veterinary

sciences

Article
Femoral Neck Thickness Index as an Indicator of Proximal
Femur Bone Modeling
Pedro Franco-Gonçalo 1,2,3 , Ana Inês Pereira 1,4 , Cátia Loureiro 5 , Sofia Alves-Pimenta 2,3,4 , Vítor Filipe 5,6 ,
Lio Gonçalves 5,6 , Bruno Colaço 2,3,4 , Pedro Leite 7 , Fintan McEvoy 8 and Mário Ginja 1,2,3, *

1 Department of Veterinary Science, University of Trás-os-Montes and Alto Douro, 5000-801 Vila Real, Portugal;
[email protected] (P.F.-G.); [email protected] (A.I.P.)
2 Animal and Veterinary Research Centre (CECAV), University of Trás-os-Montes and Alto Douro,
5000-801 Vila Real, Portugal; [email protected] (S.A.-P.); [email protected] (B.C.)
3 Associate Laboratory for Animal and Veterinary Sciences (AL4AnimalS),
University of Trás-os-Montes and Alto Douro, 5000-801 Vila Real, Portugal
4 Department of Animal Science, University of Trás-os-Montes and Alto Douro, 5000-801 Vila Real, Portugal
5 Department of Engineering, University of Trás-os-Montes and Alto Douro, 5000-801 Vila Real, Portugal;
[email protected] (C.L.); [email protected] (V.F.); [email protected] (L.G.)
6 Institute for Systems and Computer Engineering (INESC-TEC), Technology and Science,
4200-465 Porto, Portugal
7 Neadvance Machine Vision SA, 4705-002 Braga, Portugal; [email protected]
8 Department of Veterinary Clinical Sciences, Faculty of Health and Medical Sciences,
University of Copenhagen, 1870 Copenhagen, Denmark; [email protected]
* Correspondence: [email protected]

Simple Summary: Canine hip dysplasia development results in femoral neck modeling and an
increase in thickness. The main objective of this work was to describe a femoral neck thickness
index to quantify femoral neck width and to study its association with the degree of canine hip
dysplasia using the Fédération Cynologique Internationale scoring scheme. A total of 53 dogs (106 hips)
were randomly selected for this study. Two examiners performed femoral neck thickness index
estimation to study intra- and inter-examiner reliability and agreement. Statistical analysis tests
Citation: Franco-Gonçalo, P.; Pereira, showed excellent agreement and reliability between the measurements of the two examiners and the
A.I.; Loureiro, C.; Alves-Pimenta, S.;
examiners’ sessions. All joints were scored in five categories by an experienced examiner according
Filipe, V.; Gonçalves, L.; Colaço, B.;
to the Fédération Cynologique Internationale criteria, and the results from examiner 1 were compared
Leite, P.; McEvoy, F.; Ginja, M.
between these categories. The comparison of mean femoral neck thickness index between hip
Femoral Neck Thickness Index as an
dysplasia categories using the analysis of variance test showed significant differences between groups.
Indicator of Proximal Femur Bone
Modeling. Vet. Sci. 2023, 10, 371.
These results show that femoral neck thickness index is a parameter capable of evaluating proximal
https://doi.org/10.3390/ femur bone modeling and that it has the potential to enrich conventional canine hip dysplasia scoring
vetsci10060371 criteria if incorporated into a computer-aided diagnosis software.

Academic Editor: Gerardo Fatone

Abstract: The alteration in the shape of the femoral neck is an important radiographic sign for scoring
Received: 21 April 2023 canine hip dysplasia (CHD). Previous studies have reported that the femoral neck thickness (FNT) is
Revised: 18 May 2023 greater in dogs with hip joint dysplasia, becoming progressively thicker with disease severity. The
Accepted: 21 May 2023 main objective of this work was to describe a femoral neck thickness index (FNTi) to quantify FNT
Published: 24 May 2023 and to study its association with the degree of CHD using the Fédération Cynologique Internationale
(FCI) scheme. A total of 53 dogs (106 hips) were randomly selected for this study. Two examiners
performed FNTi estimation to study intra- and inter-examiner reliability and agreement. The paired
t-test, the Bland-Altman plots, and the intraclass correlation coefficient showed excellent agreement
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
and reliability between the measurements of the two examiners and the examiners’ sessions. All
This article is an open access article joints were scored in five categories by an experienced examiner according to FCI criteria. The results
distributed under the terms and from examiner 1 were compared between FCI categories. Hips that were assigned an FCI grade of A
conditions of the Creative Commons (n = 19), B (n = 23), C (n = 24), D (n = 24), and E (n = 16) had a mean ± standard deviation FNTi of
Attribution (CC BY) license (https:// 0.809 ± 0.024, 0.835 ± 0.044, 0.868 ± 0.022, 0.903 ± 0.033, and 0.923 ± 0.068, respectively (ANOVA,
creativecommons.org/licenses/by/ p < 0.05). Therefore, these results show that FNTi is a parameter capable of evaluating proximal
4.0/).

Vet. Sci. 2023, 10, 371. https://doi.org/10.3390/vetsci10060371 https://www.mdpi.com/journal/vetsci

Vet. Sci. 2023, 10, 371 2 of 11

femur bone modeling and that it has the potential to enrich conventional CHD scoring criteria if
incorporated into a computer-aided diagnosis capable of detecting CHD.

