Microsoft Word - Main document copia.docx Zurich Open Repository and Archive University of Zurich Main Library Strickhofstrasse 39 CH-8057 Zurich www.zora.uzh.ch Year: 2019 Accuracy of an automated three-dimensional technique for the computation of femoral angles in dogs Longo, Federico ; Savio, Gianpaolo ; Contiero, Barbara ; Meneghello, Roberto ; Concheri, Gianmaria ; Franchini, Federico ; Isola, Maurizio Abstract: Aims: The purpose of the study was to evaluate the accuracy of a three-dimensional (3D) automated technique (computer-aided design (aCAD)) for the measurement of three canine femoral angles: anatomical lateral distal femoral angle (aLDFA), femoral neck angle (FNA) and femoral tor- sion angle.Methods:Twenty-eight femurs equally divided intotwo groups (normal and abnormal) were obtained from 14 dogs of different conformations (dolicomorphic and chondrodystrophicCT scans and 3D scanner acquisitions were used to create stereolithographic (STL) files , which were run in a CAD platform. Two blinded observers separately performed the measurements using the STL obtained from CT scans (CT aCAD) and 3D scanner (3D aCAD), which was considered the gold standard method. C orrelation coefficients were used to investigate the strength of the relationship between the two mea- surements.Results: A ccuracy of the aCAD computation was good, being always above the threshold of R2 of greater than 80 per cent for all three angles assessed in both groups. a LDFA and FNA were the most accurate angles (accuracy gt;90 per cent).Conclusions: The proposed 3D aCAD protocol can be considered a reliable technique to assess femoral angle measurements in canine femur. The developed algorithm automatically calculates the femoral angles in 3D, thus considering the subjective intrinsic femur morphology. The main benefit relies on a fast user-independent computation, which avoids user-related measurement variability. The accuracy of 3D details may be helpful for patellar luxation and femoral bone deformity correction, as well as for the design of patient- specific, custom-made hip prosthesis implants. DOI: https://doi.org/10.1136/vr.105326 Posted at the Zurich Open Repository and Archive, University of Zurich ZORA URL: https://doi.org/10.5167/uzh-172071 Journal Article Accepted Version Originally published at: Longo, Federico; Savio, Gianpaolo; Contiero, Barbara; Meneghello, Roberto; Concheri, Gianmaria; Fran- chini, Federico; Isola, Maurizio (2019). Accuracy of an automated three-dimensional technique for the computation of femoral angles in dogs. Veterinary Record, 185(14):443. DOI: https://doi.org/10.1136/vr.105326 Research article 1 2 3 Accuracy of an automated three-dimensional technique for the 4 computation of femoral angles in dogs 5 6 7 F. Longo a, d *, G. Savio b, B. Contiero a, R. Meneghello c, G. Concheri b, F. Franchinib, M. Isola a 8 9 a Department of Animal Medicine, Production and Health, University of Veterinary Medicine, 10 Padova, Italy 11 b Laboratory of Design Tools and Methods in Industrial Engineering, Department of Civil, 12 Architectural and Environmental Engineering, University of Engineering, Padova, Italy 13 c Department of Management and Engineering, University of Padova, Vicenza, Italy 14 d Clinic for Small Animal Surgery, Vetsuisse Faculty University of Zurich, Zurich, Switzerland 15 16 17 18 19 20 21 * Corresponding author. Tel: +39 049 8272608. 22 E-mail address: flongo@vetclinics.uzh.ch (F. Longo). 23 24 25 26 27 28 29 30 31 32 33 34 Abstract 35 The purpose of the study was the evaluation of the accuracy of a three-dimensional (3D) automated 36 technique (aCAD) for the measurement of three canine femoral angles: anatomical lateral distal 37 femoral angle (aLDFA); femoral neck angle (FNA); and femoral torsion angle (FTA). 38 Twenty-eight femurs equally divided in 2 groups (normal and abnormal) were obtained from 14 39 dogs of different conformations (dolicomorphic and chondrodystrophic). 40 Computed tomographic (CT)-scans and 3D scanner acquisitions were used to create 41 stereolithographic (STL) files which were run in a computer-aided-design (CAD) platform. Two 42 blinded observers performed separately the measurements using the STL obtained from CT-scans 43 (CT aCAD) and 3D scanner (3D aCAD), which was considered the gold standard method. 