Sie sind auf Seite 1von 7

Journal of Forensic and Legal Medicine 63 (2019) 11–17

Contents lists available at ScienceDirect

Journal of Forensic and Legal Medicine


journal homepage: www.elsevier.com/locate/yjflm

Research Paper

Accuracy and reliability of virtual femur measurement from CT scan T


a,b c d
Normaizatul Afizah Ismail , Nurliza Abdullah , Mohamad Helmee Mohamad Noor ,
Poh Soon Laic, Mohamed Swarhib Shafieb, Faridah Mohd Norb,∗
a
Kulliyyah of Medicine and Health Sciences, Universiti Islam Antarabangsa Sultan Abdul Halim Mu'adzam Shah, 09300, Kedah, Malaysia
b
Universiti Kebangsaan Malaysia Medical Centre, Jalan Yaacob Latif, Bandar Tun Razak, 56000, Cheras, Kuala Lumpur, Malaysia
c
National Institute of Forensic Medicine, Hospital Kuala Lumpur, Ministry of Health Malaysia, Jalan Pahang, 50586, Kuala Lumpur, Malaysia
d
Radiology Department, Hospital Kuala Lumpur, Ministry of Health Malaysia, Jalan Pahang, 50586, Kuala Lumpur, Malaysia

ARTICLE INFO ABSTRACT

Keywords: Background: In the application of scientific human skeletal variation in medico-legal matters, virtual anthro-
Virtual method pology is the current technique performed to examine skeleton and its body parts. Hence, this study was con-
Forensic anthropology ducted to assess the accuracy and reliability of virtual femur measurement through intra and inter-observer error
Bone analysis, and comparison was made between the virtual and conventional methods.
Malaysia
Methods: A total of 15 femora were examined with four parameters i.e. maximum length of femur (FeMl),
Forensic
diameter of femoral head (FeHd), transverse diameter of midshaft (FeMd) and condylar breadth (FeCb).
Osteometric board and vernier calipers were employed for the conventional method, while CT reconstructed
images and Osirix MD software was utilised for the virtual method.
Results: Results exhibited that there were no significant differences in the measurements by conventional and
virtual methods. There were also no significant differences in the measurements by the intra or inter-observer
error analyses. The intraclass correlation coefficients (ICC) were more than 0.95 by both intra and inter-observer
error analyses. Technical error of measurement had displayed values within the acceptable ranges (rTEM < 0.08
for intra-observer, < 2.25 for inter-observer), and coefficient of reliability (R) indicated small measurement
errors (R > 0.95 for intra-observer, R > 0.92 for inter-observer). By parameters, FeMl showed the highest R
value (0.99) with the least error in different methods and observers (rTEM = 0.02–0.41%). Bland and Altman
plots revealed points scattered close to zero indicating perfect agreement by both virtual and conventional
methods. The mean differences for FeMl, FeHd, FeMd and FeCb measurements were 0.01 cm, −0.01 cm, 0.02 cm
and 0.01 cm, respectively.
Conclusion: This brought to suggest that bone measurement by virtual method was highly accurate and reliable
as in the conventional method. It is recommended for implementation in the future anthropological studies
especially in countries with limited skeletal collection.

1. Background In Malaysia, virtual autopsy was introduced in 2010 by the Ministry


of Health to the National Institute of Forensic Medicine, Hospital Kuala
Virtual anthropology is a new term coined by researchers, high- Lumpur for a trial research project in assessing the effectiveness of
lighting the importance of skeletal imaging technique as a tool to assist virtual autopsy. With the trial project, all post-mortem computed to-
in the investigation of forensic anthropology cases.1–4 The field of vir- mography (PMCT) images of forensic cases were kept as virtual records.
tual anthropology has emerged from the routine application of multi- This has become a great source for further researches in the field of
detector computed tomography (MDCT) in forensic medicine and for- anthropology. Since skeletal collection in Malaysia is very limited due
ensic anthropology in the mid-1990s.3–6 Recently, CT has been used to law practices and religious restrictions, research in anthropology has
routinely in several medico-legal institutions across the world, where remained inactive. It is hoped that virtual bone images will become an
forensic anthropologists are regularly being exposed to post-mortem CT alternative method, and more research can be conducted, particularly
scan.7–9 on virtual bone measurements.