Keywords: dog; canine hip dysplasia; femoral neck thickness index; bone modeling; osteoarthritis

1. Introduction
Canine hip dysplasia (CHD) is an inherited orthopedic disease predominant in large
and giant dog breeds that causes lameness and disability. Phenotypic expression of CHD
is influenced by genetic defects and environmental stresses that trigger hip joint laxity
and incongruency, which often leads to bone modeling and progression to secondary
osteoarthritis (OA) [1,2]. Molecular tests for the diagnosis of CHD have already been devel-
oped, but they still have not achieved acceptable diagnostic accuracy for the disease [3,4].
Radiography has remained the established imaging technology for diagnosing CHD, as
it plays an important role in the selection of breeding stock with the aim of reducing the
incidence of the disease in offspring [1,2]. In general, there is not a consistent relationship
between clinical signs and radiographic joint changes [5,6]. The ventrodorsal hip extended
(VDHE) view is recommended worldwide for CHD screening [7]. However, in young
animals, a ventrodorsal hip stress view can also be used to evaluate the hip joint laxity [1,2].
Hip laxity is considered a main risk factor for CHD, but there are some important dif-
ferences in the progression and final severity of CHD [5,6]. An image of good technical
quality of the VDHE view requires radiographic images without pelvic tilting and with
adequate femur extension and alignment [7]. However, in the last decades, despite the
widespread use of CHD screening radiographs, the prevalence of the disease remains high
in some breeds due to a number of factors. These include: variability between radiologists’
assessments, which is due to different levels of expertise among radiologists; screening
systems that are not yet sufficiently standardized and strict, allowing for some subjectivity;
the late appearance of unequivocal pathognomonic radiographic signs, at an age stage
in which sometimes the dog has already entered the breeding pool; and the absence of
evaluation of potential early hip joint changes in this view [8–10]. Therefore, the current
incidence of CHD highlights the need to introduce new parameters to improve diagnosis,
such as the hip congruency index, which has the potential to confer greater objectivity to
the assessment of hip congruency, consequently benefiting overall diagnostic and scoring
accuracy [11].
Worldwide, there are three main international entities for CHD scoring: the Fédération
Cynologique Internationale (FCI), with implementation in the countries of continental
Europe, the British Veterinary Association/The Kennel Club (BVA/KC), used mainly in
the United Kingdom, and the Orthopedic Foundation for Animals (OFA), used in the
United States of America [1,6,12]. All of these scoring systems place great emphasis
on bone modeling and OA [12]. They also take a qualitative evaluation approach to
these parameters, which leaves some margin for interpretation and error [8,13,14]. Early
screening for CHD, commonly referred to as the PennHIP method, is based on distinct hip
abnormalities. It assesses hip joint laxity by analyzing the femoral head separation from the
acetabulum under stress by measuring the distraction index (separation distance divided
by the radius of the femoral head) [1,2,6].
In CHD, early joint osteoarthritis and a consequent increase in joint fluid leads to
incongruency, laxity of the soft tissue of the hip joint, and subluxation that results in
abnormal stresses placed on the bony and soft tissue components of the joint. The joint
capsule is attached to the margin of the acetabulum and around the femoral neck. The
femoral neck is considered to be normal when its diameter narrows slightly directly below
the head. The presence of biomechanical imbalance results in cartilage wear and tear
and in the development of mechanosensitive pathways that drive proteases to initiate
the mechanism of joint breakdown, subchondral and periosteal reaction, and new bone
Vet. Sci. 2023, 10, 371 3 of 11

production in the capsule attachment area and neighboring tissues, particularly around the
junction between the head and neck [15,16]. One of the main signs of CHD is the widening
of the appearance of the femoral neck on a craniocaudal radiograph view due to osteophyte
development in conjunction with the flattening of the femoral head, the former becoming
progressively thicker until it is indistinguishable from the head in severe stages of CHD
due to bony proliferation (exostoses) [17–19]. The FCI proposes a 5-grade classification
system to represent the severity of the disease: A (normal), B (near normal/transition), C
(mild), D (moderate), and E (severe) [6,12,16,17]. Current recommendations in CHD FCI
scoring are to include Norberg angle measurement, joint space evaluation, congruence,
osteoarthritic signs, and all aspects of hip joint changes, commonly referred as Brass’
method [17]. A previous study reports that the femoral neck thickness (FNT) is altered
in hip joints classified as near normal (grade B) by the FCI system, adopting a slightly
cylindrical shape. This morphological change becomes even more evident in hip joints
classified as moderate grade (grade D) [17].
The main objective of this study was to create a new measurement method focused
on alterations of the femoral neck as a means of determining proximal femoral changes
associated with bone modeling and OA for CHD. For this purpose, we calculated the
femoral neck thickness index (FNTi), an objective parameter that relates the minimal
FNT to the diameter of the ipsilateral femoral head, and compared it to the different
FCI grades. Our hypothesis was that there would be an association between FCI grades
and FNTi, and that FNTi would increase with disease severity. To our knowledge, the
FNTi’s association with the FCI grades has not been previously studied and could be
integrated into a classification system as a parameter for evaluating bone modeling of the
proximal femur.

2. Materials and Methods

This was a retrospective study based on the evaluation of VDHE views that were
randomly selected from the Veterinary Teaching Hospital of the University of Trás-os-
Montes and Alto Douro database and from the Danish Kennel Club database, obtained
between 2010 and 2023. Recorded data included breed, sex, and weight. The inclusion
criteria were dogs older than 12, 15, or 18 months (according to FCI recommendations for
medium, large, and giant breeds) and VDHE views with adequate technical quality in
terms of image (good bone contrast and spatial resolution) and pelvis positioning (femurs
parallel to each other, patellae centred between the femoral condyles, and pelvic symmetry)
for CHD scoring. VDHE views showing radiographic signs compatible with other hip
or hindlimb diseases, such as bone fractures, previous surgeries, neoplasia, and knee
osteoarthritis, were excluded. All five FCI categories were similarly represented in the
sample. Due to the observational nature of this study, owner consent and ethical committee
approval were waived.
The sample included 53 dogs of 13 different breeds: Portuguese Pointers (13/53,
25%), Estrela Mountain dogs (12/53, 23%), Transmontano Mastiffs (8/53, 15%), German
Shepherd dogs (5/53, 9%), Labrador Retrievers (3/53, 9%), Bernese Mountain dogs (2/53,
4%), Barbado da Terceira (1/53, 2%), French Bulldog (1/53, 2%), German Shorthaired
Pointer (1/53, 2%), Old Danish Pointer (1/53, 2%), Portuguese Sheepdog (1/53, 2%),
Portuguese Water dog (1/53, 2%), and Rottweiler (1/53, 2%). There were 19 (36%) males
and 34 (64%) females. The age ranged from 12 to 166 months, and the mean ± standard
deviation (SD) was 31.87 ± 29.07 months. The average body weight was 36.1 ± 14.1 kg.

