Abstract
Previous methods of interpreting ultrasonic liver scans are discussed. A system is described in which the performance was calibrated in terms of measurable quantities; this avoided any possibility of distortion being caused by observer bias. The results of the examinations of 30 individuals with normal livers are reported: 30 scans of each individual were analysed using a computer. The statistical basis of the regression analysis is discussed. Echo amplitude (y) decreases exponentially with range (x), and the most satisfactory treatment is to take the mean of several estimates of amplitude, after logarithmic transformation (mean ln y). There is a very good linear relationship between ln y and x, and the echo amplitude is approximately log normally distributed about this line. The apparent attenuation in the pooled group is 1.76 dB cm-1. The correlations between age, somatotype and surface area with the ultrasonic data are discussed in relation to models of the liver. The analysis enables the range of normality to be specified in the clinical situation.