HIP 1.1
High-sensitive Innominate Processing
User Manual
Miroslav Králík1
, Petra Urbanová1
, Ondřej Klíma2
, Tereza
Mikešová 1
, Martina Wagenknechtová1
, and Jana Jungerová1
1
Laboratory of Morphology and Forensic Anthropology,
Department of Anthropology, Faculty of Science, Masaryk
University, Kotlářská 2, Brno 611 37, Czech Republic
2
IT4Innovations Centre of Excellence, Brno University
of Technology, Božetěchova 1/2, 612 66 Brno, Czech Republic
September 18, 2018
Contents
1 User Manual 3
1.1 Starting the program . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 General features of the program . . . . . . . . . . . . . . . . . . . 3
1.3 Reference skeletal collections . . . . . . . . . . . . . . . . . . . . 3
1.4 Scanning process . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Important technical note . . . . . . . . . . . . . . . . . . . . . . . 6
2 Procedure 6
2.1 Image loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Selecting the side . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Opening the TPS ﬁle . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Saving measured data . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.6 Enhancing/zooming in on a view . . . . . . . . . . . . . . . . . . 10
2.7 Measuring – digitalization of points . . . . . . . . . . . . . . . . . 10
2.8 Control of point conﬁguration . . . . . . . . . . . . . . . . . . . . 12
2.9 Sex diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.10 Selection of reference samples . . . . . . . . . . . . . . . . . . . . 13
2.11 Traditional morphometry . . . . . . . . . . . . . . . . . . . . . . 13
2.12 Visualization of individual dimensions . . . . . . . . . . . . . . . 13
2.13 Linear discriminant analysis . . . . . . . . . . . . . . . . . . . . . 14
2.14 Stepwise linear discriminant analysis . . . . . . . . . . . . . . . . 15
2.15 Geometric morphometry . . . . . . . . . . . . . . . . . . . . . . . 16
2.16 Results export . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.17 Coming soon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Appendices 24
A Osteometric landmarks deﬁnitions 24
A.1 Pubic view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
A.2 Iliac view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
B A list of rejected dimensions for traditional morphometry 28
B.1 Pubic view, Iliac view . . . . . . . . . . . . . . . . . . . . . . . . 28
C Recommended conﬁgurations for geometric morphometry 29
C.1 Pubic view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
C.2 Iliac view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
About the program
HIP 1.1 program is the second functional version of the HIP computer program
designed for morphometric estimation of the sex of an unknown skeletal ﬁnding
based on the pelvic bone. The authors of the program are Miroslav Králík ,
Petra Urbanová , Ondřej Klíma , Tereza Mikešová, Martina Wagenknechtová,
and Jana Jungerová. The program works with standardized 2D images
of bones from a desktop scanner and allows the case to be classiﬁed into a
group according to sex using methods of traditional and geometric morphometry.
One-dimensional and multi-dimensional statistics computing procedures
and graphical procedures are based on the R software, distributed under the
GNU (GPL) license, and its extending libraries from various authors.
Acknowledgements
For access to documented ﬁles and versatile help let us thank to:
Professor MUDr. Miloš Grim, DrSc. (Institute of Anatomy, First Faculty of
Medicine, Charles University, Prague, CZ), doc MUDr. Ondřej Naňka, Ph.D.
(Institute of Anatomy, First Faculty of Medicine, Charles University, Prague,
CZ), doc. RNDr. Ladislava Horáčková, Ph.D. (Department of Anatomy, Faculty
of Medicine, Masaryk University, Brno, CZ), Professor Sotiris K. Manolis
(University of Athens, Greece), and Dr. Constantinos Eliopoulos (Liverpool
John Moores University, GB).
For assistance with organization and further cooperation we thank Mgr.
Tomáš Mořkovský (Department of Anthropology, Faculty of Science, Masaryk
University) and RNDr. Michaela Králíková, Ph.D. (Museum of Vyškov Region,
Bučovice). We thank Bc. and Bc. Alžběta Štastná and Mgr. Martin Čuta,
Ph.D. for the English corrections of this text.
The computer program was created thanks to the ﬁnancial support of the
Higher Education Development Fund (project FRVS/2034/2012) and Masaryk
University Development Fund (project MUNI/FR/0284/2014).
License
The program is freely available and is distributed under the GPLv3 license. The
oﬃcial license text is available at http://gnu.org/licenses/gpl.html.
Citation
Please, cite the HIP 1.1 as follows:
M. Králík, P. Urbanová, O. Klíma, T. Mikešová, M. Wagenknechtová, J. Jungerová,
HIP 1.1 - High-sensitive Innominate Processing, Masaryk University, Faculty of
Science, Department of Anthropology, Laboratory of Morphology and Forensic
Anthropology (LaMorFA), Brno, 2017. http://www.sci.muni.cz/lamorfa/
veda-a-vyzkum#projekty.
Technical and software speciﬁcations
HIP 1.1 is designed for Windows operating system, versions Vista and higher.
The program was not tested on lower versions of the Windows operating system
1
(Windows XP, NT, 98, 95) and probably will not work properly. The program
was written in C ++ and R [2], and created in QtToolkit application, version
4.8.2, the user interface was created in QtDesigner. It uses Bioconductor R,
version 2.15.1. Rtools were used to compile Qt libraries and the entire application,
Rtpp, Rinside, and examples from QtDensity (under the GPL licence)
were used for compiling Qt and R. Qt itself is licensed under LGPL. During
the creation of the program, the computational and graphic functions of the following
packages of the program R [18] were used: base [18], stats [18], graphics
[18], MASS [24], shapes [7], klaR [25], and Momocs [4]. These packages also
use functions of other packages: scatterplot3d [11], rgl [1], methods [18], sp [18],
ReadImages [12], ade4 [6], spdep [3], boot [5], Matrix [2], lattice [20], nlme [15],
maptools [10], foreign [17], grid [18], deldir [21], coda [16], and splines [18].
Development of versions
Program HIP 0.3 was ﬁnalized in 2012. The software included only one reference
sample (Athens) and contained an additional modul for sex estimation by
the human clavicle.
Program HIP 1.0 was ﬁnalized in 2015. It was intended for the hip bone
only (clavicle was omitted) and two additional reference samples were included
(Prague and Brno).
Program HIP 1.1 was ﬁnalized in 2017. Some new computational routines were
added and some bugs were ﬁxed.
Please send your inquiries, comments and suggestions to the address mailto:
mirekkralik@seznam.cz (Miroslav Králík).
2
1 User Manual
1.1 Starting the program
The program is freely available for download at the website of the Department
of Anthropology MU:
http://www.sci.muni.cz/lamorfa/veda-a-vyzkum#projekty
and also at the site:
https://sourceforge.net/projects/hip-project/
Click the Download link to download the ZIP ﬁle to your computer and
unzip it. The start the program, double-click the HIP.exe ﬁle icon in the
bin folder. The shortcut (command script) of this HIP.cmd ﬁle is also located
at the top level of directories and can be seen immediately after unpacking and
opening the downloaded folder.
1.2 General features of the program
The use of the program consists of preparation/recording of a suitable standardized
2D image of the pelvic bone (synonyms: hip bone, innominate bone).