44 The correlation coefficients were used to investigate the strength of the relationship between the 45 two measurements. 46 The accuracy for the aCAD computation was good, being always above the threshold of R2> 80% 47 for all three angles assessed in both groups. ALDFA and FNA were most accurate angles (accuracy 48 > 90 %). 49 The proposed 3D aCAD protocol can be considered a reliable technique to assess femoral angle 50 measurements in the canine femur. The developed algorithm automatically calculates the femoral 51 angles in 3D, thus considering the subjective intrinsic femur morphology. The main benefit relies 52 on a fast user-independent computation, which avoid user-related measurement variability. The 53 accuracy of 3D details may be helpful for patellar luxation and femoral bone deformity correction 54 as well as for the design of patient specific custom-made hip prosthesis implants. 55 56 Keywords: Accuracy, Dog; Femur; Computed tomography; Three-dimensional constructions; 3D 57 scanner 58 Introduction 59 60 The state of art for the measurement of angles in the canine femur has been traditionally limited to 61 multiple orthogonal radiographs (RX),1-3 which were gradually overtaken by the computed 62 tomography (CT)-scans 4,5 and magnetic resonance (MRI) evaluations. 6,7 These latter two 63 diagnostic techniques exhibit satisfactory aptitudes in terms of bone and images manipulation, 64 avoiding the positioning issue that frequently characterizes the radiographic evaluation.4,8 However, 65 CT and MRI lack on real three-dimensional (3D) measurement of angles since that almost for all 66 the values proposed by the literature were achieved with two-dimensional (2D) imaging. 6,9,10 67 Recently a 3D Python-based algorithm run on a computer-aided-design (CAD) software 68 (Rhinoceros version 5, Robert McNeell & Associates) was presented as a novel methodology for 69 the computation of femoral angles in the canine femur.11,12 The femoral angles computed, 70 differently from those obtained using different diagnostic techniques, 1-10 were measured in a real 71 3D fashion. The main benefit relies on automated measurements, which are independent from the 72 points selected by the operator, bone orientation and conformation as well. As a result, the operator-73 related measurement variability is decreased as the manual manipulation of the bone model and the 74 identification of target anatomical landmarks are not required. The repeatability and reproducibility 75 of the proposed protocol were assessed and compared with manual measurements made with 76 radiographs and CT reconstructions, finding that the 3D protocol was the most repeatable and 77 reproducible method.12 This conclusion was, also, supported by the automated design of the 3D 78 protocol, which restricts the potential user-related errors only to the operations required for the 79 creation of the mesh model and, therefore, remarkably decreases the computational time.11 80 However, the accuracy of 3D measurements, described as the difference of a measured value from a 81 true value, was not assessed and needed to be investigated. Therefore, the purpose of this study was 82 to determine the accuracy of our aCAD protocol for the computation of three femoral angles in 83 dogs: anatomical lateral distal femoral angle (aLDFA); femoral neck angle (FNA); and femoral 84 torsion angle (FTA). 85 Polygonal mesh models were created from 3D reconstructions of CT images and femoral angles 86 were computed with the developed protocol. The values obtained were compared to the 87 measurements performed with the same aCAD protocol but executed on polygonal mesh models 88 generated by 3D scans, which due to its high-resolution 3D nature, was assumed as the gold 89 standard technique for this study. 90 The second object of this study was to assess the efficacy of the aCAD protocol for the 91 measurement of femoral angles in either normal or abnormal femurs. 