PURPOSE: We used texture analysis in conjunction with an alternative method of analyzing the amplitude histogram using a radiofrequency (RF) signal to differentiate ultrasonograms of normal and cirrhotic livers. This method segments the region of interest (ROI) into multiple layers (sub-ROIs). In each sub-ROI of a homogeneous medium, the histogram of enveloped-amplitude of RF backscattered echoes resembles a Rayleigh distribution. Theoretically, the values of the signal-to-noise ratio (SNR), skewness, and kurtosis for Rayleigh statistics are constant and independent of the mean scattering intensity, which is contributed by such undesirable effects as tissue attenuation, beam diffraction, and incident waveforms. These values, which averaged overall sub-ROI, should provide an unbiased estimator. METHODS: We studied 36 normal livers and 28 cirrhotic livers, all confirmed by clinical findings including laboratory and pathology data; the SNR, skewness, and kurtosis values of the disease groups were compared. At the same time, these values were estimated using the conventional method, which did not segment the ROI into multiple sub-ROIs. The unpaired t-test was used to determine statistical significance. RESULTS: With the new method, all values obtained from cirrhotic livers differed significantly from those obtained from normal livers, and the standard deviation of these values was smaller than those obtained using the conventional method. CONCLUSIONS: These results suggest that the new method can be used to diagnose the cirrhotic liver objectively.
PURPOSE: We used texture analysis in conjunction with an alternative method of analyzing the amplitude histogram using a radiofrequency (RF) signal to differentiate ultrasonograms of normal and cirrhotic livers. This method segments the region of interest (ROI) into multiple layers (sub-ROIs). In each sub-ROI of a homogeneous medium, the histogram of enveloped-amplitude of RF backscattered echoes resembles a Rayleigh distribution. Theoretically, the values of the signal-to-noise ratio (SNR), skewness, and kurtosis for Rayleigh statistics are constant and independent of the mean scattering intensity, which is contributed by such undesirable effects as tissue attenuation, beam diffraction, and incident waveforms. These values, which averaged overall sub-ROI, should provide an unbiased estimator. METHODS: We studied 36 normal livers and 28 cirrhotic livers, all confirmed by clinical findings including laboratory and pathology data; the SNR, skewness, and kurtosis values of the disease groups were compared. At the same time, these values were estimated using the conventional method, which did not segment the ROI into multiple sub-ROIs. The unpaired t-test was used to determine statistical significance. RESULTS: With the new method, all values obtained from cirrhotic livers differed significantly from those obtained from normal livers, and the standard deviation of these values was smaller than those obtained using the conventional method. CONCLUSIONS: These results suggest that the new method can be used to diagnose the cirrhotic liver objectively.