Available online at www.sciencedirect.

com

ScienceDirect
Procedia Environmental Sciences 32 (2016) 344 – 351

International Conference – Environment at a Crossroads: SMART approaches for a sustainable

future

Using fractal analysis in modeling trends in the national economy

Ion C. Andronache, Daniel Peptenatu*, Ana-Maria Ciobotaru, Andreea Karina Gruia,
Nina Margareta Gropoèilă
University of Bucharest, Research Center for Integrated Analysis and Territorial Management, 2-14 Regina Elisabeta, Bucharest, 030018,
Romania

Abstract

The study aims to develop a methodology for analyzing the trend of concentration of economic activities in the main core of
national polycentric network. A spatial projection has been realized in ArcGIS platform, for the period 2000-2012, containing the
number of companies, number of employees, turnover and profit. The original images were converted to images created 8-bit
grayscale (using open-source software ImageJ) are then analyzed by Higuchi 1D Fractal Dimension algorithm using open-source
software IQM. Higuchi fractal dimension provided us very useful information about more or less chaotic disposed in space by
the number of companies, the profit, the number of employees and the turnover. From data analysis we observed a increase of
Higuchi dimension with the growing economic influence of the main centers of development for all the four parameters. We
observed that Higuchi analysis could help in identifying of the increasing of the influence of the main centers on the
neighbouring territories, in periods of economic development, or decreasing of influence in the periods of the economic crises.
For testing and confirmation of the veracity of Higuchi analysis the results obtained were confirmed by gray box-counting
analysis using Fractal Analysis System and Fraclac/ ImageJ. Hence, the Higuchi’s fractal dimension is a very good, fast and
versatile method for modeling evolution of concentration trends in the national economy.

©
© 2016
2016TheTheAuthors. Published by by
Authors.Published Elsevier B.V B.V.
Elsevier This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of ECOSMART 2015.
Peer-review under responsibility of the organizing committee of ECOSMART 2015
Keywords:Fractal dimension, concentration trends, complexity, territorial management;

1. Introduction

Spectacular dynamics of emerging economies require specific approaches, more complex to highlight structural
changes and evolution trends [1, 2, 3, 4] .The complex analysis in territorial profile of economic dynamics can allow
the identification of trends of clustering at development poles levels of polycentric networks. The development of

* Corresponding author. Tel.: +400788154399.

E-mail address:peptenatu@yahoo.fr

1878-0296 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of ECOSMART 2015
doi:10.1016/j.proenv.2016.03.040
 Ion C. Andronache et al. / Procedia Environmental Sciences 32 (2016) 344 – 351 345

these economic clusters contributing in developing of polarization capacity of settlements; the changes in the
entrepreneurship sector are felt at the settlements levels from subordinate network [5, 6, 7]. Also, the analyze of
concentration trends has an important role in the studies of human activities pressure on the natural subsystem [8, 9,
10]. Fractal dimension is a measure of complexity, providing a measure of the degree an object "fills" space. The
fractal dimension quantifies the degree of irregularity and fragmentation of a geometrical structure or of an object in
nature. The degree of irregularity is derived from the properties of the fractal object which is a fragmented or broken
geometric figure that if divided into parts, each part is approximately a smaller copy of the whole. In nature these
copies are not exact representations but are nevertheless scale invariant [11].
Fractal methods can be very useful for quantitative assessment and classification of images. For these tasks,
determining the fractal dimensions of 2D digital images has been very successful in recent years [12, 13, 14, 15, 16].
The methods involved include the well known Box counting method or the Minkowski dilation method. It is also
possible to use gray value statistics, differential box counting, a variation method, a blanket method or frequency
analysis. In situation of gray images with spatial projection of turnover, companies number, profit, employees
number, could be use 2D and but also 1D fractal analysis techniques. The purpose of this article is to develop a
robust research method addressing the analyze concentration trends in the economy, based on fractal analysis that
can use in sustainable development and management. To be applicable a method should preferably be really simple
and easily understandable to non-specialists in the field. We used in this paper a very simple and fast method (1D
Higuchi dimension) which can be adapted very easily to actual problems and which can be used in multidisciplinary
applications.