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 Materials & Methods 111 112 Fourteen canine paired pelvic limbs were collected. The cadavers were euthanized for reasons 113 unrelated with this project and a signed informed consent was requested before proceeding with 114 imaging acquisition and femur disarticulation. The study was conducted in a double-blind fashion 115 by two observers (an orthopaedic surgeon and an engineer). Moreover, one experienced radiologist 116 acquired all radiographic and CT images. He, also, anonymised all CT scans using a legend and 117 separately packed every femur sample to prevent any conditioning for the observers. 118 Specimens were first radiographed with digital radiographic equipment (Kodak Point of Care CR-119 360 System, Carestream Health). A standard ventro-dorsal and latero-lateral views were performed. 120 121 CT scans 122 CT scans were then acquired with four multi-detector row CT scanner (Toshiba Asteion S4, 123 Toshiba Medical Systems Europe). Dogs were positioned with a supine recumbence with legs 124 adducted, extended and tied above the stifles. An amperage of 150 mA, exposure time of 0.725 s 125 and voltage of 120 kV were set on. A slice thickness of 1 mm (reconstruction interval 0.8 mm) was 126 applied. CT images were reconstructed with a high-resolution filter for bones with the following 127 bone window (window length 1000 Hounsfield units, HU; window width 4000 HU). A 3D volume 128 reconstruction was done using a DICOM-processing software (Osirix version 2.7, pixmeo SARL). 129 The first observer isolated with Osirix every anonymized femur by cropping the tibia and pelvis, 130 avoiding unintentional modification of the profiles of the femoral head and condyles. Once the 131 femur model was separated, it was segmented using the procedure described by Longo et al.12 132 Briefly, using the region of interest (ROI) and 2D/3D growing region software functions, the 133 observer found the mean density femur values, which usually are major than 300 Hounsfield unit 134 (HU) and then set-up the segmentation parameters in a dedicated tool window. As a result, a 135 bitmapped (newly generated imaging series) was created and 3D reconstructed, through surface 136 rendering function. Finally, a 3D stereolithograophic (STL)13 file was saved and imported in the 137 Rhinoceros platform.11,12 138 139 3D scans 140 STL files were generated from 3D scans to obtain reference models on which compare femoral 141 angles measured on CT. Femurs were disarticulated at coxo-femoral and femoral-tibial joints, 142 dissected free from soft tissues excluding the patella and fabellae and stored in plastic bags at a -143 20°. A 3D scanner (Cronos 3D dual, Open technologies) was used for the femoral analysis. The 144 second observer positioned every anonymised femur on a circular rotating platform. The scanning 145 of the femur was performed adopting a triangulation technique, based on cameras, characterized by 146 a predetermined convergence angle and a fringe projector. The platform was automatically rotated 147 of a predetermined angle sequence, obtaining at least 5 to 10 acquisitions. A 3D geometrical bone 148 model was generated superimposing and aligning the multiple views of the model, obtained per 149 each sequence, by means of an engineering software (Optical RevEng, Open technologies). 150 Cleaning, filtering and closing-holes phases were used to delete model inaccuracies such as noises 151 and local spikes. As a result, a high-resolution mesh model of the bone was obtained and saved as a 152 STL file. The accuracy of the 3D scanner is ±30 µm.14 Similar results were obtained by the internal 153 verification procedure based on ISO (10360-8:2013) at the Laboratory of Design Tools and 154 Methods in Industrial Engineering. Considering that the 3D scanner accuracy is higher more than an 155 order of magnitude compared to CT axial resolution (0.8 mm), it is possible to assume the 3D scan 156 models as reference. 