Summary and Future Trends - Computational Intelligence in Biomedical Engineering - page 324

Summary and Future Trends
8.1 Overview of Progress
Computational intelligence (CI) has only recently emerged as a discipline
and is becoming more effective in solving biomedical-engineering problems,
specifically those in which human intervention can be replaced by some form of
automated response. A major part of this topic has been devoted to reviewing
everyday biomedical problems that have benefited from recent CI applications.
Until now, most CI research into biomedical-engineering problems has focused
on classifying disorders based on pathological characteristics. The motivation
is to first replicate the diagnostic skill of the medical specialist and if possible
provide more reliable diagnosis. In this endeavor, two trends in the application
of CI have emerged; the first is classification of the pathology or disease and the
second is the modeling of biological systems from where the pathology arises.
The former has enjoyed cautionary acceptance within the clinical environment
because physicians easily recognize the concepts of classification and pattern
recognition. The latter is, however, less well accepted by clinicians because CI
techniques tend to present as “black box” solutions, whose success depends
on the practitioner's knowledge. It is, therefore, important to recognize that
more research is needed to bring CI to the forefront of medical technology
and to quell the remaining fears. For now, let us briefly review what has been
achieved so far to summarize the topic.
One of the most successful applications of CI has been the diagnosis of
cardiovascular diseases (CVDs). This is due to several factors, firstly, CI
applications here have focused primarily on the QRS wave complex, which is
easily measured from the recorded ECG. This wave is periodic and can occa-
sionally be corrupted by electrode noise or superposition of action potentials
from other sources. The average peak-to-peak voltage of the QRS complex
is, however, considerably larger than sources of noise thereby making signal
processing ideal for retrieving the original waveform. In addition, the heart
mechanics that produce the QRS waveform are well understood and existing
QRS-detection algorithms are suciently accurate for practical application.
Obtaining a clear and clean QRS wave is no longer a major problem as is
information loss due to preprocessing.
Cardiologists examine the QRS waveform for abnormalities and when used
in conjunction with CI techniques, physical measurements such as ST segment
325