The pelvic bone is evaluated from two standardized views (Pubic view and Iliac
view). In the current version 1.1, it is pssible to evaluate separately each pelvic
bone view or combine both pelvic bones views in one analysis. The program
supports sexual diagnosis of an unknown case using traditional morphometry
– distances/measurements (linear distance between measurement points, or interlandmark
distances), both single dimensional visualization and comparison
of a selected interlandmark distance, as well as multidimensional Linear Discriminant
Analysis (LDA) of interlandmark distances. In addition to traditional
morphometry, the program also oﬀers geometric morphometry, namely the Generalized
Procrustes Analysis (GPA) and Linear Discriminant Analysis (LDA)
of Procrustes shape coordinates, i.e., sex estimation purely on the basis of shape,
without the inﬂuence of size variables.
1.3 Reference skeletal collections
Sex estimation of an unknown case is based on a comparison with documented
reference bone collections, that is, the bones of people whose sex in life was
documented and is known with certainty. Version HIP 1.1 contains data from
three documented collections of pelvic bones:
(A) Pelvic bones come from the collection deposited at the Institute of Anatomy,
First Faculty of Medicine, Charles University in Prague. The collection includes
pelvic bones of 100 individuals, 99 individuals are represented by
both pelvic bones, one skeleton is represented only by the right pelvic
bone. The collection contains pelvic bones of 54 males and 46 females.
The skeletal remains are autopsy material. The majority of individuals (90
individuals) are part of the so-called Pachner Collection - skeletons macerated
at the Czech anatomical institute during the 1930s (1933–1936),
8 females represent skeletons macerated at the German anatomical institute
in 1884–1902, in two cases the pelvis bones belong to individuals of
unknown origin. The whole population is assumed to originate from the
3
lower socioeconomic strata. In most cases, a complete pelvis (assembled
from both pelvic bones and the sacrum) was available, isolated pelvic bones
were present in 7 individuals. The condition of the skeletal material was
varied and included abrasion of the bone surface, fractures and breakage of
pelvic bone parts, especially at the junction areas of the individual pelvis
bones (the pubic bone near facies symphysialis and the area of spinae iliacae
posteriores). (Note: All 100 individuals were selected into the database,
of which 98 individuals have both bones, 1 individual (male P_058) only
the bone of the left side, 1 individual (female P_092) only the bone of the
right side.)
(B) Bones from the autopsy collection from the Department of Anatomy, Faculty
of Medicine, Masaryk University. From the Brno bone collection which
originated from the autopsies of the dead human bodies used in the education
of medics, there were 29 pelvic bones available, of which 27 are
represented by the two pelvic bones, the two skeletons are represented only
by the right pelvic bone. This collection includes pelvic bones of 23 males
and 6 females. In all cases, the pelvic bones were isolated (i.e. they did not
form the pelvis with the other pelvic bone and the sacrum). In this collection,
the pelvic bones were preserved in an excellent condition, there were
no breaks or abrasions of the bone surface. (Note: The complete collection
has more individuals, but some were excluded because of an unknown sex
(NA), in one case (female B_089) there was probably a pathology (rickets?),
27 individuals both bones, 1 individual (female B_029) right bone,
1 individual (male B_235) right bone.
(C) Bones from the collection of The University of Athens Human Skeletal Reference
Collection, National and Kapodistrian University of Athens, Greece.
The pelvic bones originate from The University of Athens Human Skeletal
Reference Collection, stored at the Department of Animal and Human
Physiology, Athens, Greece. The collection contains documented skeletal
remains of 225 individuals that were obtained by exhuming graves from
several cemeteries in the area of Athens. The group contains skeletal remains
of people who died between 1960 and 1996, aged between early childhood
and old age (Eliopoulos et al., 2007). Individuals were predominantly
from lower or middle socio-economic backgrounds. After the exclusion of
children, visibly pathological and strongly fragmentary cases, the studied
collection of pelvic bones includes a total of 186 individuals. 171 individuals
are represented by both pelvic bones, in 5 individuals the pelvic
bone is present only on the right side and in 10 skeletons only on the left
side. This collection includes pelvic bones of 102 males and 84 females. All
pelvic bones were also isolated and their condition was good. The most
common were abrasions or fractures of facies symphysialis, spina ischiadica,
and spinae iliacae posteriores. (Note: 171 individuals have both bones,
5 individuals (female A_018, female A_119, male A_126, female A_188,
male A_200) have the right bone only, 10 individuals (male A_078, female
A_095, male A_107, male A_140, female A_144, female A_167,
male A_177, female A_190, male A_192, male A_209) have the left bone
only).
In sum, the program contains reference data from three documented bone
4
collections: 198 pelvic bones form the collection of the Institute of Anatomy,
First Faculty of Medicine, Charles University, 56 pelvic bones from the collection
of the Faculty of Medicine, Masaryk University, and 357 pelvic bones from The
University of Athens Human Skeletal Reference Collection. In total, there are
611 pelvic bones from 315 individuals (179 males, 136 females).
Due to the side diﬀerences of the pelvic bone, the program compares an
unknown case primarily with the part of the reference collection of bones coming
from the same side of the body. However, the opposite side can be also chosen.
1.4 Scanning process
The starting-point of the evaluation is scanning the bone in standardized positions
using the CanoScan 4400F 2D desktop scanner. It is a scanner with a
horizontal glass platen of a format slightly larger than A4. The desktop scanner
provides fast, high-quality recording of images of formations with relatively ﬂat
surface. However, it is necessary to standardize the images:
(a) Pubic view: place the pelvis on the outer surface of the pubic bone so
that the ﬂat part of ossis pubis adheres best to the glass platen. Place the
bone in the centre of the platen so that it is recorded whole and orient it
so that the most rearward point on the tuber ischiadicum and the most
backward protruding point on the edge of the bone in the area of the hip
joint (acetabulum) are on a line parallel to the longitudinal axis of the
scanner (corresponding to the direction of movement of the lamp). The
lamp moves from the cranial (crista iliaca) to the caudal region of the
pelvic bone (tuber ischiadicum), not vice versa. In this position the bone
does not stay by itself, it should be held or weighted down during scanning
(e.g. with a large amount of plasticine).
(b) Iliac view: place the pelvic bone on the outer surface of the hip bone (the
wing) on the scanner platen so that the ﬂat part of the wing of the hip
bone is as close as possible to the glass plane. Place the entire bone in
the centre of the platen in order to record it whole and orient it so that the
most rearward/medial point (i.e. on the side facing the sacrum) on the tuber
ischiadicum and the most rearward point on the edge of the bone in the area
of spina iliaca posterior superior are on a line parallel to the longitudinal
axis of the scanner (corresponding to the direction of movement of the lamp).
The lamp moves from the cranial (crista iliaca) to the caudal region of the
pelvic bone (tuber ischiadicum), not vice versa. In this position the bone
usually stays by itself and does not need holding while being scanned.
The bones of the reference collection were scanned into full-color TIF ﬁles,
100% in size, resolution 150 dpi in both directions. The program also loads JPG
and PNG ﬁles.
Because this resolution has been used with reference cases and has been set
as optimal for compromise between image details and ﬁle size, we recommend it
to be used. But this is not a requirement; the program also works with images
with other resolutions if scanning 100% in size is retained.
5
1.5 Important technical note
The 2D desktop scanner was chosen because it is now aﬀordable for almost
everyone. This is related to our commitment to free use of the new methodology
and software developed under the GPL license. This would be very limited
if the methodology required expensive equipment to record three-dimensional
morphology such as a laser scanner or a 3D digitizer.
However, the use of a desktop scanner has its limitations. Diﬀerent desktop
scanners provide almost identical images of the conﬁguration on the scanner
plane, but the three-dimensional formations over the platen (beyond its plane)
they convert into 2D image slightly diﬀerently. Therefore, as with any other
recording device, it is always needed to calibrate images. Due to this, we have
recorded all existing cases with the Canon CanoScan 4400F scanner. For other
scanners, the data needs to be adjusted. However, an automatic calibration
of the scanner is not yet part of this program. Therefore, do not use HIP
1.1 results based on images from recording devices other than this particular
desktop scanner (Canon CanoScan 4400F) as conclusive.