2. Material and methods

We analyzed 52 images, annual spatializations of turnover, companies number, profit and employees number,
for the period 2000-2012, to the whole country (fig. 1-4). The images were obtained by GIS modeling. Cartographic
models were made in the ArcGIS platform, where each territorial administrative unit was assigned values of
turnover. For each individual year, values were represented respecting the principle of selection the same number of
intervals and the same minimum and maximum values for each of them to be able to observe differences arising
from year to year.

Fig.1. The spatial projection of concentration trends of turnover (lei)

346 Ion C. Andronache et al. / Procedia Environmental Sciences 32 (2016) 344 – 351

Fig. 2. The spatial projection of concentration trends of companies number

Fig. 3. The spatial projection of concentration trends of profit (lei)

Fig. 4. The spatial projection of concentration trends of employees number

 Ion C. Andronache et al. / Procedia Environmental Sciences 32 (2016) 344 – 351 347

We use for the determination of 1D Higuchi Dimension software IQM (Interactive Quantitative Morphology).
IQM is an image and signal analysis software framework written in Java, developed by Helmut Ahammer (Medical
University of Graz, Austria). The Higuchi dimension is a relatively fast algorithm for calculating the fractal
dimension of 1D images. The length of the attractor (line) is determined at increasing resolution [17]. We have
decided for rows or columns method. Using this method only global characteristics of the object under investigation
are examined. The gray values (2D images) projected onto 1D signals or several image rows, columns were
extracted in order to gather a batch of 1D signals. Extraction of rows and columns can imply a little loss of
information, and the fractal dimension of the whole image can be calculated more precisely. This projection
resembles the summing up of gray values, and two 1D signals are constructed and determinate Higuchi’s fractal
dimension of the obtained landscapes.
Hence, an extracted 1D signal of an image is an intersection of the gray value surface with a two dimensional
plane and therefore, the intersection theorem for fractals can be applied (Eq. 1):

‫ܦ‬ଵ஽ ൒  ‫ܦ‬ଶ஽ ൅  ‫ܦ‬௣௟௔௡௘ Ȃ ‫ܧ‬ (1)

with D1D the fractal dimension of the 1D signal, D2D the fractal dimension of the gray value surface in a three
dimensional Euclidian space E~3, and a plane with DPlane ~214.
The Higuchi fractal dimension was estimated by calculating the overall mean of the rows and columns (Eq. 2):

ଵ ଵ ଵ
ୌ ൌ  ଶ ቂ୑ σ୑ ୒
୨ୀଵ ୌ ൫”୨ ൯ ൅  ୒ σ୧ ୌ ሺ…୧ ሻቃ (2)

Where: ୌ = 1D Higuchi dimension; DHሺ‫݆ݎ‬ሻ= Higuchi dimension of every row; DHሺܿ݅ሻ= Higuchi dimension
of every column, ‫ = ܯ‬dimensions value of DHሺ‫݆ݎ‬ሻand ܰ = dimensions value of DHሺܿ݅ሻ
The Higuchi method estimates the fractal dimension by a linear regression of a double logarithmic plot [18, 19].
The calculation of two directionally dependent fractal dimensions allows the distinguishing of directional
dependencies, which cannot be resolved by classical 2D methods. In addition to this advantage, it is always possible
to average the two different values and get a value identical to the classical methods. As Higuchi Dimension is
bigger the more increase the influence of the polarization centers over the neighbouring villages and small towns.
To check the accuracy of the method we realized fractal analyses using grayscale images. We choose Box-
Counting implemented in FracLac / ImageJ and Fractal Analysis System for Windows. ImageJ is an open source
image processing software package, developed by Wayne Rasband (National Institutes of Health, Bethesda,
Maryland, USA) and designed for scientific multidimensional image analysis.
Fractal Analysis System for Windows is an software developed by Hiroyuki Sasaki (National Institute of
Livestock and Grassland Science, Tsukuba, Japan) and can generate % image of RGB and thinning image, extract
red or green area, calculate fractal dimension and coverage, from color image, gray scale, and binary, 3D sliced
(layer). As Gray Box-counting fractal dimension is bigger than 2, the more the degree of complexity increases,
indicating a spatial evolution of concentration centers and vice versa.
Box counting consists of determining the number of N (ε) cells required to cover the structure to be measured,
depending on the size ε of these cells. For different values of the cell size ε, cells that cover the figure are counted,
and then they are represented in logarithmic coordinates log N = f (log ε). Finally, the slope of a linear regression is
an estimate of the fractal dimension. The mathematical expression is (Eq. 3):