157 158 159 160 161 Automated-CAD measurements from CT reconstructions (CT aCAD) and 3D scanner 162 (3D aCAD) 163 Both observers imported each CT (Fig. 1) or 3D (Fig. 2) STL file in the CAD software where the 164 aCAD protocol was used to measure femoral angles. The aCAD computation was performed 165 following the same procedure steps described by Savio et al.11 In brief, the vertices inside the 166 femoral medullary canal (internal mesh) were selected and deleted. This operation is needed to 167 improve the quality of axis drawing and angle measurements, as the presence of internal vertices 168 may interfere with the automatic computation. Then, the femoral analysis was initiated by clicking 169 on the femoral head. To compute the femoral angles, the developed algorithm first identifies points, 170 planes and axis into the femur mesh. It performed all the measurements in few minutes through four 171 automatic phases: 1) femur alignment; 2) proximal femoral long axis computation; 3) analysis of 172 the proximal femoral epiphysis; 4) analysis of the distal femoral epiphysis. During these two final 173 phases, the vertices representing the femoral head and condyles were superimposed by spheres (Fig. 174 1 and 2).11,12 Finally, aLDFA, FNA and FTA angles were displayed on the screen and recorded by 175 the observer. 176 177 Groups 178 Considering radiographic, CT and visual gross evaluation, the specimens were examined for 179 evidence of osteoarthritis (OA) and difference of breed conformation (dolicomorphic vs 180 chondrodystrophic). The femurs were divided in two groups. Group 1 was assigned as normal, 181 adopting the following inclusion criteria: femurs were obtained from dolicomorphic breeds with no 182 evidence of OA. Whereas the second category was more heterogenic and included femurs either 183 affected by OA regardless of conformation or taken from chondrodrystophic breeds (Fig. 3). 184 The radiologist radiographically evaluated the degree of OA and converted the OA score to a 185 numeric scale (0= none; 1= mild; 2= moderate ;3= severe).15,16 186 187 188 Statistical analysis 189 The statistical analyses were performed using a commercially available software (SAS 9.4, SAS 190 Institute Inc., Cary, NC, USA). Normality distribution hypothesis was assessed by Shapiro-Wilk 191 test. A linear regression analysis was applied, considering the gold standard method (3D aCAD) as 192 the independent variable and the CT aCAD as the dependent variable. 193 The adjusted R2 was used to quantify the strength of the relationship between the angle measured 194 through CT aCAD (observer 1) and 3D aCAD (observer 2) techniques. Adjusted R2 values > 80 % 195 were considered acceptable. The hypotheses of the linear model on the residuals were graphically 196 assessed. 197 The descriptive statistics (means, standard deviations, medians and interquartile ranges) were 198 calculated for each angle (aLDFA, FNA and FTA) measurements for both imaging techniques. 199 The paired Student t-test was performed to compare the data recorded with CT aCAD and the gold 200 standard. Statistical significance of P-value was set at < 0.05. 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 Results 217 218 Twenty-eight femurs divided in two groups (1 = normal, 2= abnormal) of 14 femurs each, were 219 used for this study. The specimens were obtained from dogs of different breeds and conformations: 220 3 mixbreed dogs, 2 Dachshunds, 2 French bouledogs and 1 Pug, German shepherd, Labrador R., 221 Bernese mountain dog, Segugio italiano, Amstaff and Great Dane. Ten dogs were intact males, 3 222 were spayed females and 1 was a not-spayed female. The overall mean body weight was 19.5 kg 223 (range 4-44 kg), whereas the body weight means of the groups were: group 1 (16.1 kg, range 13-28 224 kg) group 2 (19.3 kg, range 4-44 kg). The overall mean age was 9.5 years (range 2-15 years). The 225 mean age of group 1 was 4.7 years (range 2-8 years), while group 2 had a mean age of 12.5 years 226 (range 9-15 years). 227 Group 1 included 14 dolicomorphic femurs with no evidence of radiographic OA. Within the 14 228 femurs of the group 2, there were: 4 chondrodystrophic femurs not affected by OA, 6 229 chondrodystrophic femurs affected by OA (mean OA score: 1) and 4 dolicomorphic femurs affected 230 by OA (mean OA score: 2). 231 All data regarding the 3 angles and for both CT aCAD and 3D aCAD measurements were normally 232 distributed (Shapiro-Wilk test >0.9). The values of the angles recorded were well aligned along 233 regression lines in almost all the samples, excepted for some femurs included in group 2 (Fig. 4) 234 The adjusted R2 value of the CT aCAD and 3D aCAD measurements resulted always above the 235 acceptance criterion of 80%, regardless of the angle measured and the group considered. Overall, 236 the coefficients calculated for all 28 femurs were: aLDFA > 95%; FNA > 95% and FTA > 86% 237 (Fig. 4). Specifically, within group 1 the coefficients were: aLDFA > 93%; FNA > 93% and FTA > 238 98%, while within group 2: aLDFA > 97%; FNA > 94% and FTA > 82% (Fig. 4). Technique-239 related means, medians and interquartile ranges values for the 3 angles are displayed in Table 1. 