2.1. Radiographic Measurements

The minimal FNT was measured by drawing a straight line, roughly perpendicular to
the anatomical axis of the femoral neck, connecting the two closest points between the prox-
imal and distal margins of the femoral neck in a VDHE view. The femoral head diameter
was determined as a diameter of a circle outlining the margin of the femoral head (Figure 1).
These measurements were performed by examiner 1 (E1) in two independent sessions
 scored using FCI criteria: grade A (normal hip—Norberg angle (NA) > 105° and excellent
congruency); grade B (borderline or transitional hip joint—NA around 105° and mild in-
congruency); grade C (slight CHD—NA around 100°, centre of femoral head outside of
dorsal acetabular margin, and moderate incongruency); grade D (moderate CHD—NA >
90°, signs of osteoarthritis, and obvious incongruency); and grade E (severe CHD—NA <
90°, signs of osteoarthritis, and severe incongruency) [12,17]. The NA was measured be-
Vet. Sci. 2023, 10, 371 tween a line joining the centre of the circle encompassing the femoral heads and another 4 of 11
line connecting each centre of the femoral head with the ipsilateral, effective cranial ace-
tabular rim [20]. Some radiographic parameters related to the femoral head centre posi-
tion and the dorsal acetabular margin and joint space were also considered for final FCI
(S1
jointand S2)[17].
scoring to evaluate repeatability and by examiner 2 (E2) to test the reproducibility
using The E1 and E2
specific measurements
DICOM viewerwere
andperformed by I.P. and(Dys4Vet
editor software P.F., respectively,
version and theaccessed between
2.0,
hip FCI scoring was performed by M.G. in a single-blind fashion for each examiner (i.e.,
1 October and 31 December 2022). The FNTi was determined by dividing the FNT by the
E1 and E2 were unaware of each FNTi measurement and FCI scores, and the FCI scorer
femoral
was unawarehead diameter.
of FNTi measurements).

Figure 1. Detail of the left hip joint of a dog showing a circle delimiting the head of the femur with
Figure 1. Detail of the left hip joint of a dog showing a circle delimiting the head of the femur with
31.5 mm diameter (FHd; dashed line) and a line joining the proximal and distal borders of the fem-
31.5 mmtodiameter
oral neck determine(FHd; dashed
its thickness line)
of 28.7 mmand a line
(FNT), joining thetoproximal
corresponding and distal
an index (FNTi) of 0.91.borders of the femoral
neck to determine its thickness of 28.7 mm (FNT), corresponding to an index (FNTi) of 0.91.

In order to associate FNTi with the five FCI grades for CHD, the hip joints were
scored using FCI criteria: grade A (normal hip—Norberg angle (NA) > 105◦ and excellent
congruency); grade B (borderline or transitional hip joint—NA around 105◦ and mild incon-
gruency); grade C (slight CHD—NA around 100◦ , centre of femoral head outside of dorsal
acetabular margin, and moderate incongruency); grade D (moderate CHD—NA > 90◦ ,
signs of osteoarthritis, and obvious incongruency); and grade E (severe CHD—NA < 90◦ ,
signs of osteoarthritis, and severe incongruency) [12,17]. The NA was measured between
a line joining the centre of the circle encompassing the femoral heads and another line
connecting each centre of the femoral head with the ipsilateral, effective cranial acetabu-
lar rim [20]. Some radiographic parameters related to the femoral head centre position
and the dorsal acetabular margin and joint space were also considered for final FCI joint
scoring [17].
The E1 and E2 measurements were performed by I.P. and P.F., respectively, and the
hip FCI scoring was performed by M.G. in a single-blind fashion for each examiner (i.e., E1
and E2 were unaware of each FNTi measurement and FCI scores, and the FCI scorer was
unaware of FNTi measurements).

2.2. Statistical Analysis

SPSS computer software (SPSS Statistics for Windows Version 27.0: IBM Corp., Ar-
monk, NY, USA) was used to perform the statistical analysis.
Vet. Sci. 2023, 10, 371 5 of 11

Parametric tests were used for statistical analysis [21]. The paired t-test was used for
comparison of duplicate E1S1–E1S2 measurements to evaluate repeatability, as well as
E1S1–E2 measurements to evaluate reproducibility [22]. The Bland–Altman analysis and
the intraclass correlation coefficient (ICC) were used to investigate intra- and inter-examiner
agreement and reliability, respectively. In the Bland–Altman method, the 95% limits of
agreement (LA) were calculated as the mean difference (d) ± 1.96 standard deviation
(SD) [22–25]. Measurements were considered in agreement when the 95% confidence
interval (CI) of the mean differences included zero and equivalent when the 95% upper
and lower LA were small (irrelevant difference) [23,24]. The ICC was considered as poor,
moderate, good/acceptable, and excellent reliability when the lower limit of 95% CI
was <0.50, ≥0.50–0.75, ≥0.75–90, and ≥0.90, respectively [25,26]. Cohen’s d was used to
measure the effect size when significant differences between measurements were registered:
negligible < 0.20, small ≥ 0.20, medium ≥ 0.50, and large ≥ 0.80 [27].
The comparison of FNTi values of E1S1 measurement between FCI categories was
performed using the Welch’s ANOVA, followed by the post hoc Games–Howell test. The
null hypothesis was that there were no significant differences in the FNTi mean values
between the FCI categories [28]. A p-value of <0.05 was considered statistically significant.
The statistical analysis was performed considering each joint individually.

