Académique Documents
Professionnel Documents
Culture Documents
I
n all areas of medical research there is a typical length, but surface area increases
common physiological theme. Com- only as the square. Accordingly, if one
plexity is the salient feature shared by species is twice the size of another, it is
all such systems; afeature that is attracting likely to be eight times heavier, but have
more and more attention in physical sys- only four times the surface area. Thus, the
tems as well. Until recently, scientists larger plants and animals must com-
have assumed that understanding such pensate for their bulk; respiration depends
systems in different contexts, or even un- on surface area for the exchange of gases,
derstanding various as does cooling by evaporation from the
physiological sys- skin, and nutrition by absorption through
tems in the same or- membranes. One way to add surface to a
ganism, would require given volume is to make the exterior more
completely different irregular, as with branches and leaves;
models. One of the another way is to hollow out the interior.
most exciting pros- The human lung, with about 300 million
pects f o r t h e new alveoli, approaches the more favorable
scaling addressed ratio of surface to volume enjoyed by our
herein is that it may evolutional ancestors, the single-celled
well provide a unify- microbes.
ing theme to many The classical scaling concepts in biol-
investigations, ogy, while of great importance, are not
which up until now capable of accounting for the irregular
have been con- surfaces and structures seen in hearts,
sidered unrelated. lungs, intestines and brains. Experiments
The swirling spiral suggest that biological processes are not
of the cochlea and the continuous, homogeneous, and regular,
finely branched struc- but rather are discontinuous, inhomoge-
ture of the bronchial neous, and irregular. Thus, a new way of
tree suggest the com- analyzing such processes is required. This
plex interrelations new perspective is that of fractal geometry
among biological and fractal statistics. Herein, we briefly
development, form, review how these approaches have
and f u n c t i o n . R e - enabled scientists to peer behind the veil
lationships that de- obscuring our understanding of a number
pend on scale can of physiological processes.
have profound im-
M
ysiology. Consider, for ex- athematical Fractals
the standard problems in A basic theme in science is that of
increases as the cube of a the invariance or symmetry of laws
0739-51 75/92/$3.0001992
40 IEEE ENGINEERING IN MEDICINE AND BIOLOGY June 1492
I
with respect to various types of with an equilibrium distribution of veloc- transformation expands or contracts the
transformations that may be performed in ities characterized by a constant tempera- observed scales in the same way at each
ordinary space and time, or even in certain ture, and copying it in all directions point.
abstract mathematical spaces. I n throughout the given volume. In this way, A more general scaling transformation
dynamics, there are the familiar sym- a disorganized system may be self-similar is x’= axP, with a and B both being con-
metries in space and time which lead to under scale changes and can be thought of stants, so that different values of x, e.g.,
conservation laws. For example, transla- as being formed by repetitious translations locations in space, are weighted differ-
tional invariance implies conservation of of a generating element. The idea of form- ently. Large x is affected more than small
linear momentum, whereas invariance ing a large set of points from a smaller x for I3 > I , and vice versa for I3 < 1. This
with respect to rotation implies conserva- generating set is used below to provide latter type of scaling was used by D’Arcy
tion of angular momentum. Using the one definition of dimension for sets of Thompson [3] in scaling anatomical struc-
known symmetries of a given system points. tures. It appears quite often in the form of
provides a framework to which the details In the 1960s, Benoit Mandelbrot began allometric growth laws in botany, as well
of structure and motion must conform. In discussing a new geometry of nature [I], as in biology. This particular kind of scal-
some cases, it is one that embraces ing has been successfully used in biology
not known initial- the irregular for over a century. However, the sig-
ly just what sym- shapes of objects nificance of the existence of such al-
metry a system such as coastlines, lometric growth laws have not always
possesses bkcause lightning bolts, been properly appreciated. This scaling
of lack of fun- New scaling cloud surfaces, often refers to a process that has “infinite”
damental infor- and molecular levels of substructure that repeat in an ever
mation about
interactions
methods may trajectories. It was
soon realized that
decreasing cascade to even smaller scales,
over which one averages to obtain the
a m o n g t h e ele-
ments of the sys-
provide a these geometrical
ideas could be ap-
allometric growth law. Processes such as
these are often described by functions that
tem. However, a
time-honored unifying theme plied in other areas,
including non-
are continuous but not differentiable.
