曹 景文, 伊藤 貴司, 小西 辰男

Jing-Wen TSAO, Takashi ITOH, Tatsuo KONISHI

アロカ株式会社

Aloka Co., Ltd.

キーワード : Histogram, B-mode image, Tissue characterization

We have derived two closed form formulas to express the histograms or the probability density functions (pdf) of amplitude demodulated (AM) image and its log-compression version (AML), but took both effects of frequency dependent attenuation (FDA) and backscattering (FDB) into account. Thus, with the aid of these two formulas and earring out a series of computer evaluations will enable us feasibly to address specifically the effects of physical and equipment parameters on the pdf, and hence the statistics of AM and AML images.
Parameters such as coefficient (β) of FDA, power law (n) of FDB, site (z₀) and range (z) of ROI, system gain (C), center frequency (f₀) and bandwidth (σ₀) of transducer etc., were evaluated and the results indicate that, (1) apparent parametric variations in statistics are shown especially in AML image, but to AM image, the three statistics SNR, skewness and kurtosis appear to be insensitive to z₀, C and n, this suggests that these three statistics could serve as unbiased parameters for histograms analysis, (2) due to the presence of attenuation the AM pdf loses its consistency as the Rayleigh pdf, and its SNR becomes a monotonic decreasing function of β, (3) keeping the mean of gray level in ROI constant by gain control in every observations, then statistics of either AM or AML image will be independent of z₀ and n, (4) the lower bound required to avoid the error induced by insufficient quantized gray-scales was observed for statistic entropy>3.5, and (5) in general, in a attempt to analyze histograms in a more objective fashion, it is wise to obtain histograms under the same measurement conditions.