3. Results
Measurements were performed on 106 hip joints. The FNTi mean ± SD in E1S1
was 0.86 ± 0.06; in E1S2 it was 0.86 ± 0.06; and in E2 it was 0.87 ± 0.06. The main
statistical analysis results related to the intra-examiner (repeatability) (p > 0.05 in paired
t-test; ICC = 0.94 [95% CI, 0.92–0.96]; d ± SD −0.001 ± 0.019 [95% LA, −0.038, 0.036]) and
inter-examiner (reproducibility) (p < 0.05 in paired t-test; ICC = 0.93 [95% CI, 0.90–0.95];
d ± SD −0.007 ± 0.021 [95% LA, −0.08,4 0.034]) measurements of FNTi are presented in
detail in Table 1, Figures 2 and 3.

Table 1. Statistical analysis of the intra- and inter-examiner agreement and reliability of the FNTi in
measuring 106 hips.

95% LA
Paired t-Test
Comparison Cohen’s d ICC [95% CI] d ± SD Lower Upper
p-Value
Bound Bound
Intra-
E1S1–E1S2 0.583 – 0.94 [0.92,0.96] −0.001 ± 0.019 −0.038 0.036
examiner
Inter-
E1S1–E2 <0.001 0.14 0.93 [0.90,0.95] −0.007 ± 0.021 −0.048 0.034
examiner
Abbreviations: CI, confidence interval; d, mean difference; E, examiner; FNTi, femoral neck thickness index; ICC,
intraclass correlation coefficient; LA, limits of agreement; S, session; SD, standard deviation.

A total of 19 (18%) hip joints were scored as FCI grade A, and the FNTi mean ± SD
was 0.809 ± 0.024; 23 (21%) hip joints were scored as FCI grade B, and the FNTi mean ± SD
was 0.835 ± 0.044; 24 (23%) hip joints were scored as FCI grade C, and the FNTi mean ± SD
was 0.868 ± 0.022; 24 (23%) hip joints were scored as FCI grade D, and the FNTi mean ± SD
was 0.903 ± 0.033; and 16 (15%) hip joints were scored as FCI grade E, and the FNTi
mean ± SD was 0.923 ± 0.068. Data were assessed for normality using the Shapiro–Wilk
test (pA = 0.501; pB = 0.275; pC = 0.739; pD = 0.983; pE = 0.383), and Levene’s test indicated
unequal variances among group samples (pA,B,C,D,E < 0.05). Significant statistical mean
differences using Welch’s ANOVA followed by the post hoc Games–Howell test were
verified in FCI categories with means marked with different letter superscripts (p < 0.05)
(Table 2, Figure 4).
Vet.Sci.
Vet. Sci.2023, 10,x371
2023,10, FOR PEER REVIEW 6 6ofof11
11

Examiner11differences
Figure2.2. Examiner
Figure differencesbetween
betweenSessions
Sessions11andand22(E1S1;
(E1S1;E1S2).
E1S2).The
Thegreen
greenline
linerepresents
represents
the
themean
meanof ofthe
the differences
differences (−0.001)
(−0.001)and
andthe
thered
redlines
linesrepresent
representthe
thelower
lowerand
andupper
upper95%
95%limits
limitsof
of
agreement, −0.038 and 0.036, respectively. FNTi–femoral neck thickness
agreement, −0.038 and 0.036, respectively. FNTi–femoral neck thickness index. index.

Figure
Figure3.3.Examiner
Examiner1,1,Session
Session11(E1S1)
(E1S1)and
andExaminer
Examiner22(E2)
(E2)measurement
measurementdifferences.
differences.The
Thegreen
greenline
line
represents
represents the mean of the differences (−0.007) and the red lines represent the lower and upper95%
the mean of the differences (−0.007) and the red lines represent the lower and upper 95%
limits of agreement, −0.048 and 0.034, respectively. FNTi–Femoral neck thickness index.
limits of agreement, −0.048 and 0.034, respectively. FNTi–Femoral neck thickness index.

A total of 19 (18%) hip joints were scored as FCI grade A, and the FNTi mean ± SD
was 0.809 ± 0.024; 23 (21%) hip joints were scored as FCI grade B, and the FNTi mean ± SD
was 0.835 ± 0.044; 24 (23%) hip joints were scored as FCI grade C, and the FNTi mean ± SD
was 0.868 ± 0.022; 24 (23%) hip joints were scored as FCI grade D, and the FNTi mean ±
SD was 0.903 ± 0.033; and 16 (15%) hip joints were scored as FCI grade E, and the FNTi
 mean ± SD was 0.923 ± 0.068. Data were assessed for normality using the Shapiro–Wilk
test (pA = 0.501; pB = 0.275; pC = 0.739; pD = 0.983; pE = 0.383), and Levene’s test indicated
unequal variances among group samples (pA,B,C,D,E < 0.05). Significant statistical mean dif-
Vet. Sci. 2023, 10, 371 ferences using Welch’s ANOVA followed by the post hoc Games–Howell test were veri- 7 of 11