Corresponding author. Pathology Department, Universiti Kebangsaan Malaysia Medical Centre, Jalan Yaacob Latif, Bandar Tun Razak, 56000, Cheras, Kuala
Lumpur, Malaysia.
E-mail address: faridah.nor@ukm.edu.my (F.M. Nor).

https://doi.org/10.1016/j.jflm.2019.02.010
Received 19 September 2018; Received in revised form 9 February 2019; Accepted 15 February 2019
Available online 21 February 2019
1752-928X/ © 2019 Elsevier Ltd and Faculty of Forensic and Legal Medicine. All rights reserved.
N.A. Ismail, et al. Journal of Forensic and Legal Medicine 63 (2019) 11–17

Practice on anthropological bone measurement was done by the 2.1. Statistical analysis
conventional methods i.e. osteometric board and calipers, which is the
gold standard in forensic anthropology. However, the introduction of Normality of the study sample was performed by the Shapiro-Wilk's
multi-detected computed tomography (MDCT) in forensic service had test as the sample size was small (n = 15) and the data were analysed
inevitably offered an alternative in bone measurements. The introduc- by SPSS v. 23.0 and Excel, Microsoft Office 2013. The intra and inter-
tion of MDCT has allowed a significant improvement in three dimen- observer differences were analysed by paired t-test and ANOVA in GLM
sional (3D) image resolution, which can be visualized on dedicated (General Linear Model). The absolute differences between two sets of
software in x, y and z planes. The benefit is great, and studies are done measurements were calculated to obtain maximum absolute difference.
to explore its full potential. As the result, researchers had resorted to In assessing the reliability of measurements for intra and inter-observer
radiological method to conduct bone measurements on CT images.10–18 error, the intraclass correlation coefficient (ICC) was performed, and
Previously, in order to quantitatively assess the accuracy of the the Bland and Altman plot was generated for inter-method measure-
existing anthropological methods and/or to formulate new standards, ments. In ICC, correlation coefficient was obtained to demonstrate the
human bone collections were accessed; however, with the availability strength of the relationship (similarities) between two measurements.
of large digital samples, this requirement is to be abolished.2 Introdu- The value for reliability coefficient ranges from 0 to 1. A coefficient
cing a new method (virtual measurement) over the current method value below 0 indicates ‘no reliability’, > 0 to < 0.2 is slight reliability,
(conventional measurement) requires a series of test on accuracy and 0.2 - < 0.4 is fair reliability, 0.4 - < 0.6 is moderate, 0.6 - < 0.8 is
reliability,19,20 and this is important to ensure that these measurements substantial, while 0.8–1.0 is almost perfect reliability.25
would produce the same result. Accuracy is the term used to express The Bland and Altman plots were used to provide an illustration of
deviation of measurement from its true value, while reliability is the the spread of differences in measurements, mean difference and upper
degree of variation in repeated measures, which is also known as pre- and lower limits of agreement to compare between the virtual and
cision.21–24 If the measurements of both methods are accurate and re- conventional measurements. The Bland and Altman plot is a method
liable, the new method may replace the established method. In this comparison technique, and the averages of two measurements are
case, the virtual method is preferable as it is non-invasive, and will plotted along the horizontal axis, and the differences between two
solve the issue of scarce bone collection. methods are plotted along the vertical axis. There is no such ‘accep-
It is the aim of this study to assess the accuracy and reliability of table’ range for Bland and Altman plots. If each measurement method
femur measurements by virtual method and conventional methods. produces accurate measurements, the only difference between the
Intra- and inter-observational error analyses were important evalua- means would be due to random error, and any systematic difference is
tions in this study. The results of this pilot study will set a new direction called the bias. The bias between two methods is measured by taking
to future researches in virtual anthropology in Malaysia. the mean difference of values. The upper and lower limit of agreement
is calculated by using Equations (1) and (2) as follows:

Upper limit of agreement = + 1.96 x SD (1)