We are working on creating a calibration feature for other scanners and
cameras, and we expect it to be included in the program in one of the later
versions.
2 Procedure
2.1 Image loading
After starting the program, chose from the Pubic view or Iliac view in the
upper tabs. After selecting e.g. Iliac view, chose Load image in the left
sidebar and load from the folder on the computer the corresponding standardized
image of the pelvic bone viewing the outer surface of the bone (Figure 1).
The image opens and the name and type of the graphic ﬁle appears in the
upper left column in a cell labelled File name. At the same time, the program
automatically retrieves and displays the image resolution in both directions
(Horizontal DPI, Vertical DPI). Image resolution is used to calibrate the
size. Note: Upper tabs can be moved freely and their order can be changed
according to individual need.
2.2 Selecting the side
In the cell labelled Body side in the left menu, it is necessary to manually
specify whether the bone is from the right or left side. If it is the bone from the
right side (default setting), nothing changes. If the bone is from the left side,
change right to left and the image ﬂips – a bone from the left side will look like
a bone from the right side. However, the data will show that it is the bone from
the left side. We created this step because of the asymmetry of sight, the same
osteometric points may be evaluated diﬀerently, depending on whether they are
on the right or left side.
6
Figure 1: Screenshot of the program window in the Pubic view digitization tab
(top) and Iliac view (bottom). In the left menu there are control functions,
measurements are made in the main window.
7
2.3 Opening the TPS ﬁle
In addition to loading the image, it is possible to load data by Load .tps ﬁle
directly in the TPS format (see Example bellow), which is commonly used by
J. Rohl’s morphometric programs and is the output of his tpsDig 2D digitizer
[19]. If the corresponding image is linked to the data in the TPS ﬁle, it opens
directly when the TPS ﬁle is loaded if it is in the same folder. However, the data
must in the number and sequence correspond to the osteometric points deﬁned
below.
For example, in the same folder we have image 001b.tif on which the pelvis
is in the Pubis view, and a TPS ﬁle 001b.tps, which contains the coordinates
of 24 deﬁned osteometric points from this image. The TPS ﬁle is a regular
text ﬁle with a * .tps ending, containing the Cartesian coordinates (x, y) of
the osteometric points of the view displayed in a list, in the ﬁrst row is their
number (here LM=24) and the last line is refers to the respective image (here
IMAGE=001b.tif):
LM=24
974.00000 560.00000
621.00000 768.00000
519.00000 854.00000
552.00000 944.00000
484.00000 1054.00000
529.00000 1218.00000
390.00000 826.00000
443.00000 832.00000
451.00000 537.00000
534.00000 564.00000
601.00000 668.00000
367.00000 576.00000
400.00000 504.00000
385.00000 477.00000
346.00000 418.00000
519.00000 220.00000
584.00000 244.00000
815.00000 349.00000
888.00000 368.00000
693.00000 619.00000
583.00000 369.00000
724.00000 414.00000
743.00000 474.00000
566.00000 486.00000
IMAGE=001b . t i f
2.4 Saving measured data
We also chose the modiﬁed TPS format to store digitized data. Clicking Export
.tps file can save measured/digitized data in TPS format to a folder on your
computer so you can return to it later. If the image was digitized directly (i.e.,
we did not open the TPS ﬁle, but the TIF ﬁle) and the data were stored in
8
another folder, the image is saved automatically with them.
The program diﬀers from the tpsDig digitizer [19] by storing also an incomplete
set of points (for example, when some points cannot be measured due to
bone damage). In this case, the NaN code will be written instead of the missing
point in both coordinates.
Additionally, the TPS ﬁle stored in the HIP program contains an indication
of whether it is a bone from the right or the left side. A TPS ﬁle of a bone form,
the right side with only three points can look like this (pubic view):
LM=24
974.00000 560.00000
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
443.00000 832.00000
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
519.00000 220.00000
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
NaN NaN
IMAGE=001b . t i f
ID=0
SIDE=right
2.5 Example
If you do not have your own bone images, the program contains a number
of already measured examples in the examples folder, divided by individual
views. It is possible to directly upload data in the TPS format of already
measured bones using the Load .tps file and then work with them. To test
digitalization of points, however, it is better to use the Load image to open
an unprocessed TIF image of the bone (Warning: Opening the correct TPS
belonging to the relevant view is checked by the program so that e.g. iliac view
data cannot be opened in the pubic view analysis but it does not check it while
opening an image of a bone; therefore, it is necessary to make sure to open the
correct image in the corresponding view).
9
2.6 Enhancing/zooming in on a view
You can adjust the image to the window size by clicking the Fit button. You
can zoom in or out by dragging the slider above this button or with a mouse
wheel (if the mouse pointer is in the slider area, you can also move it with the
mouse wheel). While zoomed in, you can move the sliders at the edges of the
main window or by moving the blue square in the bottom left view. When
measuring, always use an appropriate zoom to place the osteometric points on
the image as accurately as possible.
2.7 Measuring – digitalization of points
Measurement on an image means the manual placement of the cursor to the position
of the relevant osteometric point given by the deﬁnition and recording the
Cartesian coordinates (x, y) of that point. These coordinates are the basis for
both traditional morphometry and geometric morphometry. When measuring,
place the arrowhead of the cursor to the position of the relevant osteometric
point and click the right mouse button. The following dialogue appears (Figure
2):
Figure 2: Dialogue of osteometric points digitalization on a pelvic bone in Pubic
view.
By clicking on the given point, you can see its position on the bone schema
and its deﬁnition at the same time.
If you click outside of the dialogue window, you will return to the loaded
bone image. It is necessary ﬁrst to thoroughly study the deﬁnitions of
all osteometric points. If you are not sure whether you correctly understand
a deﬁnition of a point, click on Sample cases of landmark # and in the web
browser some examples of the positions of the point on real bones or other
explanatory diagrams will be displayed. Deﬁnitions of osteometric points are
given at the end of this manual.
10
Figure 3: Numerical label and deﬁnition of point 1 in Iliac view in the frame
made by the cursor.
Double-clicking on a point in the list returns to the bone image where the
digitized point is shown as a light green dot with the appropriate number. The
point can be moved by the mouse to correct its position at any time. Numerical
labels of osteometric points can be switched on/oﬀ by the Show labels button
at the top left.
If you place the cursor over an osteometric point in the image, the cursor
changes into a hand, the number of the point and its deﬁnition are displayed
in a frame (Figure 3). If you right-click the point, a dialogue (Figure 4) will
appear, which will allow us to manipulate its position with the numerical value
of its x,y coordinates. Click the Remove button to delete the point if it is entered
incorrectly. However, we can only deactivate the point by clicking the
Deactivate button if, due to uncertainty about its position or bone preservation
in the given place, we only want to exclude it from the current analysis
and compare the results obtained with and without it. The deactivated point
changes colour to blue. Be sure to always activate all deactivated points before
completing the analysis; otherwise the program saves their coordinates as NaN,
that is, as if the point was missing.