fied in FCI categories with means marked with different letter superscripts (p < 0.05) (Table
2, Figure 4).
Table 2. Statistical descriptive analysis of the femoral neck thickness index by FCI categories.
Table 2. Statistical descriptive analysis of the femoral neck thickness index by FCI categories.
Mean 95% CI
FCI Mean 95% CI
FCI N Mean * SD Lower Upper Min Max
Categories N Mean * SD Upper Min Max
Categories Bound
Lower Bound Bound
Bound
A 19 0.809 a 0.024 0.797 0.823 0.771 0.864
A 19 0.809 a 0.024 0.797 0.823 0.771 0.864
B 23 0.835 a,b a,b 0.044 0.812 0.859 0.757 0.952
B 23 0.835 0.044 0.812 0.859 0.757 0.952
CC 24 0.868 b,c 0.022 0.856 0.880 0.831 0.918
24 0.868 b,c 0.022 0.856 0.880 0.831 0.918
DD 24 c,d c,d 0.033 0.886 0.921 0.841 0.969
24 0.903
0.903 0.033 0.886 0.921 0.841 0.969
EE 16 16 0.923
0.923 d d 0.068
0.068 0.887
0.887 0.959
0.959 0.781
0.781 1.036
1.036
** Means withdifferent
Means with differentsuperscripts
superscriptsareare statistically
statistically different
different (p < 0.05)
(p < 0.05) in thein thehoc
post post hoc Games–Howell
Games–Howell test that
test that Welch’s
followed followed Welch’s
ANOVA. ANOVA. CI,
Abbreviations: Abbreviations: CI, FCI,
confidence interval; confidence
Fédérationinterval; FCI,Internationale;
Cynologique Fédération
SD, standard deviation.
Cynologique Internationale; SD, standard deviation.

Figure
Figure 4.
4. Box
Boxand
andwhisker
whiskerplot
plotpresenting
presentingthethe femoral
femoral neck
neck thickness
thickness index
index (FNTi)
(FNTi) classified
classified in
in
Fédération
Fédération Cynologique Internationale (FCI) categories (A–E). The box plots show frombottom–up
Cynologique Internationale (FCI) categories (A–E). The box plots show from bottom–up
minimum
minimum datadata value,
value,lower
lowerquartile
quartilevalue,
value, median
median value,
value, upper
upper quartile
quartile value,
value, maximum
maximum data
data value,
value, and outliers (distant from the rest of the data).
and outliers (distant from the rest of the data).

4.
4.Discussion
Discussion
The
The objective
objective of
of this
this study
study was
was toto create
create aa methodology
methodology that that would
would allow
allow aa more
more
objective
objective radiographic
radiographicevaluation
evaluationof ofthe
thechanges
changesthat
that the
the proximal
proximal femur
femur undergoes
undergoes in in
CHD
CHD(namely,
(namely,thickening
thickeningofofthe neck),
the andand
neck), to understand
to understand this this
relationship across
relationship the var-
across the
various
ious FCIFCI degrees
degrees of CHD
of CHD in in
thethe VDHE
VDHE view.
view. TheTheuse
useofofa aparametric
parametrictest
testin
in statistical
statistical
analysiswas
analysis wasbased
basedononthetheCentral
CentralLimit
LimitTheorem,
Theorem,which
whichstates
statesthat
thatin
inlarge
large sample
sample sizes
sizes
(n >>30),
(n 30),the
thedistribution
distributionof ofstandardized
standardizedsamples’
samples’means
meanstends
tendsto tobe
benormally
normallydistributed
distributed
independently of the distribution of the population from where it originated [21]. The FCI
scoring can be considered a dynamic system that has undergone regular updates over the
years in order to harmonize CHD classifications in different countries and to improve the
reliability of associating morphological alterations with the genetic profile of the animal.
We highlight here the CHD panelist meetings of Dortmund 1991 and Copenhagen 2007
and 2022 [29].
Vet. Sci. 2023, 10, 371 8 of 11

In dysplastic hips, progressive bone modeling and osteophyte formation are induced
by osteoarthritic pathways that result in subchondral and periosteal response with new
bone production. Particularly around the junction between the head and neck, biomechani-
cal stresses in the hip joint occur at a faster rate, and new bone and osteophytes are placed in
some areas and reabsorbed in other areas of the femur and acetabulum [15]. These changes
in the bone structure are concurrent with hip osteoarthritis [30]. Pinna et al. (2022) observed
a significant increase in the thickness of the femoral neck in VDHE views of hips classified
as grade B, a grade assigned to hips considered healthy and in which no osteoarthritic signs
are visible [17]. In another study, Andronescu et al. (2015) studied the ratio between head
volume and femoral neck volume on 3D computed tomography images of dogs at high
risk of developing CHD (distraction index > 0.3) from 16 to 32 weeks of age and found a
decrease in the ratio, even though the differences between values corresponding to OA
severities were not statistically significant [18]. Therefore, the study of bone modeling
and OA as two interrelated topics seems particularly important to us. We use the FNTi
to get around the difficulty that exists in veterinary medicine to use absolute measures
in anatomic measurements due to the different sizes of dogs. The relationship between
measurements and the size of the femoral head is a strategy that has been successfully used
previously for other purposes, such as the hip distraction index (mainly in young animals
without severe bone changes) [1,6] and the hip congruency index [11]. In a previous study,
the femoral neck width was related with the femoral neck length to create a widening
index [31]. However, this methodology was not followed in this work because we think
that the relationship between the length of the femoral neck and the size of the breed is
much less studied and used than the diameter of the femoral head [6,11]. As such, there
may be greater variability in the length of the femoral neck between breeds of similar sizes,
which makes this parameter a less effective ratio measure. On the other hand, artificial
intelligence is currently being introduced in digital image analysis [32,33], and since the
identification of the femoral head seems to us to be an essential parameter in the application
of artificial intelligence to the diagnosis of CHD, any index that resorts to its use can be
more easily integrated in the near future for this purpose.
Bone modeling is admittedly one of the aspects of classification that is undervalued
by the FCI criteria. In some iterations of the FCI criteria [34], morphological changes of
the proximal femur are only explicitly mentioned in the most severe degree of dysplasia
(grade D), pointing out the characteristic mushroom-like appearance that the femoral
head assumes. Other changes related to bone modeling of the proximal femur have been
overlooked and are presumably relegated to the topic of “osteoarthritic signs”, addressed
with a yes/no question [34,35], which makes the evaluation less accurate. The BVA/KC
uses a phenotypic evaluation criteria based on a points system, addressing changes in
bone shape from minor modeling up to severe OA [19]. However, it is still a qualitative
scoring system, so there is also some subjectivity in the analysis because it ends up being
dependent on the level of experience of the examiner. Taking this into consideration, there
is a perceived need for novel approaches that grant more objectivity to the assessment of
bone modeling and, in addition, CHD scoring.
Our results showed that the methodology behind the FNTi had excellent reliability
and agreement. Given the results of the paired t-tests and Bland-Altman plots with the
mean differences near zero and the narrow 95% CI, there was no evidence of bias between
the examiners’ measurements, and they can be considered statistically similar. The intra-
examiner ICC was 0.94, and the lower bound of the 95% CI was 0.92, which translates
to excellent reliability. This indicates adequate repeatability and reproducibility of the
described FNTi determination methodology. Significant mean differences between FNTi
examiners’ measurements were observed in the paired samples t-test. However, the
recorded effect size was negligible (d = 0.14) and, considering the corresponding Bland–
Altman plot (Figure 3), the 95% LA can be interpreted as clinically small because 95% of all
calculated inter-examiner differences lie in the short range of −0.048 to 0.034. Furthermore,
the inter-examiner ICC shows excellent reliability. This shows that even though the two
Vet. Sci. 2023, 10, 371 9 of 11