240 The t-test showed that there was not a statistically significant difference (P < 0.05) in the mean 241 difference values of each paired measurements for every angle assessed, excepted for FTA 242 measurement in group 2 (Table 2) 243 Discussion 244 245 This study investigated the accuracy of a novel automated 3D technique (aCAD) for the 246 computation of canine femoral angles. We used the correlation coefficients to assess the strength of 247 the relationship between the angle measurements performed by the observers in Rhino starting from 248 STL files created either from CT-scans (CT aCAD) or 3D scanner (3D aCAD). The aCAD 249 methodology has, looking at the accuracy investigation, been satisfactory for all three angles 250 assessed (> 82%). 251 This suggests that the CT aCAD measurements were comparable to the 3D aCAD measurements, 252 which represented our reference standard method of assessment. The practical consequence is that 253 the developed 3D protocol is not only repeatable and reproducible12 but also may be considered 254 enough accurate. However, a validation of the 3D scanner on bone measurements needs to be 255 performed to corroborate this subjective assumption. 256 The accuracy of a test is a description of how close a measured value is to an assumed true value; 257 which means that a “true” value must be both identifiable and measurable, thus providing an 258 unequivocal gold standard against which new tests may be assessed.4,16 In this study, we have 259 assumed 3D scanner measurements of femoral anatomic specimens as the gold standard method for 260 two main reasons. First, 3D scanner allows for creating detailed and precise geometrical bony 261 models17 that could nicely reproduce the original femoral morphology. We have applied white spray 262 onto the femoral specimens and waited at least 24 hours before the image acquisition with the 263 scanner. The aim was to increase the visualization of the femoral cortices, decreasing the radio-264 transparency of the bone and thus improve the quality of the femoral captures. Second, 3D scanner 265 allows the user to work with real 3D files, which we cannot obtain from other reported two-266 dimensional techniques. 9,18 It may be argued that we could have either measured the femoral angles 267 on digital photography images of femur specimens or calculated them directly onto the bones. 268 Although, the quantification of an established “true” value for a such variable measurement (angle) 269 depends on arbitrary anatomic landmarks, in the authors’ opinion a comparison between a 3D 270 technique (aCAD) with a 2D gold standard method (digital photography) wouldn’t be feasible. The 271 reason is attributable to the structured differences of the methodologies tested. 272 A direct measurement of femoral angles onto femurs specimens could have represented an 273 alternative gold standard. However, we believe that such method couldn’t represent an accurate 274 methodology as well because precise anatomic reference lines needed to be drawn, increasing the 275 risk of operator-measurement errors. 276 Overall, the aLDFA and FNA were the most accurate angles since that their correlation coefficients 277 were always above the 90% threshold, regardless of the groups considered. FTA measurements 278 were still satisfactory but showed a lesser accuracy. These results partially confirmed the data that 279 we previously presented.12 Specifically, the aLDFA represents the most repeatable, reproducible 280 and accurate angle to measure. The FNA, which resulted as the lesser repeatable and reproducible 281 angle to be quantified with three different diagnostic techniques (RX, CT and aCAD computation), 282 here exhibited comparable values between CT aCAD and 3D aCAD. Whereas, the measurements 283 recorded for the FTA resulted as the most out of range from the real values, but still within the 284 established threshold of acceptance (> 80%) in both normal and abnormal femurs. 