୪୭୥ ேሺఌሻ
‫ܦ‬଴ ൌ Ž‹ఌ՜଴ ቆ భ ቇ (3)
୪୭୥
ഄ

whereD0 is the box-counting fractal dimension, ε = the side length of the box, and N (ε) = the number of
contiguous and non-overlapping boxes of side ε required to cover the area of the object [20, 21].
As the zero limits cannot be applied to digital images, D0 was estimated by means of the equation (Eq. 4):

‫ܦ‬଴  ൌ ݀ (4)
348 Ion C. Andronache et al. / Procedia Environmental Sciences 32 (2016) 344 – 351

where d = the slope of the graph log[N(ε)] against log(1/ ε)21.

Correlation between 1D Higuchi dimension and Gray Box-counting using FracLac/ImageJ and Fractal Analysis
System were statistically analyzed using regression analysis (Pearson Correlation). Pearson Correlation (R) refers to
the interdependence or co-relationship of variables. In the context of regression examples, correlation reflects the
closeness of the linear relationship between X (1D Higuchi dimension) and Y (Gray Box-counting using
FracLac/ImageJ and Fractal Analysis System). R lies between -1 and 1 with:
R = 0 is no linear correlationͲ ൏ ܴ ൏ ͲǤʹ = very weak
ͲǤʹ ൏ ܴ ൏ ͲǤͶ = weak
ͲǤͶ ൏ ܴ ൏ ͲǤ͸ = moderate
ͲǤ͸ ൏ ܴ ൏ ͲǤͺ = strong
ͲǤͺ ൏ ܴ ൏ ͳ = very strong
R = 1 is perfect positive (slope up from bottom left to top right) linear correlation
R = -1 is perfect negative (slope down from top left to bottom right) linear correlation.

3. Results and discussion

The dimension values of distinct images were examined according to each of the individual methods. Firstly,
projection, extraction of rows, columns was carried out to get 1D signals for the calculation of the Higuchi
dimension. The slopes of the linear regressions of double logarithmic plots were determined, and the estimated
values of the fractal dimensions were calculated by linear regressions.
After the determination of Higuchi dimension resulted the following values (Table 1):
Table 1. Higuchi Dimension of turnover, companies number, profit and employees number (2000-2012)

Year Turnover Companies number Profit Employees number

2000 1.3449 1.3243 1.3455 1.3381
2001 1.3681 1.3371 1.3708 1.3407
2002 1.3786 1.3396 1.3804 1.3406
2003 1.3894 1.3474 1.3897 1.3414
2004 1.3984 1.3549 1.3909 1.3475
2005 1.4073 1.3616 1.3991 1.3481
2006 1.4093 1.3701 1.3870 1.3476
2007 1.4149 1.3717 1.3871 1.3590
2008 1.4121 1.3581 1.4040 1.3497
2009 1.4235 1.3619 1.4101 1.3663
2010 1.4236 1.3624 1.3863 1.3646
2011 1.4277 1.3637 1.4044 1.3658
2012 1.4231 1.3594 1.4009 1.3646

From data analysis is observing a increase of Higuchi dimension with the growing economic influence of the
main development centers (centers of development) for the four parameters: turnover, companies number, their
profit and employees number. We have identified the existence of several centers of concentration (influence): the
most important are: Bucharest, Ploieèti – Braèov, Alba-Iulia-Sibiu-Deva, Timièoara-Arad, Constanìa.
During the economic crisis, the influence of the polarization centers decreased, this is quantified by Higuchi
Dimension, by the decreasing of its value in 2008 (for turnover and companies number) and in 2006 and 2010 for
profit and employees number. In all four cases in 2012 compared to 2011 was achieved a slight decline.
For comparison, the images were used to calculate gray fractal dimension (using FracLac / ImageJ and Fractal
Analysis System for Windows) (fig. 5-6).
 Ion C. Andronache et al. / Procedia Environmental Sciences 32 (2016) 344 – 351 349