examiners themselves produce somewhat different values, their measurements are clearly
related and functionally consistent. Both examiners had a similar amount of practice time
with the method beforehand, so no conclusions can be drawn based on experience. On the
other hand, one would expect the mean differences to be greater among bigger FNTi mean
values due to the subjectiveness of the annotation imposed by exostosis and/or osteophyte
formation in the concave fossa, making the bone contours more intricate, which could
have possibly caused disparities between sessions and examiners. Ultimately, by analyzing
the two diagrams (Figures 2 and 3), we can see that the mean difference values do not
increase or decrease in proportion to the mean FNTi values, thus concluding that there is
no proportional bias.
The Welch’s ANOVA of FNTi values in FCI categories revealed a statistically significant
main effect (p < 0.001), indicating that not all FCI categories had the same mean FNTi value.
Post hoc comparisons using the Games–Howell post hoc procedure were conducted to
determine which pairs of the five categories’ means differed significantly. The results
indicate that different FCI categories have mean FNTi values that increase gradually with
the degree of severity of the disease: grade A hips had a statistically significant lower
mean FNTi value than grades C, D, and E (p < 0.05); grade B had significantly lower
mean FNTi value than grades D and E (p < 0.05); and grade C had significantly lower
mean FNTi value than grade E (p < 0.05). Therefore, the null hypothesis is rejected, which
supports our assumption that FNTi changes with FCI CHD grades. This parameter can help
distinguish between CHD grades, potentially improving FCI scoring criteria. However,
there is some overlap in the FNTi values corresponding to the different FCI categories,
specifically between adjacent categories. This is one of the reasons that an assessment
solely dependent on this is ambiguous and, therefore, impractical. The “whiskers” of the
box plot corresponding to category E stretch over a wider range of values than the other
box plots, overlapping almost every other category. This can be explained by how long
hip subluxation has been in place and by the difficulty of adequate delimitation of the
femoral head for its diameter measurement in joints with very severe bone osteoarthritic
deformations. In severe cases with long-term subluxation, the surface of the femoral
head cartilage at the non-articulated margin of the head and neck becomes thickened
due to a lack of contact with the opposite acetabular surface [15]. Hence, a long-term
subluxated hip favors the appearance of the so-called mushroom head deformity, which
inflates the diameter of the femoral head compared to the neck thickness, thus producing
lower FNTi values pertaining to category E, values that lie in the lower “whisker” of the
box plot. We strongly advocate that this parameter should never be used by itself to classify
hips, but rather in complementarity with other parameters. On the other hand, given the
recognized variability that exists in the progression of CHD, femoral neck bone modeling,
and femoral head size, it is expected that more robust FNTi results will be obtained if a
study is performed using only one breed in the sample [5].
Since our study sample was sourced from two different databases and included a
variety of breeds, the results can be more easily extrapolated to general canine populations
at risk of CHD. Nonetheless, as a limitation of this study, it is important to note that some
breeds prone to CHD are clearly over-represented (Portuguese breeds); others, on the
contrary, are under-represented (Labrador Retriever and Rottweiler). As such, in the future,
there is a need to conduct more studies on a wider range of breeds and patients.

5. Conclusions
This study describes a methodology that allows for the evaluation of bone modeling
of the proximal femur in the VDHE view, which can be used in the future with confidence
as a criterion for CHD scoring. The FNTi shows adequate intra- and inter-examiner
measurement agreement and reliability. Mean FNTi values are gradually higher in the
different FCI categories, with statistically significant differences. The FNTi method shows
potential to make CHD classification more objective if incorporated as a scoring criterion,
Vet. Sci. 2023, 10, 371 10 of 11

and it is also promising in view of its application through a computer-aided diagnosis

capable of autonomously identifying and classifying CHD.