When digitizing the points deﬁned by perpendiculars, HIP 1.1 makes it easier
to ﬁnd their position by automatically displaying blue perpendiculars (blue
dotted lines). The program can also determine the smallest/greatest distance
between selected points, the displayed distances are in units after calibration
(i.e. in millimetres). This is how it is automatically calculated in Iliac view:
length of a perpendicular from point 3 to the line connecting points 1 and 2,
length of a perpendicular from point 16 to the line connecting points 14 and
15, line connecting points 6 and 21 (length of the pelvic bone); in Pubic view:
distance of points 18 and 22 (width of the lower branch of the pubic bone).
The measurements are similar in both pelvic bone views, the diﬀerences are
only in the number of deﬁned points and preferred conﬁgurations.
11
Figure 4: The dialogue for manipulation with an osteometric point after clicking
on it.
2.8 Control of point conﬁguration
In digitalization, in the left column in a cell labelled GM Shape types is shown
the suﬃciency of bone coverage with points in terms of the application of geometric
morphometry (does not concern traditional morphometry). Based on the
morphometric analysis, we have selected several conﬁgurations of osteometric
points (described in the Appendix of this manual) for each view, which suﬃciently
describe the shape of the bone and are not redundant. Thus, there is
information on which of the preferred conﬁgurations of a given bone is available
(e.g. Optimal conﬁguration). Without fulﬁlling the so-called minimum deﬁned
conﬁguration, no further analysis will take place and the dialogue will indicate
that the analysis cannot be performed. With insuﬃcient number of points, the
posterior probability often does not reach value 95% or higher and the results
of the sex diagnosis cannot be considered reliable. Deﬁnitions of the preferred
conﬁgurations (the only ones allowed in this version of the program) are also
listed at the end of the manual for each of the views.
2.9 Sex diagnosis
Once the digitalization is complete, we proceed to sex estimation. In this step,
the program oﬀers several options.
12
Figure 5: Landmark distance visualization tab from the Traditional
morphometry tab after displaying the diﬀerence in the 1-3 dimension in Iliac
view.
2.10 Selection of reference samples
First, it is necessary to choose which comparative/reference samples to compare
the case against. In the right menu/Database tab, select one, two or all three
reference ﬁles (Athens, Brno, Prague) in the Sample ﬁeld. Next, in the Body
side ﬁeld, choose which side of the body the reference samples should come
from (Auto – according to the side of the assessed bone, Right, Left).
2.11 Traditional morphometry
Now select the top Traditional morphometry tab. From the coordinates of the
osteometric points, all available dimensions (distance between all pairs of measured
points) are automatically calculated and converted to millimetres based
on the resolution of the images. In the left column we can then choose from
several tabs.
2.12 Visualization of individual dimensions
At the bottom left of this tab is the Landmark distance visualization tab,
which allows you to graphically display the distribution of any dimension. Select
a view that you want to evaluate (Iliac view or Pubic view), select the two
points whose distance interests you and click on Plot distance. Using the
probability density plot and the box and whisker plot, where females are always
red and males blue in colour, the sex diﬀerences in the selected dimension in
the reference ﬁle are shown. The green line shows the location of the unknown
case. Just by displaying the individual dimensions, it is possible to detect cases
extremely masculine or extremely feminine. However, most individuals are in
the overlapping zone of male and female values.
13
2.13 Linear discriminant analysis
The group sorted according to sex can include an unknown case based on a
multi-dimensional combination using Linear Discriminant Analysis - LDA [9].
The method searches in a multidimensional space for a combination of variables
that maximizes diﬀerences between groups (in our case between the sexes).
The LDA program calculates the lda function of the MASS package [24], the
prediction of the unknown case sex is performed using the predict function from
the MASS package [24].
Select the middle tab of the traditional morphometry labelled Linear discriminant
analysis. Start by choosing in the left column a view you want to
evaluate (Both views together, Pubis view, Ilium view). If you select Both views
together, the calculation will be based on the combined data of both views. Then
you can choose from three options: (a) use all dimensions, (b) exclude some of
all dimensions (i.e., negative selection), or (c) select only some (i.e. positive
selection) from all dimensions.
a) To calculate discriminant analysis from all available dimensions, select Use
all distances. However, due to the circumstances described below, this
is quite inappropriate. You must always carefully choose the dimensions for
discrimination.
b) If you want to calculate discriminant analysis without some dimensions, select
the option Avoid selected distances. This option automatically disallows
the calculation of discriminant analysis from the dimensions that the
authors of the HIP program determined to be inappropriate. Their list is
provided in an appendix at the end of this manual. These are dimensions
based on pairs of points that vary greatly in the reference sample(s), often
approach zero, and at the same time are very small in relation to the total
size of the bone. This leads to a high degree of accidental eﬀects (due to an
inaccuracy of measurement, occurrence of pathologies, damage to the surface
of the bone, etc.) in the process of discrimination. The program blocks
all (except for justiﬁed exceptions) dimensions whose values (or part of the
values) are lower than 10 mm. When this option is selected, the list of the
“prohibited” dimensions appears in red at the top right. If there is a reason
to add to these inappropriate dimensions more to exclude from the analysis,
select the two points that deﬁne the rejected dimension in the windows
named Distances on ...view (... is the name of the view) and click the
Add to list button on the right. The dimension is added to the blocked
dimensions list and will be black. However, it can be unblocked anytime by
selecting it and pressing the Remove from blocked button.
c) If you want to calculate discriminant analysis only from selected dimensions,
select Use only selected distances. Choose the appropriate dimensions
either from experience or by viewing individual dimensions and their degree
of dimorphism in the Landmark distance visualization tab. In the
Distances on ...view (... is the name of the view) windows, select the
chosen dimensions and add them to the list by clicking the Add to list
button.
Hence, the data can be selected for analysis at two levels, the ﬁrst level is
14
the activation and deactivation of the points in the measurement, the second
level is the blocking or the selection of the dimensions before the analysis.
Once the dimensions to enter the analysis are set, you can proceed to the
Linear Discriminant Analysis itself. Click the Run analysis button. The program
selects from the reference collection database all the dimensions that we
have allowed and which are currently available from the evaluated case. The
LDA will run on these dates, the program calculates a new discriminant function,
and then uses it to include an unknown case. During the analysis (while
the program calculates) a green progress bar indicator moves below the analysis
start button.
Several result tables are displayed after the analysis is complete:
A) The table on the left shows the Prior probability, i.e. the original probability
of the case classiﬁcation by sex even without an analysis that is set
to the proportion of the given sex in the reference sample (selection). Posterior
probability indicates the likelihood of correct case classiﬁcation and sex
estimation. In the sex with a higher probability, the cells are highlighted in
yellow. Posterior probability values of 0.95 and higher are generally considered
as practically sure to be allocated in the appropriate class. In addition,
there is a classiﬁcation table that shows the total number of males and
females in the sample (Cases) and the number (Misclassified) and percentage
(Misclassified [%]) of incorrectly classiﬁed males and females.
B) On the right there is a table of selected results of discriminant analysis
and classiﬁcation. The value of a discriminant score for the unknown case
(Value of disc. function) is displayed. Additionally, there are the total
percentage of Accuracy of discrimination of reference cases, the total
number of Misclassified cases regardless of sex, the Cut-off point, and
the total number of Cases and Variables in the analysis.
C) The coeﬃcients of discriminant function (LD1) for each dimension, the mean
of each of the dimensions used (Male mean, Female mean) and the respective
value of the unknown case variable (Case) are given in the table below
(Coefficients of discriminant function).
D) The probability density plot (Probability density) and the box and whisker
plot of the discriminant score values of the reference cases (women in red,
men in blue), and a green line indicating the value of the discriminant score
of the analysed case are displayed bellow (similarly to individual dimensions).