2. Materials and methods
Lower limit of agreement = + 1.96 x SD (2)
A pilot study was conducted on 15 femora, which was intact and
free from deformities, trauma or diseases. The samples were taken from The assessment of accuracy for intra and inter-observer was ana-
multitude of bone collections with unknown identity from the National lysed by technical error of measurements (TEM), relative TEM (rTEM)
Institute of Forensic Medicine (IPFN). The femora were scanned by and coefficient of reliability (R). The TEM is an accuracy index, which
PMCT using TOSHIBA Aquilion 64 CT Scanner (Toshiba Medical calculates standard deviation between repeated measures. The lower
System, Japan), with detector collimation of 1 mm, pitch 0f 0.844. The the TEM value, the higher the reliability of the results. The coefficient of
exposure setting for tube voltage was 120 kVp with automatic exposure reliability (R) ranges from 0 (not reliable) to 1 (completely reliability),
control for tube current and FOV of 320 mm. All images were re- and R value above 0.95 is indicative of negligible errors.23 The standard
constructed with 2.0 mm thickness and slice interval of 1.6 mm. Prior to of ‘acceptable measurement error’ varies depending on the magnitude
the study, ethical approval was obtained from the Institutional Ethics of the effect one is testing. In anthropological measurements, any
Committee of the University. measurement with an R value below 0.90 should be treated with due
Observers in this study were selected based on their skill and ex- caution, especially if the associated rTEM is more than 5%.26
perience level. One observer was an anthropological expert (Observer The formulas for TEM calculation are shown in Equations (3)–(5).
A), another observer was well-trained in bone measurement (Observer 26,27
In equation (3), absolute TEM will be generated, where the Σd2 is
B), while the third observer was a novice (Observers C). For inter-ob- the summation of deviations raised to the power of two, and n is the
server error analysis, measurements by the conventional method were number of samples. Then, the absolute TEM is transformed into relative
performed by Observer A and B, whilst measurements on virtual TEM (using Equation (4)) in order to obtain the percentage of error
method were performed by Observer B and C. Each observer for intra- corresponding to the total variable average value (VAV). The VAV is
observer error analysis replicated the measurements twice on two dif- obtained by summing up the average for both sets of measurements,
ferent days, and an average was taken in centimetre (cm). and divided by two. In Equation (4), rTEM is the relative technical error
In conventional method, measurements were taken by osteometric of measurement expressed in percentage, and VAV is the variable
board (WARD’S Forensics, with a precision of 0.1 cm) and vernier ca- average value. In order to get the R value, Equation (5) is used, where R
lipers (TED PELLA, INC, with a precision of 0.01 cm). In virtual method, is the coefficient of reliability, TEM is the technical error of measure-
measurements were taken by a digital ruler in Osirix MD (precision of ment, and SD is the standard deviation.
0.01 cm). The OsiriX MD V.8.5 was installed in a 27-inch iMac com-
puter, and measurements were performed on 3D volume rendering d²
TEM =
images. 2n (3)
The parameters were maximum length of femur (FeMl), diameter of
femoral head (FeHd), transverse diameter of midshaft (FeMd) and rTEM (%) =
TEM
x 100
condylar breadth (FeCb). Definitions of all the landmarks were de- VAV (4)
scribed (Table 1). Illustrations of parameters in both conventional and
virtual methods were given (Figs. 1 and 2). (TEM ) 2
R=1
SD 2 (5)

12
N.A. Ismail, et al. Journal of Forensic and Legal Medicine 63 (2019) 11–17

Table 1
Definition of landmarks for all parameters.
Label Parameters Definition of landmarks

FeMl Maximum length of femur Distance from the most superior point on the head of femur to the most inferior point on the distal condyles.
FeHd Maximum diameter of femoral head The maximum diameter of femoral head, wherever it occurs.
FeMd Transverse diameter of midshaft Distance between the medial and lateral surfaces of midshaft of femur.
FeCb Condylar breadth Distance between the two most laterally projecting points on the two condyles of femur.