Author Contributions: M.G. and P.F.-G. contributed to conception and design of this study. B.C.,
F.M., P.F.-G., P.L., S.A.-P. and M.G. organized the database. M.G., P.F.-G., A.I.P. and C.L. defined the
methodology. B.C., C.L., F.M., L.G., A.I.P., V.F. and M.G. performed validation and data analysis.
P.F.-G. wrote the first draft of the manuscript. All authors have read and agreed to the published
version of the manuscript.
Funding: This work was financed by project Dys4Vet (POCI-01-0247-FEDER-046914), co-financed
by the European Regional Development Fund (ERDF) through COMPETE2020, the Operational
Programme for Competitiveness and Internationalisation (OPCI).
Institutional Review Board Statement: Ethical review and approval were waived for this study due
to the observational and retrospective nature of this study.
Informed Consent Statement: This study did not involve humans.
Data Availability Statement: Not applicable.
Acknowledgments: The authors acknowledge the UTAD Veterinary Teaching Hospital and Danish
Kennel Club for allowing access to images from their data archive. The authors are also grateful for
all the conditions made available by FCT—Portuguese Foundation for Science and Technology, under
the projects UIDP/00772/2020, LA/P/0059/2020 and Scientific Employment Stimulus—Institutional
Call—CEECINST/00127/2018 UTAD.
Conflicts of Interest: Author P.L. was employed by Neadvance Machine Vision SA. The remaining
authors declare that the research was conducted in the absence of any commercial or financial
relationships that could be construed as a potential conflict of interest. Bartolome N., Segarra
S., Artieda M. et al. (2015) A genetic predictive model for canine hip dysplasia: integration of
genomewide association study (GWAS) and candidate gene approaches. PLoS ONE 10, e0122558.

References
1. Smith, G.K.; Paster, E.R.; Powers, M.Y.; Lawler, D.F.; Biery, D.N.; Shofer, F.S.; McKelvie, P.J.; Kealy, R.D. Lifelong Diet Restriction
and Radiographic Evidence of Osteoarthritis of the Hip Joint in Dogs. J. Am. Vet. Med. Assoc. 2006, 229, 690–693. [CrossRef]
[PubMed]
2. Ginja, M.M.; Gonzalo-Orden, J.M.; Melo-Pinto, P.; Bulas-Cruz, J.; Orden, M.A.; San Roman, F.; Llorens-Pena, M.P.; Ferreira, A.J.
Early hip laxity examination in predicting moderate and severe hip dysplasia in Estrela mountain dog. J. Small Anim. Pract. 2008,
49, 641–646. [CrossRef] [PubMed]
3. Bartolomé, N.; Segarra, S.; Artieda, M.; Francino, O.; Sánchez, E.; Szczypiorska, M.; Casellas, J.; Tejedor, D.; Cerdeira, J.; Martínez,
A.; et al. A genetic predictive model for canine hip dysplasia: Integration of Genome Wide Association Study (GWAS) and
candidate gene approaches. PLoS ONE 2015, 10, e0122558. [CrossRef] [PubMed]
4. Bruun, C.S.; Bank, A.; Ström, A.; Proschowsky, H.F.; Fredholm, M. Validation of DNA test for hip dysplasia failed in Danish
Labrador Retrievers. Anim. Genet. 2020, 51, 617–619. [CrossRef]
5. Lopez, M.; Schachner, E. Diagnosis, Prevention, and Management of Canine Hip Dysplasia: A Review. Vet. Med. Res. Rep. 2015, 6,
181–192. [CrossRef]
6. Ginja, M.M.D.; Silvestre, A.M.; Gonzalo-Orden, J.M.; Ferreira, A.J.A. Diagnosis, Genetic Control and Preventive Management of
Canine Hip Dysplasia: A Review. Vet. J. 2010, 184, 269–276. [CrossRef]
7. Martins, J.; Colaço, B.; Alves-Pimenta, S.; Gonzalo Orden, J.M.; Ferreira, A.J.; Ginja, M.M. Effect of the Dog positioning on X-ray
Table on Hip Dysplasia Parameter Evaluation. Vet. Comp. Orthop. Traumatol. 2019, 32, 376–382. [CrossRef]
8. Coopman, F.; Verhoeven, G.; Paepe, D.; Van Bree, H.; Duchateau, L.; Saunders, J.H. Inter-Observer Agreement on the Radiographic
Assessment of Canine Hip Dysplasia. Vlaams Diergeneeskd. Tijdschr. 2007, 76, 417–422.
9. Genevois, J.-P.; Cachon, T.; Fau, D.; Carozzo, C.; Viguier, E.; Collard, F.; Remy, D. Canine Hip Dysplasia Radiographic Screening.
Prevalence of Rotation of the Pelvis along Its Length Axis in 7012 Conventional Hip Extended Radiographs. Vet. Comp. Orthop.
Traumatol. 2007, 20, 296–298.
10. Powers, M.Y.; Biery, D.N.; Lawler, D.F.; Evans, R.H.; Shofer, F.S.; Mayhew, P.; Gregor, T.P.; Kealy, R.D.; Smith, G.K. Use of the
Caudolateral Curvilinear Osteophyte as an Early Marker for Future Development of Osteoarthritis Associated with Hip Dysplasia
in Dogs. J. Am. Vet. Med. Assoc. 2004, 225, 233–237. [CrossRef]
11. Franco-Gonçalo, P.; Moreira da Silva, D.; Leite, P.; Alves-Pimenta, S.; Colaço, B.; Ferreira, M.; Gonçalves, L.; Filipe, V.; McEvoy, F.;
Ginja, M. Acetabular Coverage Area Occupied by the Femoral Head as an Indicator of Hip Congruency. Animals 2022, 12, 2201.
[CrossRef]
Vet. Sci. 2023, 10, 371 11 of 11

