Clinical applications of bioelectricity - Biomedical Engineering Desk Reference - page 372

four indices in order to extract more information from
the original impedance spectra. However, the indices are
still useful for quantification of various aspects of re-
sponses to treatment or test substances, an example of
which is given by Emtestam et al. (2007).
The finding of correlation between tissue structures
(as seen in the microscope) with impedance properties
(Nicander et al., 1996) triggered the idea of a potential
diagnostic decision support tool intended to assist the
doctor in a clinical environment. This idea is not new, for
example, Fricke and Morse found a difference in capacity
of tumors of the breast compared to normal breast tissue
already in 1926. However, this finding is completely
unspeciflc. Almost any tissue alteration can be detected
by electrical impedance, but in order to be clinically
useful, the method has to be able to differentiate benign
alterations from malignant alterations, or be able to dis-
tinguish one disease from another in order to select
a specific and adequate therapy. In other words: there is
a difference between statistical significance and clinical
significance. A p -value < 0.001 may sound convincing in
a statistical comparison, but may mean nothing in the
clinic unless both sensitivity and specificity are good
enough in the intended application. Thus, more in-
formation had to be extracted from the impedance
spectra, and both the electrical impedance indices, which
were sufficient for characterizing elicited skin reactions,
and classical Cole-style models, were found inadequate
to distinguish various disorders, according to the Ollmar
group.
In skin testing, the central area of the volar forearm is
very popular, mainly because of ease of access. It is also
considered very homogeneous and stable, compared to
other areas of the human body, and in extrapolation of this
belief there have been studies without randomization of
test sites within the volar forearm region. In search of
suitable statistical tools to enhance discrimination power,
this belief was challenged by ˚ berg et al. (2002), who used
linear projection methods (in this case PARAFAC) to
extract clinical information from the impedance spectra.
The results showed systematic differences within the test
area, which may not be important in comparison of strong
reactions to, for example, detergents, but would have
devastating impact on the outcome of a comparison of
cosmetic preparations, where only small differences
would be expected on normal skin, unless randomization
of test sites is built into the study protocol.
It is known that diabetics are prone to develop ulcers
difficult or impossible to heal, and therefore some dif-
ference in the skin properties might be present even
when no clinical signs of ulceration, not even slight ery-
thema, are present. Lindholm-Sethson et al. (1998)
found such a difference in the skin between diagnosed
diabetics without clinical signs in comparison with
a healthy control group, but the difference was hidden

behind more pronounced factors, such as age and sex,
which were identified using PCA (principal component
analysis). It seems that multivariate methods, such as
PCA, make better use of the information inherent in
electrical impedance spectra than simple indices or
lumped parameter models, and that the extracted in-
formation sometimes reflects clinically interesting phys-
iological or pathological conditions. In certain cases it
might be possible to establish a strong correlation be-
tween specific principal components and well-defined
physiological or pathological conditions.
To date, most skin studies involving electrical imped-
ance are based on pure surface electrodes. Due to the
extreme heterogeneity of the skin, such measurements
(at least at low frequencies, as demonstrated in a simula-
tion study by Martinsen et al., 1999), reflect mainly the
conditions in the stratum corneum and the integrity of its
inherent skin barrier. However, only living cells get irri-
tated or sick, and if important information about an al-
teration in the living strata of skin resides in a relatively
low frequency range, such information will be over-
shadowed or diluted by the intact stratum corneum . The
dilution factor might be 1:100 or even 1:1000, and
strongly frequency dependent! In several skin reactions or
diseases, the skin barrier will be more or less destroyed by
the chemical assault from the outside (fast event), or
by sloppy maintenance of the stratum corneum provided
by the damaged or sick living epidermis from the inside
(slow event), and then surface electrodes would be suf-
ficient. If not, something has to be done about the stratum
corneum, and a number of methods have been tried, such
as aggressive electrode gels (which will add to tissue
damage), peeling creams or simply grinding or stripping
off the outermost layer. The effect of tape stripping to
various degrees of damage on skin impedance is illustrated
in a topic by Ollmar and Nicander (2005), and shows the
dramatic overshadowing power of the intact stratum
corneum on properties residing in the living strata
of skin.
A new approach toward solving the stratum corneum
dilemma has been presented by Griss et al. (2001), using
electrodes furnished with micromachined conductive
spikes, thin enough not to leave any damage after removal
and short enough to only short circuit the stratum
corneum without reaching blood vessels and nerve end-
ings in deeper strata. This concept, originally developed
as an improvement of ECG and EEG electrodes, has
been further developed by the Ollmar group ( ˚ berg
et al., 2003a) to an electrode system intended to facili-
tate skin cancer detection even when the stratum
corneum happens to be intact on top of any skin tumor.
For a clinical diagnostic decision support tool, it is not
enough to detect an alteration; it must also reliably dis-
tinguish disease from other alterations. Key concepts in
this context is sensitivity and specificity, as well as