Table 2
Descriptive statistics for all parameters in conventional and virtual methods.
Parameters Conventional Method Virtual Method

Observer A Observer B Observer B Observer C

FeMl Min 41.70 41.70 41.72 41.67


Max 47.70 47.70 47.73 47.71
Mean 43.50 43.47 43.48 43.42
SD 1.63 1.63 1.63 1.66
FeHd Min 3.70 3.72 3.69 3.63
Max 5.00 5.02 5.02 4.90
Mean 4.31 4.33 4.32 4.27
SD 0.36 0.37 0.37 0.35
FeMd Min 2.10 2.08 2.12 2.15
Max 2.80 2.86 2.82 2.96
Mean 2.44 2.43 2.45 2.48
SD 0.19 0.19 0.18 0.19
FeCb Min 6.70 6.72 6.68 6.66
Max 8.80 8.83 8.83 8.84
Mean 7.75 7.78 7.78 7.79
SD 0.55 0.54 0.55 0.52

Fig. 1. The parameters as measured by the conventional method. *FeMl - maximum length of femur, FeHd - maximum diameter of femoral head,
FeMd - transverse diameter of midshaft, FeCb - condylar breadth, min–
minimum, max – maximum, SD− standard deviation.

significant difference in the measurements between different observers


and methods.

3.2. Interclass correlation coefficient

The maximum absolute difference of measurement in intra and


inter-observer error analyses were calculated for each parameter
(Table 3). The highest absolute difference was observed in the inter-
observer error analysis for the virtual method (FeMl = 0.96 cm). The
intraclass coefficients (ICC) for all parameters were mostly above 0.95,
with CI mostly more than 0.94 suggesting a strong correlation between
measurements within the same observer and between observers,
pointing to a high degree of reliability between the observers in their
measurements.

3.3. Technical Error Measurement and reliability test

Fig. 2. The parameters as measured by the virtual method. The accuracy of measurements was assessed by ‘Technical Error
Measurement’ (TEM) (Table 4). The rTEM for intra-observer error
analysis was below 0.08%, which indicated negligible error. By inter-
3. Results observer error analysis, similar findings were observed in rTEM, where
most values were below 2.25%, which were highly reliable (R > 0.92).
3.1. Descriptive statistics Of all parameters, FeMl showed the least error by different methods and
observers (rTEM ranging from 0.02 to 0.41%) and the highest R value
The data were normally distributed for all parameters (FeMl, FeHd, (0.99).
FeMd, FeCb) by Shapiro-Wilk's test (p > 0.05). The descriptive statis-
tics for all parameters were tabulated separately for each observer using 3.4. Bland and Altman plots
the average values (Table 2). Overall, the measurements by Observer C
were slightly deviated from measurements by the other observers. The Bland and Altman plots were presented for all parameters to com-
intra-observer error analysis was further analysed by paired t-test, and pare measurements made by the virtual and conventional methods
the results showed no significant difference in both methods (Fig. 3). The plots showed that the measurement difference were all
(p > 0.05). The inter-observer error analysis was performed by within the upper and lower limits of agreement. The FeMl measure-
ANOVA in general linear model (GLM), and results showed no ments by virtual method were consistent with those by conventional

13
N.A. Ismail, et al. Journal of Forensic and Legal Medicine 63 (2019) 11–17

Table 3
Intraclass correlation (ICC) for intra- and inter-observer measurement by two different methods.
Parameters Intra-observer Inter-observer

Conventional (Observer A) Virtual (Observer B) Conventional (Observers A & B) Virtual (Observers B & C)

FeMl Maximum 0.10 0.18 0.10 0.96


ICC 1.00 1.00 1.00 0.99
CI > 0.99 > 0.99 > 0.99 > 0.98
FeHd Maximum 0.07 0.28 0.07 0.17
ICC 0.99 0.98 0.99 0.98
CI > 0.99 > 0.94 > 0.99 > 0.94
FeMd Maximum 0.10 0.14 0.10 0.14
ICC 0.98 0.95 0.98 0.95
CI > 0.95 > 0.86 > 0.95 > 0.86
FeCb Maximum 0.05 0.05 0.11 0.25
ICC 0.99 0.99 0.99 0.99
CI > 0.99 > 0.99 > 0.99 > 0.98

FeMl - maximum length of femur, FeHd - maximum diameter of femoral head, FeMd - transverse diameter of midshaft, FeCb - condylar breadth, min – minimum, max-
maximum, ICC-Intraclass correlation, CI-Confidence interval.