285 The computation ability of the developed protocol in femurs of different dimensions, conformations 286 (dolicomorphic and chondrodystrophic breeds) as well as in femurs affected or unaffected by OA, 287 represented a key point of our project. Previously, the described 3D protocol was performed mrely 288 on normal femurs, free of orthopaedic diseases.11,12 289 The femoral angles measured by the observers are commonly quantified in the preoperative 290 planning of patellar luxation ,5,19 which is frequently caused by femoral deformities.20,21 These 291 skeletal malformations cause imbalanced joint loading and when they are either severe or lately 292 diagnosed (chronic), they may lead to OA which deforms the articular profiles.22-24 In this study, 10 293 out of 28 femurs were affected by OA, of which one (femur 19) had a severely arthritic femoral 294 head (OA score: 3) (Fig. 5) and two (femurs 25 and 26) had the condylar profiles altered (OA score: 295 2). The massive remodeling of the articular profiles, above all of the femoral head, represents both a 296 challenge for the computational analysis and a plausible explanation for a less than perfect accuracy 297 detected for the FTA. The algorithm needs to correctly identify and fit the original sphere of the 298 femoral head and condyles. During the pilot developing phase, the algorithm was set up to exclude 299 from the analysis all the vertices that belong to external components of the femoral head fitting such 300 as osteophytes, which could potentially alter the computational analysis.11 The FTA correlation 301 coefficient obtained for the computation of abnormal femurs (R2=82%) means that the algorithm 302 effectively analyses also deformed femoral heads but not as accurately as for FNA and aLDFA 303 computation (≥ 92%). Considering the satisfactory FTA accuracy in the normal group (R2 FTA > 304 98%), we attribute the lower FTA accuracy in abnormal femurs mainly to the difficulty of analysing 305 severely altered femoral head profiles. However, the accuracy obtained was still major than 80% 306 threshold (R2=82%). 307 The descriptive statistic displayed in Table 1 shows that the values measured for FNA and FTA fall 308 within the ranges described in the literature: FNA (125°-138°) 3,25 and FTA (12-40°). 2, 25 309 The FNA and especially FTA reference ranges are wide.2,3,25 In the authors’ opinion this is 310 concerning and need to be clarified as femoral torsion is frequently detected in case of patellar 311 luxation and need to be often corrected. The accepted clinical tolerance for FTA suggests that there 312 is a variable either depending on the femur morphology or on the observer ability which influences 313 the angle measurements. Explanations may rely on the identification of the target points such as the 314 center of the femoral head and neck, which could be challenging for the observer, especially in the 315 case of severe OA. Our FTA mean ranges from 20-22° (table 1), which agrees with our previous 316 results11,12 and with the literature ranges.2,25 However, sometimes a 27° reference value for femoral 317 torsional deformity is assumed,20 and therefore the obtained FTA mean implies that our 3D 318 technique identifies a more retroverted position of the femoral head. Whether this result may have a 319 clinical impact could not be answered with this study and therefore need to be further investigated. 320 The aLDFA mean values, accordingly with those already found by the authors’ 11,12 are slightly 321 lower than the reported range (aLDFA 94-98°).3, 25 We impute this result mainly to morphologic 322 heterogeneity of the femurs computed. We analyzed a range of femurs of different dimensions 323 (small to large dogs) and conformations (dolicomorphic and chondrodystrophic), while the data 324 reported in literature were obtain mainly in large dolicomorphic dogs.16,25 It is plausible to expect 325 that chondrodystrophic dogs as well as small size breeds may be characterized by different values 326 regarding frontal and torsional femoral alignment. Furthermore, the t-test analysis exhibited a not 327 significant difference for each paired of values assessed. In almost for all the cases evaluated, the 328 CT aCAD measurements tended towards underestimating the femoral angle values compared to the 329 gold standard, but this tendency was statistically significant only for the femoral torsion evaluation 330 in the group of abnormal femurs (Table 2). 