The graphs are created by the function density of the stats package
[18] and the functions plot, lines, and polygon of the graphics package [18].
2.14 Stepwise linear discriminant analysis
The top tab of traditional morphometry is the Stepwise linear discriminant
analysis. In the left column, select the view you want to evaluate (Views
combined, Pubic view, Iliac view).
Since discriminatory analysis is very sensitive to the ratio of ﬁle size to
number of discriminant variables, it is generally recommended that the ratio of
the number of individuals and the number of variables be at least 5:1 (5 cases
per variable) or higher [13]. It is therefore appropriate to include only a few
15
of the most eﬀective variables (in this case, dimensions) in the model. This
is the purpose of Stepwise Linear Discriminant Analysis (Stepwise LDA). The
program uses functions from the MASS [24] and klaR [25] packages, in particular
the stepclass and lda functions. Prediction is again performed by the function
predict from the MASS package [24].
Afterwards, proceed as described above with the LDA. Choose the dimensions
from which the analysis is to be calculated. The stepwise procedure is
then realized by a default setting of the function stepclass using the function
ucpm from the klaR package [25] to calculate the criteria for assigning a variable
to the resulting model. The calculation includes a leave-one-out randomization
method, i.e. n-fold cross-validation, where n is the number of samples (cases)
of the reference collection(s). The number of cases (and thus the number of
cross-validated samples) will vary according to the selection of bone views and,
above all, the selection of reference collections. One case is randomly selected
from the reference collection and a discriminant function is created on the rest
of the collection (with n–1 number of cases). The pre-selected case is then used
as validation (test) and the discriminant function is used on it. The whole process
is repeated n-times. Details of the preconﬁgured stepwise procedure are
described in the manual of the klaR package [25].
Click on the Run analysis button. Due to the stepwise procedure and
verifying, the calculation is time consuming and may take from a few tens of
seconds to several tens of minutes depending on the number of individuals and
dimensions. The ongoing calculation is again indicated by the movement of
the green progress bar below the start button of the analysis. Completion of
the calculation is indicated by the window Stepwise LDA has finished. The
program lists results in the same structure as the classic LDA (Figure 6).
In order to reach a compromise between the calculation speed and the estimation
reliability, the preferred method of working with the program (according
to the authors of the program) is to use a classical Linear Discriminant
Analysis on a small number of selected dimensions. We recommend that you
deliberately choose 3 well-discriminating dimensions in each view and use these
six for the analysis. If it is not successful (posterior probability does not reach
0.95 or higher), choose a few other dimensions and perform the discriminant
analysis again. Given the growth patterns of the pelvic bone, it is the right
choice (if you have both views available) to select a few dimensions from each of
the two views [14]. A certain aid for choosing the dimensions can be the selection
of dimensions in the picture (Figure 7). If there is a complete set of points
from both views (and therefore all possible dimensions), some of these dimensions
(Figure 7) was part of a set of selected dimensions in most of the many
thousands of stepwise discriminant analyses performed. They can therefore be
considered the most reliable and be used deliberately. If these dimensions are
not available (due to the lack of clarity or absence of landmarks), it is advisable
to choose dimensions similar to them in direction and position.
2.15 Geometric morphometry
An alternative to traditional morphometry is evaluation using geometric morphometry
(GM). GM procedures are most often based on data in the form of
Cartesian coordinates of landmarks and have the advantage that we can evaluate
the shape and size of objects separately [23, 22].
16
Figure 6: Top and bottom part of the results in the Stepwise linear
discriminant analysis from the Traditional morphometry tab after calculating
the discrimination in Iliac view.
Select the Geometric morphometry tab at the top, and in the left column
choose the appropriate view (Pubic view analysis, Iliac view analysis, Views
combined). If you select both views together, the LDA calculation will be based
on the combined data of both pelvic bone views after each has gone through an
individual superposition (see below).
At the beginning of the procedure, the program performs a checking of the
digitalized point conﬁguration. As mentioned above, the appropriate conﬁgurations
were designed to adequately represent the shape being evaluated and at
the same time not contain redundant points – from a group of diﬀerently deﬁned
points that lie close to each other and strongly correlate, always only one is
17
Figure 7: Dimensions of both pelvic bone views, which most often appear in
the results of Stepwise discriminant analysis.
used (for more details see the Recommended Conﬁgurations for geometric morphometry
at the end of this manual). For Pubic view, Optimal Conﬁguration
(13 points), Reduced Conﬁguration (9 points), and Minimal Conﬁguration (7
points) are deﬁned. For Iliac view, Optimal Conﬁguration (13 points), Reduced
Conﬁguration A (11 points), Reduced Conﬁguration B (10 points), Reduced Conﬁguration
C (8 points), and Minimal Conﬁguration (5 points) are deﬁned. The
program automatically checks which conﬁgurations are available. Then it calculates
using the one which has the largest number of points. If a conﬁguration
has an extra point compared to a lower conﬁguration but the combination of
the points is not higher, the excess point will not be used in the calculation.
Selected point conﬁgurations are superimposed by Generalized Procrustes
Analysis (GPA) [8]. Firstly, the program calculates the centroid size (CS),
which is the square root of the sum of squared distances of a set of landmarks
from their centroid (CS can only be compared between conﬁgurations of the
same points). Secondly, a standardization of the conﬁgurations being compared
is performed by removing the diﬀerences in the centroid size (CS = 1 in all
cases), in the position and rotation of all conﬁgurations. The conﬁgurations are
superimposed on each other so that the sum of squares of point distances within
the clusters of individual points is minimal. Finally, after the superposition, only
the shape diﬀerences between the conﬁgurations remain in the data – new socalled
Procrustes shape coordinates. The size of the original conﬁgurations is
stored separately as CS. For the GPA calculation, the program uses the procOPA
function from the shapes package [7].
Procrustes shape coordinates as shape indicators and centroid size CS as a
18
size indicator are then used as input variables for Linear Discriminant Analysis
by sex. Its procedure has already been described above. When analysing, it is
possible to choose whether you want to discriminate purely based on the shape
(Procrustes shape coordinates), which is the default setting, or to combine the
coordinates with the centroid size. In this case, check the Include centroid
size box and the centroid size will be included among the LDA variables.
Click on the Run analysis button. The results of sex discrimination (Figure
8, 9, 10) have a similar structure to that of traditional morphometry, including
graphs, as described above. In addition, this analysis shows graphically the
shape of the unknown case in a superposition with the average shape (consensus)
for males and females. The case being evaluated is represented by a green line,
females are in red and males in blue. The representation is complemented by
a deformation grid tps (thin plate spline) that represents how a plastic square
grid would deform, if it was covered by the points conﬁguration of the unknown
case, the points were attached to the grid, and then moved to the average female
and male conﬁguration. The largest shape changes are shown by thickening
(darkening) or by diluting (lightening) of the grid. The grid density and shape
change ampliﬁcation to highlight the graphical visualization of the diﬀerence
can be adjusted to the left by changing the Grid size and Amplification
values before the calculation begins. The tps grid is calculated using the tps.grid
function from the Momocs package [4]. The graph is complemented with a point
graph of Procrustes shape coordinates in the same colours (males blue, females
red, the unknown case green).
Figure 8: Top results in Iliac view analysis from the Geometric
morphometry tab after calculating discrimination in Iliac view.
2.16 Results export
All results are automatically saved in the program directory in the outs folder
in a separate subfolder, named by time (year-month-day hour-minute-second)
when the analysis took place (e.g. 2018-09-18 17-45-19). The results are saved
in the HTML format which can be viewed in any internet browser, data can be
simply selected and copied for further use. Images are also stored separately in
the PNG format.