Table 4
The technical error measurements (TEM) and reliability test for intra- and inter-observer error analysis.
Parameters Analysis Intra-observer Inter-observer

Conventional (Observer B) Virtual (Observer B) Conventional (Observer A & B) Virtual (Observer B & C)

FeMl rTEM 0.03 0.02 0.10 0.41


R 0.99 0.99 0.99 0.99
FeHd rTEM 0.02 0.04 0.64 1.22
R 0.99 0.99 0.99 0.98
FeMd rTEM 0.01 0.08 0.30 2.25
R 0.99 0.99 0.99 0.92
FeCb rTEM 0.04 0.04 0.69 0.81
R 0.99 0.99 0.99 0.99

FeMl - maximum length of femur, FeHd - maximum diameter of femoral head, FeMd - transverse diameter of midshaft, FeCb - condylar breadth, rTEM - relative
technical error of measurement, R - coefficient of reliability.

method with an average difference of 0.01 cm (upper limit of 0.05 cm Nevertheless, there was one measurement in FeMd (0.09 cm difference) that
and lower limit of −0.03 cm). The average differences for FeHd, FeMd was slightly above the upper limit (0.087 cm) (Fig. 3). Since this assessment
and FeCb measurements were −0.01 cm, 0.02 cm and 0.01 cm, re- is not for clinical diagnostic purposes, difference as much as 0.10 cm is
spectively. The values were distributed closely to zero, showing almost acceptable considering measurement for FeMd and FeHd are rather chal-
perfect data agreement. lenging compared to FeMl and FeCb. The findings could not be compared
with other studies due to paucity of similar studies.
4. Discussion In the literature, the accuracy of virtual bone measurements by
PMCT was found to be accurate and reliable2,14,28–32 but none has
This study assessed the accuracy and reliability of the virtual been done using femur. Among the earliest study, analyse the accuracy
method by intraclass correlation coefficient (ICC), technical error of of skull measurement between CT image reconstruction from CT and
measurement (TEM), relative TEM (%TEM), coefficient of reliability (R) measurements by calipers32. The result disclosed no significant differ-
and Bland and Altman plots. Results showed that virtual method was ence between both methods. However, the virtual method should be
highly accurate and reliable in bone measurements, even though it was used with caution, as it was stated that measurement error of 5 mm or
evident that the conventional method was more accurate in its mea- more were not tolerated. These findings had opened new avenues for
surements, as shown by less value variation made between Observers A future research in anthropologic imaging studies.
and B. An interesting study had demonstrated the accuracy and reliability
Interestingly, Observer C was found to be the contributor for the of osteometric measurements from dried skeletal elements and from CT
highest absolute difference (FeMl = 0.96 cm) and the highest rTEM images, with and without soft tissues.33 In the study, seven crania were
(FeMd = 2.25%) between Observer B and C in virtual method. utilised and enumerated by three observers. The findings demonstrated
Nevertheless, these values were acceptable for the purpose of anthro- mean percentage of differences between measurements from three data
pometric measurements. The value may be improved further, con- sources ranging from 1.4% to 2.9%. The accuracy of these measure-
sidering that Observer C was a novice, and the measurements made in ments was evaluated by technical error of measurement (TEM) and
other parameters have only slightly deviated from the other observers. relative technical error of measurement (rTEM). The intra-observer
These findings indicated that training and experience are necessary for TEM and rTEM in three observers and the craniometric variables
producing accurate measurements. ranged between 0.46 mm and 0.77 mm, and 0.56% and 1.06%, re-
The ICCs for all parameters were more than 0.95, and ICC for FeMl was spectively. The inter-observer TEM and rTEM for craniometric variables
1.00 suggesting a perfect correlation by both virtual and conventional were 2.6 mm and 2.26%, respectively. Overall, minimal differences
methods in the intra and inter-observer error analyses. The Bland and were observed from these data sources, and high accuracy was noted
Altman plots for all parameters exhibited that the points were distributed between the observers, which proved that CT image is an acceptable
within the upper and upper limits of agreement for FeMl, FeHd and FeCb. technique to measure the osteometric variables.33

14
N.A. Ismail, et al. Journal of Forensic and Legal Medicine 63 (2019) 11–17

Fig. 3. Bland and Altman plots for FeMl, FeHd, FeMd and FeCb.