In the late 18OOs,most mathematicians
strategy for con- traditional ones [2], felt that a continuous function must have
structing the equa- to various from outside the a derivative “almost everywhere,” which
tions of motion for physical sciences. means the derivative is singular only on a
a complex system
is t o a s s u m e a
investigations A common feature
of these objects.
set of points whose total length or measure
vanishes. However, some mathematicians
symmetry proper-
ty, and then deter-
heretofore which Mandelbrot
called ,fi-actuls, is
wondered if functions existed that were
continuous but did not have a derivative
mine the forces
required to main-
considered that their boun-
daries are so ir-
at any point (continuous everywhere but
differentiable nowhere). It is interesting to
tain this symmetry
over time. unrelated regular that it is
not easy to under-
note that mathematicians at the time were
reluctant to consider such unusual func-
This approach is stand how to apply tions as being worthy of serious research
the variational s i m p l e metrical attention. Similarly, there is some resis-
principle, and it ideas and opera- tance today to the shifting emphasis that
has been applied tions to them. has occurred in research, and is just begin-
to such quantities as the system’s entropy, Such fundamental concepts as dimension ning in the classroom, with regard to plac-
mass, and energy. This strategy for deter- and length measurement must be general- ing general nonlinear analysis in the
mining the dynamic properties of a system ized. Therefore, let us consider some of curriculum.
has been very useful in bioengineering, the metric peculiarities of a few unusual The motivation for considering such
where concepts such as efficiency have mathematical objects, which we sub- pathological functions was initiated with-
been introduced. A biosystem can be as- sequently use to describe some biomedi- in mathematics and not in the physical or
sumed to operate so as to maximize its cal systems. biological sciences [4]. After all, what
efficiency. The equations that produce The physical phenomena mentioned possible use can there be for a function
this effect are sought through a variational above seem to be without symmetry over that is so jagged that it has no tangents at
principle. many orders of magnitude. However, that all? In 1872, Karl Weierstrass (1815-
In other areas, where it might at first in itself is found to be the symmetry of 1897) gave a lecture to the Berlin Acad-
seem that symmetry is of little use, it tums self-similarity, just as in the case of the gas emy in which he presented functions that
out that there is often an unnoticed sym- in equilibrium discussed above. That is, had the aforementioned continuity and
metry framework present that determines these phenomena are self-similar under differentiation properties [ 5 ] .Thus, these
system behavior. As an example, the mo- scaling transformations. Mathematically, functions had the symmetry of self-simi-
tions of the individual molecules in an s c a l i n g transformations are multi- larity. Twenty-six years later, Ludwig
equilibrium gas in the lungs are random, plicative, as compared to the translational Boltzmann, who elucidated the micro-
and symmetry seems to be the last thing transformations mentioned above, which scopic basis of entropy, said that physi-
one would think about to describe the gas are additive. A system that is invariant cists could have invented such functions
exchange process. Considered another under time translation means that its in order to treat such things as collisions
way, however, equivalent randomness oc- properties at time t are unchanged under among molecules in gases and fluids.
curs everywhere in the gas, so there is a the replacement t + t + z, with z being any Boltzmann had a great deal of experience
statistical symmetry with respect to spa- time interval. The multiplicative linear thinking about such things as dis-
tial translation. The entire gas can be scaling transformationof the form x’ = a x , continuous changes of particle velocities
generated by taking a small element of changes the original distances through that occur in kinetic theory. He had spent
volume, one that contains many particles multiplication by the constant, a . This many years in trying to develop a micro-
I II
I
dynamics, referred to these functions as a InN (2) Another way to see this uses the dimen-
[I =
sionality of the set by writing the length of
~~
N
point (d = 0): it is atural Fractals
the fractal lung is “fractal dust.” A fractal property can be spatial, as
given in Fig. 4, The total length, in the fixed geometry of the mathe-
where the “dimen- L(z), of the set matical examples above; it can be tem-
sion” of the struc- after z generations is 213 of the length L(z poral, as in a series of data taken from a
ture corresponds to that measured from - 1 ) of the previous generation. so that system over an interval of time; and it can
casts of mammalian lungs. be exact or statistical. Forexample, in Fig.
