Studies on tissue characterization by texture analysis with co-occurrence matrix method using ultrasonography and CT imaging

Abstract

We used texture analysis with the co-occurrence matrix method to analyze ultrasonograms from normal and diseased livers, and X-ray CT images obtained from normal cases and cases of idiopathic interstitial pneumonia. Ten cases of normal, fatty, and cirrhotic livers; 10 cases of normal lungs; and 10 cases of idiopathic interstitial pneumonia, all confirmed by clinical findings, laboratory data, surgery, or biopsy, were the subjects of this study. We compared the results of texture analysis in normal and diseased livers under the same conditions of gain, focus, magnification rate, probe frequency, and depth of the region of interest. Here we discuss the relationship between Fisher ratio of texture analysis and pathological character. Although the normal and diseased liver groups did not differ significantly, the different pathological grades of fibrosis and the different size of nodules in the cirrhotic and normal liver groups did have different Fisher ratios. We compared the results of texture analysis with images obtained from normal cases and cases of idiopathic interstitial pneumonia. Significant differences between normal lungs and those with idiopathic interstitial pneumonia were also found. We thus think that texture analysis can be used to analyze ultrasonograms obtained from lesions of different pathological grades and to classify CT images as well.

Clinical application of a new method that segments the region of interest into multiple layers for RF amplitude histogram analysis in the cirrhotic liver

“…Quantitative tissue characterization methods using acoustic parameters of ultrasonography are being studied; however, clinically applicable methods remain scarce. We previously reported using frequency-dependent attenuation [13][14][15][16] and texture analysis 17,18 to quantify ultrasonic images using the commercially available ultrasonic systems. Here we use the histogram of the enveloped amplitude of the RF signals as a statistical characteristic of texture.…”

Texture analysis with a new method in which the region of interest is segmented into multiple layers for radiofrequency amplitude histogram analysis of fibrous rat livers

“…Signal-tonoise ratio (SNR), skewness, and kurtosis for the Rayleigh statistics appear at the top right Because this process assigned pixels in fibrous areas a value of 0 and those in nonfibrous areas a value of 255, the calculated values were expressed as a percentage of the total fibrous content, taken as 255. 6 The light microscopic images obtained by the line scanner were processed by a single investigator (R.T.), who knew neither the values obtained for the three primary statistics nor the method used to prepare the samples. Six images were processed for each sample.…”

“…The statistics for each sample were set so values obtained by both methods were displayed at the top right of the respective histogram (Figs. [4][5][6]. The three primary statistics of each sample were measured six times, and each mean was taken to represent the sample.…”

“…5,6 Texture analysis is performed via histogram analysis of radiofrequency (RF) signals from a segmented region of interest (ROI). 7,8 Clinical application of texture analysis to patients with diffuse liver disease enabled us to differentiate hepatic cirrhosis and normal livers quantitatively using the signal-to-noise ratio (SNR), skewness, and kurtosis as primary imaging statistics of the histogram.…”

Proposal of a parametric imaging method for quantitative diagnosis of liver fibrosis

“…Among these imaging modalities, ultrasound represents an attractive choice because it is noninvasive and easily performed. Several studies have attempted quantification of diffuse liver diseases using ultrasound [3][4][5][6][7][8][9], but these methods remain unreliable because it is difficult to express the local change in the echo signal by a simple model for analysis.…”