We describe our recent progress aimed at computer analysis of mammogra
ms. The overall aim is to provide the clinician with reliable quantita
tive information. We summarist a representation we have developed of t
he 'interesting' (non-adipose) tissue in a breast, then put the repres
entation to work in three ways: (i) to propose a new quantitative meas
ure to aid in diagnosing masses; (ii) to explore the possibility of re
ducing by half the radiation dose required for a mammogram; and (iii)
recalling some of the results that can be provided by differential com
pression mammography, in which mammograms are taken at two slightly di
fferent compressions. Copyright (C) 1997 Elsevier Science Ireland Ltd.