2.7 2.05

Fractal Dimension
Fractal Dimension

2.03
2.65
2.01
2.6
1.99

2.55 1.97

Turnover Companies number

Turnover Companies number
Profit Employees number Profit Employees number

Fig. 5 Gray Box-Counting Fractal Dimension (FracLac / ImageJ) for Fig. 6. Gray Box-Counting Fractal Dimension (Fractal Analysis
turnover, companies number, profit and employees number System) for spatializations turnover, companies number, profit and
employees number

In mostly, the analyses gray box-counting confirm Higuchi analyses. In both complementary analyses, discern
the trend of increasing influence of the main components of development, with small gaps during the economic
crises, when the spatial influence is less in intensity. The Gray analyses, starting from another analyses algorithm
(box-counting 2D) caught for the both analyses, restrictions spatial influence for turnovers in 2009, for numbers of
companies in 2009, 2010, 2012, for profit in 2009, 2011, 2012, and for the employees, 2008 and 2012.
The correlations coefficients between 1D Higuchi dimension and Gray Box-Counting FracLac/ImageJ and
Fractal Analysis System were very strong (R=0.8395 and R=0.8481). A very strong correlation was between the two
determinations of gray box-counting, R=0.9527). The Pearson coefficients of correlations very strong indicate that
the methodology chosen reflects the mode of manifestation in spatial of changes in economic parameteres and also
the global impact on emerging economies made by polycentric bands around big cities. To highlight the mode in
which the real evolution was manifested in space (in absolute values, in total amount nationwide) for the 4 economic
parameters analized, we realized correlative analyses between values 1D Higuchi dimension and absolute value (fig.
7-10). These analyses indicated that in general the growth of absolute values were reflected in the increase of fractal
dimension (2001, 2002, 2004, 2005, 2007) when polycentric development trained by itself a catalyst effect for the
development of emerging areas (spatial evidenced). This also applies when there is a decrease in the absolute values
of economic parameters considered when braking and economic decline of polycentric bands affected emerging
economies.
In specific crisis years 2006, 2008 to 2011 occurred differences between the absolute values of the analyzed
parameters and their spatial effects, highlighted by Higuchi 1D fractal dimension [spatial effects appear before like
a form of prediction (when first there is a decrease in emerging areas which slowdown polycentric development,
thereby decreasing their positive impact) or after the shock manifestation at polycentric level, spatial effect is
delayed, the shock is extending slower to emerging economies).
The turnover decrease from 2003 did not have a spatial meaning (the shock wasn`t felt from polycenters towards
emerging areas).

1.20E+12 1.45
80 1.38
1.00E+12
60 1.36
8.00E+11 1.4
6.00E+11 40 1.34
4.00E+11 1.35 20 1.32
2.00E+11
0 1.3
0.00E+00 1.3
2001

2004
2000

2002
2003

2005
2006
2007
2008
2009
2010
2011
2012
2000

2009
2001
2002
2003
2004
2005
2006
2007
2008

2010
2011
2012

Absolute value Higuchi 1D Absolute value Higuchi 1D

Fig. 7. Correlation between 1D Higuchi dimension and absolute value Fig. 8. Correlation between 1D Higuchi dimension and absolute
for Turnover value for Companies number
350 Ion C. Andronache et al. / Procedia Environmental Sciences 32 (2016) 344 – 351

The analyse for turnover (fig. 7) indicated that decreasing of turnover from 2009 is predicted by the decrease of
1D Higuchi dimension from 2008 and is felt spatial only in 2010 (year when were operated salary cuts with 25% for
state employees, reducing power of consumption). Pearson Correlation Ceoficient between 1D Higuchi dimension
and absolute value is very strong (R=0.9123) indicating a strong relation between development and decline of
polycentric areas and its emerging economies.
The analyse for companies number (fig. 8) confirms the capacity of anticipation of 1D Higuchi dimension, the
decrease of its value in 2008 materialized in decrease of companies number from 2009-2010 [years of deep
economic crisis, when due to tougher fiscal policies (flat tax) many companies have bankrupted]. Pearson
Correlation Ceoficient between 1D Higuchi dimension and absolute value is very strong (R=0.8844) indicating a
strong relation between development and decline of polycentric areas and its emerging economies, however
disrupted by the shock of bankruptcies chain during the economic crisis.
The capacity of predictability of 1D Higuchi dimension is evidenced also in the profit evolution (fig. 9) when
discreasing of fractal dimension from 2006 and 2010 is reflected in profit decrease nationwide from 2008 and 2011.
Pearson Correlation Ceoficient between 1D Higuchi dimension and absolute value is strong (R=0.7024) indicating a
strong relation between development and decline of polycentric areas and its emerging economies. It feels more
acutely the negative effect of the economic decrease during the crisis, when were operated drastic salary cuts, flat
tax, VAT increase which has led for that period at profit decrease and severe shock from polycenters towards
emerging economies, which have not benefit for the increasingly growing demand from polycenters.