5. we depict the time series for the beat-
imensionalities of Sets of Points to-beat interval of normal sinus rhythm of
he similarity dimension of a set of a mammalian heart. Regions of the time
points may be defined for self-similar This equation has the solution L(z) = series are magnified (Fig. 5a) to em-
sets. Such sets can be covered by trans- (2/3)‘L(O), and as z -+ the length of the
00, phasize the suggested “self-similarity” of
lating a generating element throughout the Cantor set exponentially goes to zero. the fluctuations of the data. This property
set, which is the basic idea in making a
measurement of the length of a continuous
object with a ruler. Consider a straight line
segment. Dividing the segment into N
self-similar pieces by applying a ruler of
length q , the length of the interval is then
L(q) = Nq. If L = 1, then the ruler must
have length q = l/N to exactly cover the
line.
Similarly, an area L2 can Ipe covered by
N elements, each of area I, so that L’ =
Nq2, and q = I/”’* for L = 1 . In three
dimensions, a unit cube is covered by q =
l/N1’3elementary cubes. Generalizing, a
d-dimensional object is covered by N self-
similar objects, with
I I I
simulations of the growth of dendritic
structures [ 11, 121, in which the structures
are “grown” by allowing particles to start
a large distance away from a target or seed
cluster of particles, and diffuse (random-
walk) toward the seed cluster. Such
processes are referred to as diffusion-
limited aggregation (DLA), and the math-
ematical models rely on the assumed
diffusive character of the random motion.
In steady state, the diffusion equation for
the particle number density, p, becomes
Laplace’s equation V2p = 0. This equation
is solved numerically on a grid, with the
solutions forced to satisfy certain bound-
ary conditions. (Solutions of Laplace’s
equation can have neither maxima nor
minima in a region in which the equation
holds. A field that satisfies the equation in
4. Computer simulation of a fractal lung, in which the boundary conditions a region has no sources, and therefore no
influence morphogenesis. The boundary was derived from a chest radiograph. The “lumpiness” in the region. It is interesting
model data are in good agreement with actual structural data [9]. that such a condition of smoothness is
used as part of the mathematical model
that generates irregular forms). The ag-
is more clearly revealed in the inverse on the parameters of the system. A finger gregation of particles at the seed causes
power law spectrum for this time series of frost on a window pane is a crystalline changes in the boundaries and results in
(Fig. 5b). solid that has a fractal dimension less than complicated applied mathematics
As is usual in the application of mathe- 2, but greater than 1. This dimension is a problems.
matical models to nature, natural fractals consequence of the process being con- An effective radial density for such
are more restricted than mathematical strained to unfold within a two-dimen- clusters can be defined by writing Eq. 6
ones. The ideal elements (infinite lines, sional space; if the process had unfolded for an arbitrary r < R and dividing by the
smooth planes, etc.) of Euclidean geome- in free space, the result would be falling B
Euclidean volume a r , where the positive
try are never realized in nature, and neither snow, with a fractal dimension greater integer, E, is the ordinary dimension num-
are the ideal elements of fractal geometry, than 2, but less than 3. ber of the underlying Euclidean space.