Researches were also conducted on a comparison study of skull varies depending on the reconstructed image created in the 3D volume
measurements by calipers and 3D CT images.30 Results had demon- rendering. It is recommended that a maximum slice thickness of
strated no significant differences between the conventional method 1.25 mm may be utilised, when 3D reconstructions of anatomical fea-
(calipers) and virtual method (3D CT images). Similar findings were tures or other fine details are desired.34 The higher the slice thickness,
revealed by using clavicle31 and complete skeletal measurements.33 the lesser the resolution quality. Images of low resolution quality will
Majority of these studies disclosed high accuracy in measurements affect measurements due to unclear visibility of borders, lines and
derived from dried bone element and also, from 3D-CT image of the points of landmarks. The adjustable contrast setting on the screen may
bone.31,32 Several studies have further compared measurements from sometimes play an important role.
bone with soft tissues, and measurements of the same bone following Another contributing factor is the precise definition of landmarks. In
removal of soft tissues.4,8 Another study, stated that bone measure- the literature, there were several definitions given by different re-
ments on PMCT images were accurate, albeit with significant difference searchers.12,35 One needs to apply only one definition, and adhere to
in measurements.14 These measurements were believed to be accurate the definition consistently in every measurement. The most common
i.e. with maximum difference of 0.6 mm in variable measurement, and error in measurement was evident from the midshaft diameter for in-
a difference of approximately 1.5 mm in stature estimation.14 stance, the maximum diameter of midshaft, wherever it occurs, and
Several factors such as image quality, precision of landmarks and another parameter was the maximum diameter at the most lateral to the
availability of trained personnel may contribute to the reproducibility medial point at midshaft of femur.
of bone measurements by the virtual method. Image quality in CT scan Finally, the availability of trained personnel or observer is important.