19
Figure 9: Graphs of the discriminant score in Iliac view analysis from the
Geometric morphometry tab after calculating discrimination in Iliac view.
2.17 Coming soon
In the next version, we’d like to extend the program with the following functions:
• calibration of deformation for diﬀerent types of table scanners or cameras
• descriptive statistics of input variables (coordinates, dimensions)
• integrated testing of the assumptions of parametric methods
• inclusion of other indicators of prediction and reliability of discrimination
• inclusion of other classiﬁcation procedures (e.g. logistic regression)
• new reference samples to support greater versatility and external reliability
of the methodology
• adding dimensions measured directly on the bones to the calculations of
the traditional morphometry.
20
Figure 10: Deformation grids in Iliac view analysis from the Geometric
morphometry tab after calculating discrimination in Iliac view (above) and superimposition
of the reference collection (blue points – males, red points – females)
with the analysed case (green crosses).
21
References
[1] Adler, D., Murdoch, D.: rgl: 3D visualization device system (OpenGL)
(2012), http://CRAN.R-project.org/package=rgl, r package version
0.92.892
[2] Bates, D., Maechler, M.: Matrix: Sparse and Dense Matrix Classes and
Methods (2012), http://CRAN.R-project.org/package=Matrix, r package
version 1.0-6
[3] Bivand, R., with contributions by Micah Altman, n, L.A., Assuncao,
R., Berke, O., Bernat, A., Blanchet, G., Blankmeyer, E., Carvalho, M.,
Christensen, B., Chun, Y., Dormann, C., Dray, S., Halbersma, R., an d
Pierre Legendre, E.K., Lewin-Koh, N., Li, H., Ma, J., ni Millo, G., Mueller,
W., Ono, H., Peres-Neto, P., iras, G.P., Reder, M., Tiefelsdorf, M., Yu.,
D.: spdep: Spatial dependence: weighting schemes, statistics and models
(2012), http://CRAN.R-project.org/package=spdep, r package version
0.5-53
[4] Bonhomme, V., Picq, S., with contribu tions from Dan Dkin, J.C.,
Gaucherel, C., Kriebel, R., a nd Marcelo Reginato, N.M., Telmon, N.,
Wishkerman., A.: Momocs: Shape Analysis of Outlines (2012), http:
//CRAN.R-project.org/package=Momocs, r package version 0.2-02
[5] Canty, A., Ripley, B.: boot: Bootstrap R (S-Plus) Functions (2012), r
package version 1.3-4
[6] Dray, S., Dufour, A.B.: The ade4 package: Implementing the duality diagram
for ecologists. Journal of Statistical Software 22(4), 1–20 (2007).
https://doi.org/10.18637/jss.v022.i04
[7] Dryden, I.: shapes: Statistical shape analysis, http://CRAN.R-project.
org/package=shapes, r package version 1.1-5
[8] Dryden, I., Mardia, K.: Statistical shape analysis. John Wiley & Sons,
Chisester (1998)
[9] Fisher, R.A.: The Use of Multiple Measurements in Taxonomic Problems.
Annals of Eugenics 7(2), 179–188 (1936)
[10] Lewin-Koh, N.J., Bivand, R., contributions by Edzer J. Pebesma, Archer,
E., Baddeley, A., Bibiko, H.J., Callahan, J., Carrillo, G., Dray, S., Forrest,
D., iendly, M.F., Giraudoux, P., Golicher, D., Rubio, V.G., trick Hausmann,
P., Hufthammer, K.O., Jagger, T., Luque, S.P., MacQueen, D.,
Niccolai, A., Lamigueiro, O.P., Short, T., Snow, G., Stabler, B., Turner,
R.: maptools: Tools for reading and handling spatial objects (2012),
http://CRAN.R-project.org/package=maptools, r package version 0.8-
20
[11] Ligges, U., Mächler, M.: Scatterplot3d - an r package for visualizing multivariate
data. Journal of Statistical Software 8(11), 1–20 (2003), http:
//www.jstatsoft.org
22
[12] Loecher, M.: ReadImages: Image Reading Module for R (2012), http:
//CRAN.R-project.org/package=ReadImages, r package version 0.1.3.2
[13] Meloun, M., Militký, J., Hill, M.: Počítačová analýza vícerozměrných dat
v příkladech, vol. 1. Praha: Academia (2005)
[14] Novotný, V.: Sex Determination of the Pelvic Bone: A Systems Approach.
Anthropologie 24, 197–206 (1986)
[15] Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., R Core Team: nlme: Linear
and Nonlinear Mixed Eﬀects Models (2018), https://CRAN.R-project.
org/package=nlme, r package version 3.1-137
[16] Plummer, M., Best, N., Cowles, K., Vines, K.: Coda: convergence diagnosis
and output analysis for mcmc. R News 6(1), 7–11 (March 2006), http:
//oro.open.ac.uk/22547/
[17] R Core Team: foreign: Read Data Stored by Minitab, S, SAS, SPSS, Stata,
Systat,dBase,... (2012), http://CRAN.R-project.org/package=foreign,
r package version 0.8-50
[18] R Core Team: R: A Language and Environment for Statistical Computing.
R Foundation for Statistical Computing, Vienna, Austria (2012), http:
//www.R-project.org/, ISBN 3-900051-07-0
[19] Rohlf, J.F.: tpsDig version 1.31. Program for facilitating the statistical
analysis of landmark data in morphometrics by making it easier to collect
and maintain landmark data from digitized images. Department of Ecology
and Evolution, State University of New York, Stony Brook, NY 11794-5245
(2001), http://Life.Bio.SUNYSB.edu/morph/morph.html
[20] Sarkar, D.: lattice: Multivariate Data Visualization with R. Springer,
New York (2008), http://lmdvr.r-forge.r-project.org, iSBN 978-0-
387-75968-5
[21] Turner, R.: deldir: Delaunay Triangulation and Dirichlet (Voronoi) Tessellation
(2012), http://CRAN.R-project.org/package=deldir, r package
version 0.0-21
[22] Urbanová, P.: A Study of Human Craniofacial Variation by Using Geometric
Morphometrics [online]. Disertační práce, Masarykova univerzita,
Přírodovědecká fakulta, Brno (2009 [cit 2018-09-06]), https://theses.cz/
id/hqz6vl/
[23] Urbanová, P., Králík, M., Svoboda, J., Malina, J.: 15. Appendix. Kvantitativní
popis tvaru pomocí metod geometrické morfometrie, pp. 277–291.
Nadace Universitas - Akademické nakladatelství CERM, Brno, 1. vyd. edn.
(2009)
[24] Venables, W.N., Ripley, B.D.: Modern Applied Statistics with S. Springer,
New York, fourth edn. (2002), http://www.stats.ox.ac.uk/pub/MASS4,
iSBN 0-387-95457-0
[25] Weihs, C., Ligges, U., Luebke, K., Raabe, N.: klar analyzing german business
cycles. In: Baier, D., Decker, R., Schmidt-Thieme, L. (eds.) Data
Analysis and Decision Support. pp. 335–343. Springer-Verlag, Berlin (2005)
23
Appendices
A Osteometric landmarks deﬁnitions
A.1 Pubic view
Figure 11: Overview of landmark positions in Pubic view on consensus shape
(left) and on real-life example (right).
Landmark deﬁnitions
• Landmark 1: Point on the outline of facies symphysialis where its anterior
edge refracts in the superior edge, or where the anterior and posterior
edges meet at the top of facies symphysialis.