331 332 Conclusions 333 We have shown that the automatic measurements obtained from CT derived data are 334 significantly comparable with high-resolution 3D scanner-derived data, suggesting that the tested 335 automated CAD technique is an accurate methodology for measuring femoral angles in both normal 336 and abnormal canine femurs. However, currently it is not validated what should a gold standard be 337 for 3D measurements. Therefore, further studies could be undertaken to compare anatomical versus 338 3D scanner measurements of bones. 339 The presented methodology could represent a reliable diagnostic method to adopt when a femoral 340 deformity is suspected, having the automated and 3D nature of its assessments and rapidity of its 341 computational analysis as main substantial benefits. Moreover, the precision of patellar luxation 342 planning may increase, due to the user-independent structure of measurements. Finally, the 343 possibility to correctly identify anatomic landmarks such as the original curvature of the femoral 344 head, the external and internal profiles of the femoral neck, and potentially the original morphology 345 of the acetabulum, even in the case of a severe degenerative joint disease, may extends its 346 usefulness in the future, also, for arthroplasty purposes. However, further evaluations need to be 347 done with a greater number of samples to improve the quality and the precision of the femur 348 computation in severely arthritic femoral heads. 349 350 Conflict of interest statement 351 None of the authors of this paper have a financial or personal relationship with other people 352 or organisations that could inappropriately influence or bias the content of the paper. 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 References 382 383 384 1. Bardet JF, Rudy RL, Hohn RB. Measurement of femoral torsion in dogs using a 385 biplanar method. Vet Surg 1983;12:1-6. 386 387 2. Montavon, P.M., Hohn, R.B., Olmestead, M.L., Rudy, R.L., 1985. Inclination and 388 anteversion angles of the femoral head and neck in the dog evaluation of a standard method 389 of measurement. Vet Surg 1985;14:272-282. 390 391 3. Tomlison, J, Fox D, Cook JL, et al. Measurement of femoral angles in four dog breeds. Vet 392 Surg 2007;36:593-598. 393 394 4. Oxley B, Gemmill TJ, Pink J, et al. Precision of a novel computed tomographic method for 395 quantification of femoral varus in dogs and an assessment of the effect of femoral 396 malpositioning. Vet Surg 2013;42:751-758. 397 398 399 5. Barnes, D.M., Anderson, A.A., Frost, et al. Repeatability and reproducibility of 400 measurements of femoral and tibial alignment using computed tomography multiplanar 401 reconstructions. Vet Surg 2015;44:85-93. 402 403 6. Kaiser S, Cornely D, Golder W, et al. The correlation of canine patellar luxation and the 404 anteversion angle as measured using magnetic resonance images. Vet Radiol Ultrasound 405 2001;42:113-118. 406 407 hat formatiert: Deutsch (Schweiz) 7. Ginja MMD, Ferreira, AJA, Jesus SS, et al. Comparison of clinical, radiographic, computed 408 tomographic, and magnetic resonance imaging methods for early prediction of canine hip 409 laxity and dysplasia. Veterinary Radiol Ultrasound 2009;50:135-143. 410 411 8. Jackson GM, Wendelburg KL. Evaluation of the effect of distal femoral elevation on 412 radiographic measurement of the anatomic lateral distal femoral angle. Vet Surg 413 2012;41:994-1001. 414 415 9. Dudley RM, Kowaleski MP, Drost, WT, et al. Radiographic and computed 416 tomographic determination of femoral varus and torsion in the dog. Veterinary Radiol 417 Ultrasound 2006;47:546-552. 418 419 10. Yasukawa S, Edamura K, Tanegashima K, et al. Evaluation of bone deformities of the 420 femur, tibia, and patella in Toy Poodles with medial patellar luxation using computed 421 tomography. Vet Comp Orthop Tramatol 2016;29:29-38. 422 423 11. Savio G, Baroni T, Concheri G, et al. Computation of femoral canine morphometric 424 parameters in three-dimensional geometrical models. Vet Surg 2016;45:987-995. 