[Quantitative image feature analysis of echographic textures using the normalized first moment of the power spectra:...

In order to evaluate the coarseness of echographic textures objectively, ultrasonic B-mode data from the thyroid glands in 10 normal subjects and 10 patients with chronic thyroiditis were digitized. The normalized first moment of the power spectra in the digitized matrix data that were obtained was calculated by computer processing of a Fourier transformation algorithm. On phantom study, when the location of the region to be analyzed or the gray scale level of the images was changed, the normalized first moment of the power spectra in the low frequency region of less than 1.0 cycle/mm varied significantly, and noise components in the high frequency region of more than 2.5 cycles/mm exceeded the signal components. Thus we attempted to calculate the normalized first moment of the power spectra from 1.0 to 2.5 cycles/mm in human thyroid glands. The values of the normalized first moment of the power spectra in the thyroid glands with normal (n = 20) and chronic thyroiditis (n = 20) were 0.989 +/- 0.036 and 1.065 +/- 0.063, respectively. These values were significantly different between the two groups (p < 0.01). It is concluded that the normalized first moment of the power spectra calculated within the limited frequency is useful in characterizing the echographic texture of chronic thyroiditis.