8E+10 1.45 600 1.37

6E+10 1.36
1.4 400
x 10000

1.35
4E+10
1.35 1.34
200
2E+10
1.33
0 1.3 0 1.32
2008
2009
2010
2000
2001
2002
2003
2004
2005
2006
2007

2011
2012

2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Absolute value Higuchi 1D
Absolute value Higuchi 1D

Fig. 9. Correlation between 1D Higuchi dimension and absolute value Fig. 10. Correlation between 1D Higuchi dimension and absolute value
for Profit for Employees number

Regarding the number of employees (fig. 10) we observed that the spatial effects are felt quickly (because of
higher population mobility than turnover, number of companies and profit), such that decreasing or increasing of
number of employees cause the decreasing or increasing of 1D Higuchi fractal dimension. However as Pearson
Coefficient of correlation in the profit situation is only moderately (R = 0.5095), indicating that for at least half of
values, chaotic effects occurred. The best values of Pearson coefficients of correlations indicate that the
methodology reflect the way of manifestation in spatial of changes in economic parameters and also the overall
impact on emerging economies made by polycenters around big cities.
Being a global fractal analysis and not a local one, it does not provide information on the increase or decrease of
the big centers influence of polarity on small towns and villages from its proximity. Higuchi 1D analysis only
provides information about the impact of polarization centers in their entirety on the whole territory of Romania.
We consider useful the method because it offers a solid starting point, well argued for further research, which may
be based on an analysis Higuchi 2D dimension, Sholl analysis or local fractal dimension, like cluster fractal
dimension or mass dimension fractal analysis .

4. Conclusion

Fractal dimension of landscapes obtained from surface images does change with the surface properties. The
smoother is a surface, i.e. the smaller are its unevenness at any particular scale, the greater is fractal dimension of
any landscape obtained from an image of this surface at given magnification. If a surface shows anisotropic
roughness properties (texture) then fractal dimensions of its horizontal and vertical landscapes differ from one
 Ion C. Andronache et al. / Procedia Environmental Sciences 32 (2016) 344 – 351 351

another. Fractal analysis of spatializations of economic parameters helps in differentiating and modeling
concentration trends in the national economy. Has been identified a increasing influence of the main centers on the
neighbouring territories and has been highlighted periods of restriction of influence in the time of the economic
crise.
The fractal dimensions of objects in a digital image have been investigated by classical 2D methods, such as Box
counting, for a long time. Despite providing many reliable results, these methods have several restrictions, such as
direction independence and the impossibility to restrict calculations to regions of interests. To overcome these
limitations, this study proposes the transformation of 2D image data to 1D data series and the application of time
series analyzing methods. An advantage of the 1D analysis (like in situation of the gray box-counting dimension) is
that it investigated directly the grey value surfaces without extensive segmentation and binarization of the images,
may change the images. Hence, the Higuchi’s fractal dimension, is a very good, fast and versatile method for
modeling evolution of concentration trends in the national economy.

Acknowledgements

The authors give thanks to Helmut Ahammer (Medical University of Graz, Austria) for IQM 3.2 software and to
Hiroyuki Sasaki (National Institute of Livestock and Grassland Science, Tsukuba, Japan) for Fractal Analysis
System 3.4.7. This work was supported by a grant of the Romanian National Authority for Scientific Research and
Innovation, CNCS - UEFISCDI, project number PN-II-RU-TE-2014-4-0835.