• Landmark 2: The most protruding point on the outline of eminentia iliopubica
considering the main course of the upper outline of the bone; this
point corresponds approximately to the point where the obliterated/fused
suture between os pubis and os ilium meets the upper outline of the bone.
• Landmark 3: Point on the edge of os ilium below spina iliaca anterior
inferior (where the body of the bone becomes the wing), placed in the
deepest notch.
• Landmark 4: The most medially protruding point on the outline of spina
iliaca anterior inferior considering the main course of the anterior edge of
the bone.
24
• Landmark 5: Point below spina iliaca anterior superior where labium internum
and labium externum cristae iliacae meet and crista iliaca ends.
• Landmark 6: The most medially protruding point on the outline of crista
iliaca.
• Landmark 7: Point where the lateral outline of ala ossis ilii touches the
superior margin of acetabulum.
• Landmark 8: The uppermost point on the superior edge of acetabulum
placed on the top of this edge (i.e. the point corresponding to the ﬁne
edge between articular surface of acetabulum and external bone surface,
often corresponds to the most external position).
• Landmark 9: The lowest point on the inferior edge of acetabulum placed on
the top of the acetabular rim (i.e. the point corresponding to the ﬁne edge
between articular surface of acetabulum and external bone surface, often
corresponds to the most external position), analogously to the landmark
8 at the opposite side of the acetabulum.
• Landmark 10: The most medially protruding point by facies lunata placed
on the edge of acetabulum near incisura acetabuli.
• Landmark 11: The most laterally protruding point by facies lunata placed
on the pubic part of acetabulum; regardless of whether the point lies on
the edge of os pubis or facies lunata itself, in the latter case always on the
outer edge of this facet.
• Landmark 12: Point on the lateral outline of os ischii where it touches the
inferior edge of acetabulum in incisura ischiadica minor.
• Landmark 13: Point on the outline of os ischii between acetabulum and
tuber ischiadicum placed in the deepest notch.
• Landmark 14: Point on the outline of os ischii where it touches the edge of
tuber ischiadicum, i.e. by the superior end of tuber ischiadicum in incisura
ischiadica minor.
• Landmark 15: The most laterally protruding point on the outline of tuber
ischiadicum.
• Landmark 16: The lowest point on the inferior outline of tuber ischiadicum.
• Landmark 17: Point on the medial, resp. anterior termination of tuber
ischiadicum placed in a small notch frequently located there.
• Landmark 18: Point on the medial, resp. anterior beginning of crista
phallica placed in the deepest notch.
• Landmark 19: Point on the outline of facies symphysialis where its anterior
edge refracts in the inferior edge, or where the anterior and posterior edges
meet at the bottom of facies symphysialis.
25
• Landmark 20: Point on the outline of foramen obturatum in sulcus obturatorius
placed at the apex of the angle which is formed by the anterior and
superior, or posterior (if foramen obturatum is triangular-shaped) edges
of foramen obturatum.
• Landmark 21: Point on the outline of foramen obturatum where the inferior
and posterior edges of foramen obturatum meet; generally it is the
point on the outline of foramen obturatum opposite to landmark 20.
• Landmark 22: Point on the inferior outline of foramen obturatum which
is the closest to landmark 18.
• Landmark 23: The most medially lying point on the anterior edge of
foramen obturatum.
• Landmark 24: Point on the posterior edge of foramen obturatum placed
on the vertical from landmark 22 to the longitudinal axis of foramen obturatum
given by landmarks 20 and 21.
A.2 Iliac view
Figure 12: Overview of landmark positions in Iliac view on consensus shape
(left) and on real-life example (right).
Landmark deﬁnitions
• Landmark 1: Point on the outline of the superior arm of incisura ischiadica
major at the inﬂection point, where the smooth curve of the outline of the
superior arm changes its course, usually at the top of tuberculum musculi
piriformis.
26
• Landmark 2: Point on the outline of the inferior arm of incisura ischiadica
major on the superior base of spina ischiadica.
• Landmark 3: The lowest point placed of the outline of incisura ischiadica
major which is the furthest from the line connecting landmarks 1 and 2.
• Landmark 4: Point on the inferior base of spina ischiadica, placed on
the outline of incisura ischiadica minor where the opposed facing edges of
bone (upper edge corresponding to the lower outline of spina ischiadica,
the lower edge representing outline of incisura ischiadica minor) connect
in an obtuse angle.
• Landmark 5: Point where the outline of incisura ischiadica minor connects
with the medial edge of tuber ischiadicum.
• Landmark 6: Point on tuber ischiadicum lying at the intersection of its
inferior edge with its longitudinal axis.
• Landmark 7: Point on tuber ischiadicum lying at the intersection of its
superior edge with its longitudinal axis; opposite to landmark 6.
• Landmark 8: Point on the medial outline of tuber ischiadicum which
is the furthest from the longitudinal axis of tuber ischiadicum given by
landmarks 6 and 7.
• Landmark 9: Point on the lateral outline of tuber ischiadicum placed on
the vertical from point 8 to the longitudinal axis of tuber ischiadicum given
by landmarks 6 and 7: joining of landmarks 8 and 9 makes the transverse
axis of tuber ischiadicum.
• Landmark 10: Point where the edge of os ischii below acetabulum connects
the lateral edge of tuber ischiadicum; if the edges do not meet, the point
is placed on the outline of tuber ischiadicum where the edges are closest
to each other but always above landmark 9.
• Landmark 11: Point on the outline of os ischii between tuber ischiadicum
and acetabulum where the opposed facing contours of the bone connect
in an obtuse angle.
• Landmark 12: The lowest point on the inferior outline of acetabulum
placed on the most external position of the acetabular rim.
• Landmark 13: The highest point of the superior outline of acetabulum
placed on the most external position of the acetabular rim (i.e. the point
corresponding to the ﬁne edge between articular surface of acetabulum
and external bone surface, often corresponds to the most external position
on the rim); analogously to the landmark 12 at the opposite side of the
acetabulum.
• Landmark 14: The most protruding point on the outline of spina iliaca
anterior inferior considering the main course of the upper outline of the
bone, which is the furthest from the line passing by the inferior and superior
termination (base) of the contour of spina iliaca anterior inferior.
27
• Landmark 15: The most protruding point on the outline of spina iliaca
anterior superior, which is the furthest from the line passing by the inferior
and superior termination (base) of the contour of spina iliaca anterior
inferior.
• Landmark 16: The deepest point on the outline of os ilium between spina
iliaca anterior inferior and spina iliaca anterior superior: i.e. the furthest
point from the line connecting landmarks 14 and 15.
• Landmark 17: Point on the anterior edge of os ilium where it meets with
the external edge of labium externum cristae iliacae.
• Landmark 18: The most medially (in the image: to the left) protruding
point on the outline of spina iliaca posterior superior.
• Landmark 19: The most medially (in the image: to the left) protruding
point on the outline of spina iliaca posterior inferior, in the region where
the inferior arm of facies auricaluris (on the opposite side of the bone and,
therefore, not visible on the image) ends.
• Landmark 20: Point on the outline of spina iliaca posterior inferior where
the superior arm of incisura ischiadica major ends: on facies pelvina (on
the opposite site of the bone not visible in the image) it corresponds to
the point where the inferior edge of facies auricularis in its termination
point refracts upwards.