although the latter is closer to nature than To investigate the fractal structure The resulting density:
the former. The definitions of the dimen- experimentally, it is necessary to be able
sion of the mathematical fractals given to relate the results of observations to frac-
previously require that the size of the tal measures, such as dimension. One use-
elementary covering element vanish. ful dimension for this purpose is the shows that, because D < E, the fractal
These definitions of similarity dimensions cluster ,fractal dinletision, D, which is a density decreases with distance from the
are examples of a general dimension measure of how completely a cluster of origin. The cluster dimension refers to an
defined by Hausdorff, which is now en- objects fills the region where it resides, as average density or mass distribution, but
compassed by the term fractal dimension for example a collection of small spherical it does not carry information about the
[lo], applicable even to sets that may not balls in a spherical volume. The parameter shape of the cluster.
be self-similar over all ranges of space or D may be defined for a physical object Observations of DLA-type patterns have
time. Independence of scale for properties with a lower length limit, q. If a system is been reported for many phenomena in-
of systems containing mass is limited to composed of spherical physical objects of cluding dielectric breakdown, metallic
finite spatial and/or temporal intervals. radius q, then the number of objects colloid clustering, growth of metal leaves
For example, molecular diameters and placed side by side that are required to on a two-dimensional surface, and viscous
typical periods of molecular motions may cover a distance, R, is R/2q. If, for ex- fingering in porous media. The scattering
set lower limits to self-similarity, al- ample, the smallest sphere in the cluster of of light, x-rays, and neutrons has been
though those may be extreme restrictions objects has radius Ro, the number of par- used to determine the fractal dimensions
in aparticular situation. However, it is still ticles in the cluster is: of clusters by employing the usual method
possible and useful to apply the general of generalizing scattering expressions in
idea to a natural system and define its Euclidean spaces to fractional dimen-
fractal dimension. Knowing that the frac- sional spaces. Therefore, we expect the
tal dimension of an object or process is in analogy to Eq I . The equation above scattered intensity, I(k) to obey a power-
closer to unity than it is to two, for ex- defines the clirster diniension, D. If we law equation I(k) - k-b, where k is the
ample, indicates a system or process is assign the same mass to all the spherical spatial frequency (reciprocal of the
closer to being a smooth curve, with objects in the cluster, o can be considered wavelength) that lies in the range l/R <<
respect to some set of variables, than it is to be the mass density and N the total k << l/q, where R is the scale of the entire
to being a smooth plane. mass. The parameter D may then also be cluster, and q is the scale of the individual
The dimension of a naturally occurring called the mass dimension. particle. This scaling means the
fractal is a quantitative measure of a The nonlinear dynamical processes that wavelength of the radiation is very large
qualitative property of a structure that is can produce fractals are currently the sub- compared to the particle diameters, and
self-similar over some region of space or ject of much research, but they are not very small compared to the entire cluster.
interval of time. There may be more than understood in many cases. However, there For silica particles, it is found that D = 2. I ,
one region of self-similarity, depending have been recent, successful computer while light scattering from IgG-type
P
ower laws, Noise and inverse power-law spectrum is often a found for the static spatial structure of the
Fractal Time Signals manifestation of the fractal or scale-in- mammalian lung [ 161, which indicates a
The power law relating intensity of variant nature of the underlying process. lack of characteristic length. The lack of
scattered radiation to spatial frequency Measuring heart rate by listening or by regular motion of the heart upsets the con-
mentioned above, reminds us of the feeling, would lead one to describe the ventional view of normalcy of the motion
numerous power law relations in science normal heart rate as periodic or regular. of that system. That view, one of homeos-
that have the self-similarity property. For This description implies a power spectrum tasis [ 171, holds that normal physiological
example, the inverse-square force, which that consists of sharp peaks at isolated systems strive to maintain constancy in
is fundamental in gravitation and in frequencies, rather than the broad their intemal function, and therefore at-
electricity and magnetism, has no intrinsic spectrum indicated by the data of Fig. 5 . tempt to reduce fluctuations in their vari-
scale: it has the same form at all scales The real data suggest that there is no char- ables. If perturbed, such systems would
under a linear scaling transformation.
For any type of field (sound, electromag- ~ ~ . ~ _ _ _ ~ .
1 II