15
N.A. Ismail, et al. Journal of Forensic and Legal Medicine 63 (2019) 11–17

The observer must be well-trained in positioning the bone anatomically i.e. disaster: experience gained from the 2009 Victorian bushfires. Forensic Sci Int.
in placing the landmarks36 and measuring the parameter. A random ob- 2011;205(1):15–28 2011/02/25/.
2. Franklin D, Swift L, Flavel A. 'Virtual anthropology' and radiographic imaging in the
server with insufficient knowledge in computer software will not perform forensic medical sciences. Egypt J Food Sci. 2016;6:31–43.
well. In Osirix MD, the ‘anatomical plane’ will give direction on the ana- 3. Dedouit F, Telmon N, Costagliola R, Otal P, Joffre F, Rougé D. Virtual anthropology
tomical view of the body. The positioning process will be easy, if it involves and forensic identification: report of one case. Forensic Sci Int.
2007;173(2–3):182–187 12/20/.
a single bone. In PMCT image, the body was often improperly positioned, 4. Verhoff MA, Ramsthaler F, Krähahn J, et al. Digital forensic osteology—possibilities
leading to difficulties in its measurements. The observer needs to be skilful in cooperation with the Virtopsy® project. Forensic Sci Int. 2008;174(2):152–156
in positioning the image by manually operating to its correct anatomical 2008/01/30/.
5. Wüllenweber R, Schneider V, Grumme T. [A computer-tomographical examination of
position. However, the observer should not rely on the ‘anatomical plane’ cranial bullet wounds (author's transl)]. 1977/11// Z Rechtsmed.
guide because it usually refers to the anatomical position of the whole 1977;80(3):227–246 PubMed PMID: 602451. ger.
image, and not the bone in specific. 6. Sidler M, Jackowski C, Dirnhofer R, Vock P, Thali M. Use of multislice computed
tomography in disaster victim identification—advantages and limitations. Forensic
The use of CT image bears the advantage of elimination of the soft
Sci Int. 2007;169(2):118–128 2007/07/04/.
tissue around the bone,8,31 which enables for measurements to be taken 7. Torimitsu S, Makino Y, Saitoh H, et al. Stature estimation in a contemporary
accurately than with direct bone measurements. In the standard pro- Japanese population based on clavicular measurements using multidetector com-
tocol of direct bone measurement, the soft tissue need to be removed puted tomography. Forensic Sci Int. 2017;275:316 e1-.e6.
8. Robinson C, Eisma R, Morgan B, et al. Anthropological measurement of lower limb
beforehand, which can be tedious and time-consuming. Thus, it is clear and foot bones using multi-detector computed tomography. J Forensic Sci.
that CT image has an inherent advantage compared to dry bone mea- 2008;53(6):1289–1295.
surements. Besides being quick, CT scan is also non-invasive. 9. Grabherr S, Cooper C, Ulrich-Bochsler S, et al. Estimation of sex and age of “virtual
skeletons”–a feasibility study. Eur Radiol. 2009;19(2):419–429.
This study was a pilot study, which requires further research to 10. Hishmat AM, Michiue T, Sogawa N, et al. Virtual CT morphometry of lower limb long
validate the findings with extended samples. Limitation of this study bones for estimation of the sex and stature using postmortem Japanese adult data in
involves the limited femora available from the institute. Besides, the forensic identification. Int J Leg Med. 2015;129(5):1173–1182.
11. Karakas HM, Celbis O, Harma A, Alicioglu B. Total body height estimation using
identities were unknown as to their specific sexes or ages, and that only sacrum height in Anatolian Caucasians: multidetector computed tomography-based
Observer B had the chance to measure bones in both virtual and con- virtual anthropometry. Skeletal Radiol. 2011;40(5):623–630.
ventional methods. It is recommended that this study would be ex- 12. Torimitsu S, Makino Y, Saitoh H, et al. Stature estimation based on radial and ulnar
lengths using three-dimensional images from multidetector computed tomography in
panded to other skeletal parts with larger sample size by multiple ob- a Japanese population. Leg Med. 2014;16(4):181–186 7//.
servers in all methods available for better comparison. 13. Torimitsu S, Makino Y, Saitoh H, et al. Stature estimation in Japanese cadavers based
on pelvic measurements in three-dimensional multidetector computed tomographic
images. Int J Leg Med. 2014;129(3):633–639.
5. Conclusion
14. Zech W-D, Näf M, Siegmund F, Jackowski C, Lösch S. Body height estimation from
post-mortem CT femoral F1 measurements in a contemporary Swiss population. Leg
In this study, femora measurements were found to be highly accu- Med. 2016;19:61–66 3//.
rate and reliable by both conventional and virtual methods. There were 15. Sarajlić N, Cihlarž Z, Klonowski EE, Selak I. Stature estimation of Bosnian male
population. Bosn J Basic Med Sci. 2006;6(1):62–67.
good correlation coefficients (more than 0.95) in both intra and inter- 16. Hasegawa I, Uenishi K, Fukunaga T, Kimura R, Osawa M. Stature estimation for-
observer error analyses by ICC. The technical error of measurement mulae from radiographically determined limb bone length in A modern Japanese
displayed negligible errors (rTEM < 0.04% for intra-observer and < population. Leg Med. 2009;11(6):260–266.
17. Farsinejad M, Rasaneh S, Zamani N, Jamshidi F. Relationship between the stature