12. Flückiger, M. Scoring Radiographs for Canine Hip Dysplasia—The Big Three Organizations in the World. Eur. J. Companion Anim.
Pract. 2007, 17, 135–140.
13. Geissbühler, U.; Drazovic, S.; Lang, J.; Howard, J. Inter-Rater Agreement in Radiographic Canine Hip Dysplasia Evaluation. Vet.
Rec. 2017, 180, 357. [CrossRef]
14. Fortrie, R.R.; Verhoeven, G.; Broeckx, B.; Duchateau, L.; Janssens, L.; Samoy, Y.; Schreurs, E.; Saunders, J.; van Bree, H.;
Vandekerckhove, P.; et al. Intra- and Interobserver Agreement on Radiographic Phenotype in the Diagnosis of Canine Hip
Dysplasia. Vet. Surg. 2015, 44, 467–473. [CrossRef]
15. Riser, W.H.; Rhodes, W.H.; Newton, C.D. Hip Dysplasia. In Textbook of Small Animal Orthopedics; U. Penn CAL Small Animal
Orthopedics; Lippincott Williams & Wilkins: Philadelphia, PA, USA, 1982.
16. Smith, G.K. New Paradigms for Hip Dysplasia Prevention and Control, Performance and Ethics of CHD Screening. In 12th
ESVOT Congress Proceedings; Vezzoni, A., Schramme, M., Eds.; LMU: Munich, Germany, 2004; pp. 125–131.
17. Pinna, S.; Tassani, C.; Antonino, A.; Vezzoni, A. Prevalence of Primary Radiographic Signs of Hip Dysplasia in Dogs. Animals
2022, 12, 2788. [CrossRef]
18. Andronescu, A.A.; Kelly, L.; Kearney, M.T.; Lopez, M.J. Associations between Early Radiographic and Computed Tomographic
Measures and Canine Hip Joint Osteoarthritis at Maturity. Am. J. Vet. Res. 2015, 76, 19–27. [CrossRef]
19. Dennis, R. Interpretation and Use of BVA/KC Hip Scores in Dogs. InPractice 2012, 34, 178–194. [CrossRef]
20. Gaspar, A.R.; Hayes, G.; Ginja, C.; Ginja, M.M.; Todhunter, R.J. The Norberg Angle Is Not an Accurate Predictor of Canine Hip
Conformation Based on the Distraction Index and the Dorsolateral Subluxation Score. Prev. Vet. Med. 2016, 135, 47–52. [CrossRef]
21. Kwak, S.G.; Kim, J.H. Cornerstone of Modern Statistics. Korean J. Anesthesiol. 2017, 70, 144–156. [CrossRef]
22. Watson, P.F.; Petrie, A. Method Agreement Analysis: A Review of Correct Methodology. Theriogenology 2010, 73, 1167–1179.
[CrossRef]
23. Ludbrook, J. Confidence in Altman-Bland Plots: A Critical Review of the Method of Differences. Clin. Exp. Pharmacol. Physiol.
2010, 37, 143–149. [CrossRef] [PubMed]
24. Bland, J.M.; Altman, D.G. Measuring Agreement in Method Comparison Studies. Stat. Methods Med. Res. 1999, 8, 135–160.
[CrossRef] [PubMed]
25. Koo, T.K.; Li, M.Y. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J. Chiropr.
Med. 2016, 15, 155–163. [CrossRef] [PubMed]
26. Lee, J.; Koh, D.; Ong, C.N. Statistical Evaluation of Agreement between Two Methods for Measuring a Quantitative Variable.
Comput. Biol. Med. 1989, 19, 61–70. [CrossRef]
27. Cohen, J. Statistical Power Analysis for the Behavioral Sciences, 2nd ed.; Lawrence Erlbaum Associates, Inc.: London, UK, 1988.
28. Marôco, J. Análise Estatística Com o SPSS Statistics, 7th ed.; Gráfica Manuel Barbosa & Filhos: Pêro Pinheiro, Portugal, 2018.
29. Hedhammar, Å. Swedish Experiences From 60 Years of Screening and Breeding Programs for Hip Dysplasia-Research, Success,
and Challenges. Front. Vet. Sci. 2020, 27, 228. [CrossRef]
30. Edinger, D.T.; Hayashi, K.; Hongyu, Y.; Markel, M.D.; Manley, P.A. Histomorphometric Analysis of the Proximal Portion of the
Femur in Dogs with Moderate Osteoarthritis. Am. J. Vet. Res. 2005, 66, 150–155. [CrossRef]
31. Martins, J.; Ferreira, A.J.; Ginja, M.M. Morphometric assessment of the hip joint in the Estrela Mountain Dog breed. Vet. Comp.
Orthop. Traumatol. 2012, 25, 202–210. [CrossRef]
32. McEvoy, F.J.; Proschowsky, H.F.; Müller, A.V.; Moorman, L.; Bender-Koch, J.; Svalastoga, E.L.; Frellsen, J.; Nielsen, D.H. Deep
Transfer Learning Can Be Used for the Detection of Hip Joints in Pelvis Radiographs and the Classification of Their Hip Dysplasia
Status. Vet. Radiol. Ultrasound. 2021, 62, 387–393. [CrossRef]
33. Banzato, T.; Wodzinski, M.; Tauceri, F.; Donà, C.; Scavazza, F.; Müller, H.; Zotti, A. An AI-Based Algorithm for the Auto-matic
Classification of Thoracic Radiographs in Cats. Front. Vet. Sci. 2021, 8, 731936. [CrossRef]
34. Morgan, J.K.; Stephens, M. Radiographic Diagnosis and Control of Canine Hip Dysplasia; Iowa State University Digital Press: Ames,
IA, USA, 1985.
35. FCI Hip Dyspalsia Classification. In Proceedings of the FCI/DKK Hip Dysplasia Conference, Copenhagen, Denmark, 24–25 May
2022; p. 11.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.