiva.

We computed features in different color spaces, and their channels are represented by the corresponding letter in the formulas: L*, a* and b* for L*a*b*; R, G and B for RGB; and H, S and V for HSV.

The feature computes the value of the H channel from HSV in each pixel , but taking into account the values of the neighboring pixels in the surrounding window of size s. The feature measures the quantity of vessels in a number of image rows by applying the mask M. And finally, the feature defines a set of circumferences of radius . These circumferences will define cut points with the vessels, where the width of the latter is computed by means of an active contour algorithm [64].



Download



Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.