• Landmark 21: Point on the outline of crista iliaca which is the furthest
from the landmark 6
B A list of rejected dimensions for traditional
morphometry
B.1 Pubic view, Iliac view
By default (except for justiﬁed instances), we rejected dimensions that had
unsuitable features in terms of discrimination. HIP software oﬀers an option to
avoid these dimensions from computations of traditional morphometrics. These
dimensions are (a) whose values (or only a part of the values) are within a range
of less than 10 mm and are therefore more strongly aﬀected by a measurement
error, (b) and/or depend on the relationships between landmarks of diﬀerent
types of deﬁnitions (e.g., landmarks 9 and 10 in Iliac view) and, as a result, their
values vary considerably (from a few cm to zero), and/or (c) the distribution
of their values is asymmetrical with many outlying values, (d) and/or the male
and female values almost overlap (e.g. dimension 22-24 in Pubic view).
In Pubic view, we have rejected the use of these 20 dimensions for traditional
morphometry: 2-11, 3-4, 5-6, 7-8, 9-10, 9-12, 9-13, 12-13, 12-14, 13-14, 13-15,
14-15, 16-17, 16-18, 17-18, 20-23, 21-22, 21-23, 22-23, 22-24.
In Iliac view, we have rejected the use of these 16 dimensions for traditional
morphometry: 1-19, 1-20, 2-4, 5-7, 5-8, 7-9, 7-10, 9-10, 9-11, 10-11, 10-12, 11-12,
13-14, 14-16, 15-16, 19-20.
28
Figure 13: Rejected dimension in Pubic view (left) and Iliac view (right).
C Recommended conﬁgurations for geometric mor-
phometry
After digitizing each of the views, the program automatically checks which conﬁguration
is available. Then it calculates with the conﬁguration that has the
highest number of landmarks present. The process of conﬁguration settings is
even more complicated because some points are redundant – there are more of
them relatively close together. We have chosen this procedure so that in the
case of not being able to locate one of the landmarks (e.g. due to bone damage),
a diﬀerently deﬁned landmark nearby is available that describes the given part
of the bone similarly to the missing one. When selecting a landmark for the
appropriate conﬁguration, however, the program selects only one of the closest
landmarks – the most appropriate one. In the conﬁgurations below, these groups
are indicated in brackets. Within the group, the program selects from left to
right, if the group is missing a point on the right, it proceeds to the point closest
to the left. If there are no landmarks in a group of a given conﬁguration, the
program proceeds to a conﬁguration with a lower number of landmarks where
this group does not exist.
C.1 Pubic view
Optimal conﬁguration
Optimal conﬁguration is composed of 13 points: 1, 2, 4, 6, 8, 9, 14, 15, 16, 18,
19, 20, 22. If any of these points are not present, some of them may be replaced
by another point from their group, but always only one to make the total of 13:
29
Figure 14: Example of three conﬁgurations in Pubic view. The groups of landmarks
belonging to each other are surrounded by an ellipse. Only one point is
used from each group: priority is green, if it is not present, red is used, if it is
not present either, blue is used.
1, 2, (group 4 – 3), (group 6 – 5), (group 8 – 7), (group 9 – 10 – 11), (group 14
– 13 – 12), 15, (group 16 – 17), 18, 19, 20, (group 22 – 23). If the conﬁguration
is available, the program calculates with this conﬁguration of 13 points. If it is
not possible to create this optimal conﬁguration because of missing points, it
proceeds to the reduced conﬁguration.
Reduced conﬁguration
Reduced conﬁguration is composed of 9 points: 1, 2, 6, 8, 9, 15, 16, 18, 22. If
any of these points are not present, some of them may be replaced by another
point from their group, but always only one to make the total of 9: (group
1 – 19), 2, (group 6 – 5), (group 8 – 7), (group 9 – 10 – 11), 15, (group
16 – 17), 18, (group 22 – 23). If the conﬁguration is available, the program
calculates with this conﬁguration of 9 points. If it is not possible to create
this reduced conﬁguration because of missing points, it proceeds to the minimal
conﬁguration.
Minimal conﬁguration
Minimal conﬁguration is composed of 7 points: 1, 6, 8, 9, 16, 18, 22. If any of
these points are not present, some of them may be replaced by another point
from their group, but always only one to make the total of 7: (group 1 – 19),
(group 6 – 5), (group 8 – 7), (group 9 – 10 – 11), (group 16 – 17), 18, (group
22 – 23). If the conﬁguration is available, the program calculates with this
conﬁguration of 7 points. If the conﬁguration is not available, the program
announces that analysis by the method of geometric morphometry cannot be
performed.
30
C.2 Iliac view
Figure 15: Example of ﬁve conﬁgurations in Iliac view. The groups of landmarks
belonging to each other are surrounded by an ellipse. Only one point is used
from each group: priority is green, if it is not present, red is used, if it is not
present either, blue is used.
Optimal conﬁguration
Optimal conﬁguration is composed of 13 points: 1, 2, 3, 6, 7, 8, 9, 12, 13, 15,
17, 18, 19. If any of these points are not present, some of them may be replaced
by another point from their group, but always only one to make the total of
13: 1, 2, 3, 6, 7, (group 8 – 5), (group 9 – 10), (group 12 – 11), (group 13 – 14
– 16), 15, 17, 18, (group – 20). If the conﬁguration is available, the program
calculates with this conﬁguration of 13 points. If it is not possible to create
this optimal conﬁguration because of missing points, it proceeds to the reduced
conﬁguration A.
Reduced conﬁguration A
Reduced conﬁguration A is composed of 11 points: 1, 2, 3, 6, 7, 8, 9, 12, 13,
15, 18. If any of these points are not present, some of them may be replaced
31
by another point from their group, but always only one to make the total of
11: 1, 2, 3, 6, 7, (group 8 – 5), (group 9 – 10), (group 12 – 11), (group 13 – 14
– 16), (group 15 – 17), (group 18 – 19 – 20). If the conﬁguration is available,
the program calculates with this conﬁguration of 11 points. If it is not possible
to create this reduced conﬁguration A because of missing points, it proceeds to
the reduced conﬁguration B.
Reduced conﬁguration B
Reduced conﬁguration B is composed of 10 points: 1, 2, 3, 6, 7, 8, 9, 12,
13, 15. If any of these points are not present, some of them may be replaced
by another point from their group, but always only one to make the total of
10: 1, 2, 3, 6, 7, (group 8 – 5), (group 9 – 10), (group 12 – 11), (group 13
– 14 – 16), (group 15 – 17). If the conﬁguration is available, the program
calculates with this conﬁguration of 10 points. If it is not possible to create this
reduced conﬁguration B because of missing points, it proceeds to the reduced
conﬁguration C.
Reduced conﬁguration C
Reduced conﬁguration C is composed of 8 points: 1, 6, 7, 8, 9, 12, 13, 15. If any
of these points are not present, some of them may be replaced by another point
from their group, but always only one to make the total of 8: 1, 6, 7, (group 8
– 5), (group 9 – 10), (group 12 – 11), (group 13 – 14 – 16), (group 15 – 17). If
the conﬁguration is available, the program calculates with this conﬁguration of
8 points. If it is not possible to create this reduced conﬁguration C because of
missing points, it proceeds to the minimal conﬁguration.
Minimal conﬁguration
Minimal conﬁguration is composed of 5 points: 1, 6, 7, 13, 15. If any of these
points are not present, some of them may be replaced by another point from
their group, but always only one to make the total of 5: 1, 6, 7, (group 13 – 14
– 16), (group 15 – 17). If the conﬁguration is available, the program calculates
with this conﬁguration of 5 points. If the conﬁguration is not available, the
program announces that analysis by the method of geometric morphometry
cannot be performed.
32