2.25% for inter-observer), and the coefficient of reliability (R) in- and the length of long bones measured from the X-rays; modifed trotter and gleser
dicated small error in the measurements (> 0.95). Bland and Altman formulae in Iranian population: a preliminary report. Soud Lek. 2014;59(2) 20-11.
plots were within the upper and lower limits of agreements, and the 18. Amal Hayati ZA, Khairul O, Pawita Albakri AH Sri, Noor Hazfalinda H, Normaizatul
Afizah I. Stature approximation of malays, Chinese and Indian in Malaysia using
data were closely distributed to zero suggesting an almost perfect radiographs of femur, Tibia and Fibula. Jurnal Sains Kesihatan Malaysia.
agreement. 2011;9(1):45–50.
This study has explored the benefits of virtual anthropology by in- 19. Dirkmaat DC, Cabo LL, Ousley SD, Symes SA. New perspectives in forensic anthro-
pology. Am J Phys Anthropol. 2008;137(47):33–52.
troducing an innovative, standardized method as an alternative method
20. Guyomarc’h P, Santos F, Dutailly B, Desbarats P, Bou C, Coqueugniot H. Three-di-
to dry bone osteometric technique. Virtual anthropology is an advanced mensional computer-assisted craniometrics: a comparison of the uncertainty in
technique with great possibility, and is highly recommended for future measurement induced by surface reconstruction performed by two computer pro-
grams. Forensic Sci Int. 2012;219(1):221–227 2012/06/10/.
anthropological research studies, especially in Malaysia, where bone
21. Goto R, Mascie-Taylor CG. Precision of measurement as a component of human
collections are restricted. variation. J Physiol Anthropol. 2007 Mar;26(2):253–256 PubMed PMID: 17435374.
Epub 2007/04/17. eng.
Acknowledgements 22. Ousley S. Should we estimate biological or forensic stature? J Forensic Sci.
1995;40(5):768–773.
23. Ulijaszek SJ, Kerr DA. Anthropometric measurement error and the assessment of
We would like to thank the Director General of Health Malaysia for nutritional status. Br J Nutr. 2007;82(3):165–177 Epub 03/01.
his permission to publish this article. Special thanks to Datuk Dr. Mohd 24. Adams B, Byrd J. Interobserver Variation of Selected Postcranial Skeletal Measurements.
2002; 2002.
Shah Mahmood (Head of Forensic Department), Mr. Mohd Tarmizi 25. Jamaiyah H, Geeta A, Safiza MN, et al. Reliability, technical error of measurements
Othman (radiographer), Mrs. Nur Shazuwani Ruzlan (Scientific Officer) and validity of length and weight measurements for children under two years old in
and all the staff in the National Institute of Forensic Medicine, Hospital Malaysia. Med J Malaysia. 2010;65(Suppl A):131–137 PubMed PMID: 21488474.
26. Perini TA, Oliveira GLd, Ornellas JdS, Oliveira FPd. Technical error of measurements
Kuala Lumpur for their assistance and expert advice in this study. A in anthropometry. Rev Bras Med do Esporte. 2005;11(1):86–90.
special thanks goes to UKM for funding this research project (Grant No: 27. Goto R, Mascie-Taylor CGN. Precision of measurement as a component of human
FRGS/1/2015/SKK13/UKM/02/1). variation. J Physiol Anthropol. 2007;26(2):253–256.
28. Giurazza F, Del Vescovo R, Schena E, et al. Determination of stature from skeletal and
skull measurements by CT scan evaluation. 10/10/ Forensic Sci Int.
Appendix A. Supplementary data 2012;222(1–3):398 e1-.e9.
29. Lorkiewicz-Muszyńska D, Kociemba W, Żaba C, et al. The conclusive role of post-
mortem computed tomography (CT) of the skull and computer-assisted super-
Supplementary data to this article can be found online at https://
imposition in identification of an unknown body. Int J Leg Med. 2013 May
doi.org/10.1016/j.jflm.2019.02.010. 01;127(3):653–660.
30. Sakuma A, Ishii M, Yamamoto S, et al. Application of postmortem 3d‐CT facial re-
References construction for personal identification. J Forensic Sci. 2010;55(6):1624–1629.
31. Brough AL, Bennett J, Morgan B, Black S, Rutty GN. Anthropological measurement of
the juvenile clavicle using multi‐detector computed tomography—affirming relia-
1. O'Donnell C, Iino M, Mansharan K, Leditscke J, Woodford N. Contribution of post- bility. J Forensic Sci. 2013;58(4):946–951.
mortem multidetector CT scanning to identification of the deceased in a mass 32. Hildebolt CF, Vannier MW, Knapp RH. Validation study of skull three‐dimensional

16
N.A. Ismail, et al. Journal of Forensic and Legal Medicine 63 (2019) 11–17

computerized tomography measurements. Am J Phys Anthropol. and Imaging. 2016;4:43–46 March 2016.
1990;82(3):283–294. 35. Bass WM. Human Osteology: A Laboratory and Field Manual. fifth ed. Missouri
33. Stull KE, Tise ML, Ali Z, Fowler DR. Accuracy and reliability of measurements ob- Archaelogical Society: University of Columbia; 2005.
tained from computed tomography 3D volume rendered images. Forensic Sci Int. 36. Lee S, Gong HH, Hyun JY, et al. Estimation of stature from femur length measured
2014;238:133–140. using computed tomography after the analysis of three-dimensional characteristics of
34. Ford JM, Decker SJ. Computed tomography slice thickness and its effects on three- femur bone in Korean cadavers. Int J Leg Med. 2017;131(5):1355–1362.
dimensional reconstruction of anatomical structures. Journal of Forensic Radiology

17