References

1.Carree MA,Thurik AR. The impact of entrepreneurship on economic growth, in DB Audretsch& ZJ Acs (Eds), Handbook of Entrepreneurship
Research. Boston/Dordrecht: Kluwer Academic Publishers; 2003, p. 437-471.
2.Evans DS, Leighton LS. Some empirical aspects of entrepreneurship. AmericanEconomic Review 1989, 79 (3):519-535.
3.Rotefoss B, Kolvereid L. Aspiring, nascent and fledgling entrepreneurs: an investigation of the business start-up process. Entrepreneurship &
Regional Development 2005, 17 (2): 109-127.
4. Grilo I, Thurik, AR. Determinants of entrepreneurial engagement levels in Europe and the US. Industrial and Corporate Change 2008, 17
(6):1113-1145.
5.Peptenatu D, Draghici C, Merciu C, 2012a.Characteristics of entrepreneurial profile in some emergent territorial structures in Romania, Actual
Problems Of Economics 2012a, 138: 448-458.
6.Peptenatu D, Pintilii R.D, Draghici C, Merciu C, Mateescu R.D. Management of environment risk within emergency territorial systems. Case
study-the influence area of the Bucharest city. Journal of Environmental Protection And Ecology 2012b, 13(4): 2360-2370.
7.Peptenatu D, Sîrdoev I, Prăvălie R. Quantification of the aridity process in south-western Romania. Iranian Journal of Environmental Health
Science and Engineering 2013, 11(3): 1-6.
8.Petrişor A.I. Assessment of the green infrastructure of Bucharest using Corine and urban atlas data. Urbanism.Architecture.Constructions
2015a: 6(2):19-24.
9.Petrişor AI. Using Corine Data to look at deforestation in Romania: distribution & possible consequences, Urbanism. Architecture.
Constructions, 2015b, 6(1):83-90.
10.Popescu, O.United Nations decade on biodiversity: strategies, targets and action plans, Urbanism Architecture Constructions 2015, 6(2): 37-
50.
11.Jelinek HF, Jones C, Warfel M, Lucas C, Depardieu C, et al. Understanding fractal analysis? The case of fractal linguistics. Complexus 2006,
9: 66–73
12.Klonowski W, Stepien R, Stepien P. Simple fractal method of assessment of histological image for applications in medical diagnostics.
Nonlinear Biomedical Physics 2010; 4(7). http://www.nonlinearbiomedphys.com/content/pdf/1753-4631-4-7.pdf.
13.Stepien P, Stepien R. Analysis of contours of tumor masses in mammograms by Higuchi’s fractal dimension. Biocyb Biomed Eng 2010;
30:49–56. http://www.ibib.waw.pl/bbe/bbefulltext/bbe_30_4_049_ft.pdf.
14.Ahammer H, Kroepfl JM, Hackl C, et al. Fractal dimension and image statistics of anal intraepithelial neoplasia. Chaos Soliton.Fract. 2011,
44:86–92.
15.Hinrikus H, Bachmann M, Karai D, Klonowski W, Lass J, Stepien P, Stepien R, Tuulik V. Higuchi's fractal dimension for analysis of the
effect of external periodic stressor on electrical oscillations in the brain. Med BiolEngComput 2011;49(5):585-91.
16.Muenzner JK, Biersack B, Kalie H, Andronache IC, Kaps L, Schuppan D, Sasse F, Schobert R, Gold(I) Biscarbene Complexes Derived from
Vascular-Disrupting Combretastatin A-4 Address Different Targets and Show Antimetastatic Potential. ChemMedChem 2014, 9(6):1195-204.
17.Higuchi T. Approach to an irregular time series on the basis of the fractal theory. Physica D 1988, 31: 277–283.
18.Jelinek HF, Matthews S, Succar P, McLachlan CS, Ahammer H, et al. Establishing a reference range for oligodendroglioma classification
using Higuchi dimension analysis IASTED Biomedical Engineering 2012; 764–039: 25–28.
19.Ahammer H. Higuchi dimension of digital images PLoS ONE 2011; 6(9):e24796. http://dx.doi.org/10.1371/journal.pone.0024796.
20.Russel D, Hanson J, Ott E. Dimension of strange attractors. Phys. Rev. Lett. 1980: 45(14), 1175-1178.
21.Di Ieva A, Grizzi, F, Ceva-Grimaldi, G, Carlo Russo, C, Gaetani, P, Aimar, E, et al. Fractal dimension as a quantitator of the microvasculature
of normal and adenomatous pituitary tissue, J. Anat. 2007;211(5), 673–680. doi:10.1111/j.1469-7580.2007.00804.x.