The paper considers aspects of work carried out on projects in medical diag
nostics and engineering fault detection and control, and shows the developm
ent of some ideas in signal processing and diagnostic techniques; in partic
ular regarding models, estimators and decision schemes. The fact that analy
sis of human biological systems is restricted by powerful ethical and legal
constraints would seem to lead to fundamental differences of approach to m
edical and engineering problems. Tn the reseal-ch considered this did not s
eem to be a dominant issue. The use of system simulators as reference stand
ards, and blackboard schemes for knowledge integration, has shown to advant
age in both fields. Methods for increasing the usefulness of limited data h
ave been successfully applied to decreasing the cost associated with sensor
s for a given standard of diagnostic reliability. The science of engineerin
g control and diagnostics has benefited front the use of observers based on
state-space descriptions of the system. An observer provides an alternativ
e reference model which has been used to generate residuals when faults occ
ur. A new technique employing sliding-mode observers to recreate fault sign
als is introduced. This idea is extended for use in a medical context.