Calculation of tissue attenuation with a clinical ultrasound unit

Five alternative methods for estimating tissue attenuation with a clinical ultrasound unit were evaluated.

Two homogeneous tissue mimicking phantoms (with known attenuations of 0.25 and 0.54 db/cm/MHz) and 22 human placentas were scanned. Attenuation coefficients were computed by linear regression of 1) L(f), the log spectral difference; 2) Fav (d), the average spectral frequency; 3) log[P(d)], the logarithm of the spectral power; 4) Fav (d) x log[P(d)]; and 5) (Fav(d) + log[P(d)])/2.

Each of these five heuristics provided accurate estimates for the relative attenuations of the two phantoms. For placental tissue, method 3 provided better regression fits than methods 1 and 2. Methods 4 and 5 provide the best regression fits (P < .01).

The combination of both frequency and power information in methods 4 and 5 compensates, in part, for nonlinearities introduced by tissue heterogeneity. A parameter that combines both frequency and time domain information is likely to provide more reliable estimates of attenuation than power or frequency measurements alone.