2.3. Image analysis

The gray level density histogram is a concise and simple summary of the statistical information included in the image. The calculation of the gray level histogram includes the individual pixels. Thus, the histogram provides first-degree statistical information about the image [12]. This information includes the mean gray level intensity, standard deviation of the histogram, median, minimum and maximum intensity values, variance values, entropy, uniformity, skewness, kurtosis, and size % lower, size % upper, size % mean (size % L, size % U, and size % M) values.

Entropy is a texture analysis parameter and known to measure parametric homogeneity within the ROI [21,22]. This is a measure of gray level variation in a histogram and is a parameter that indicates inhomogeneity (irregularity) in the intensity. When all data are the same, it is defined as zero, and the value increases as the distribution becomes irregular [22,23].

Uniformity refers to the gray level distribution, which indicates the closeness of the gray tones of the image to uniform distribution; a higher figure implies a more uniform distribution [22,23]. Skewness reflects the asymmetry of the distribution; if there are more points on the left side of the mean, the skewness is positive and negative when the opposite is true [23]. Kurtosis is a measure of the peak of distribution. If the histogram is a bell curve, kurtosis is 3, and if the histogram has a sharper peak, it is greater than 3 [23]. Size % L, size % U, and size % M depict the area between ±1 standard deviation and the histogram curve (size % M), the area above +1 standard deviation (size % U), and the area below –1 standard deviation (size % L) [24]. Images were transferred to a 27-inch iMac computer (Apple Inc., Cupertino, CA, USA). OsiriX V.4.9 imaging software (Pixmeo, Switzerland) was used to conduct the HA on the ROI.

The ROI included the entire CC region in the midsagittal section because it is best seen and demarcated there (Figure 1). Gray level intensity, standard deviation of the histogram, entropy, uniformity, skewness, kurtosis, and size % L, size % U, and size % M values were calculated using the ROI values. The entire image analysis algorithm was provided by software developed in-house with MATLAB (version R2009b; MathWorks, Natick, MA, USA).

Selected section for ROI (A) and marked ROI (B) of a patient.