425 426 12. Longo, F., Nicetto, T., Banzato T, et al. Automated computation of femoral angles in dogs 427 from three-dimensional computed tomography reconstructions: Comparison with manual 428 techniques. Vet J 2018;232:6–12. 429 430 13. Botsch M, Kobbelt L, Pauly M, et al. Mesh data structures. In: Botsch ed. Polygon Mesh 431 Processing, 1st Edn. MA, USA: A.K. Peters, 2010 pp. 21-28 432 433 14. https://www.growshapes.com/store/p137/Open-Technologies-3D-Scanner/Cronos-3D-Dual-434 3MP.html. Last access 12/04/2019. 435 436 15. Lopez MJ, Lewis BP, Swaab ME, et al. Relationships among measurements obtained by use 437 of computed tomography and radiography and scores of cartilage microdamage in hip joints 438 with moderate to severe joint laxity of adult dogs. Am J Vet Res 2008;69:362-370. 439 440 16. D’Amico LL, Xie L, Abell LK et al. Relationships of hip joint volume ratios with degrees of 441 joint laxity and degenerative disease from youth to maturity in a canine population 442 predisposed to hip joint osteoarthritis. Am J Vet Res 2011;72:376-383. 443 444 17. Palmer RH, Ikuta CL, Cadmus JM 2011. Comparison of femoral angulation measurement 445 between radiographs and anatomic specimens across a broad range of varus conformations. 446 Vet Surg 2011;40:1023-1028. 447 448 18. Fahrni S, Campana L, Dominguez A, et al. CT-scan vs. 3D surface scanning of a skull: first 449 considerations regarding reproducibility issues. Forensic Sciences Research 2017;2:99-99. 450 451 19. Swiderski JF, Radecki SV, Park RD, et al. Comparison of radiographic and anatomic 452 femoral varus angle measurements in normal dogs. Vet Surg 2008;37:43-48. 453 454 20. Gibbons SE, Macias C, Tonzing MA, et al. Patellar luxation in 70 large breed dogs. J Small 455 Anim Pract 2006; 47: 3–9. 456 457 hat formatiert: Deutsch (Schweiz) hat formatiert: Deutsch (Schweiz) hat formatiert: Deutsch (Schweiz) hat formatiert: Deutsch (Schweiz) 21. Brower BE, Kowaleski MP, Peruski AM, et al. Distal femoral lateral closing wedge 458 osteotomy as a component of comprehensive treatment of medial patellar luxation and 459 distal femoral varus in dogs. Vet Comp Orthop Tramatol 2017;20-27. 460 461 22. Roch SP, Gemmill TJ. 2008. Treatment of medial patellar luxation by femoral closing 462 wedge ostectomy using a distal femoral plate in four dogs. J Small Anim Pract 463 2008;249:52–158. 464 465 23. Dobbe J, du Pre` KJ, Kloen P, et al. Computer-assisted and patient-specific 3-D planning 466 and evaluation of a single-cut rotational osteotomy for complex long-bone deformity. 467 Medical & Biological Engennering & Computing 2011;49:1363–1370. 468 469 24. Milner SA, Davis TR, Muir KR, et al. Long-term outcome after tibial shaft fracture: is 470 malunion important? Journal of Bone and Joint Surgery American volume 2002; 84A:971–471 980. 472 473 25. Petazzoni, M. Radiographic measurements of the femur. In: Petazzoni, M., Jaeger, G.H. 474 (Eds). Atlas of Clinical Goniometry and Radiographic Measurements of the Canine Pelvic 475 Limb. 2nd Edn. Merial, Milano, Italy: Merial, 2008, pp. 34-54. 476 477 478 479 480 481 482 483 484 Table 1 descriptive statistics measured with both computed tomography (CT aCAD: tested 485 protocol) and 3D scanner (3D aCAD: gold standard) techniques for each angle. 486 487 Technique aLDFA FNA FTA CT aCAD Mean ± SD 92.51 ± 5.4 125.32 ± 10.2 21.96 ± 7.1 Median 92.7 127.96 21.58 IQR 7.7 8.28 8.8 3D aCAD Mean ± SD 92.55 ± 5.3 124.26 ± 10.8 20.87 ± 6.4 Median 92.2 126.8 20.2 IQR 6.95 11.85 6.25 488 Table 2 Mean difference and P-value of paired t-test calculated for each angle. 489 T- test aLDFA FNA FTA Normal Abnormal Normal Abnormal Normal Abnormal Mean - 0,14° 0,22° - 0.24 - 0.24 - 0.41 - 1.77 hat formatiert: Deutsch (Schweiz) hat formatiert: Deutsch (Schweiz) hat formatiert: Deutsch (Schweiz) Mean Difference ± SD ± 1,16 ± 0,79 ± 0.82 ± 3.82 ± 0.79 ± 3.21 P-value 0,65 0,3 0,29 0,08 0,07 0,05 490 491 492 493 494 495 496 497 498 499 500 501 502 Figure legends 503 504 505 Fig. 1. 3D computation performed in a stereolithographic file obtained from a computed 506 tomography reconstruction (CT aCAD) of a 2-years-old French Bouledog. After the 3D 507 computation, femoral axes appear in the bone model (A). The green line is the femoral head and 508 neck axis (FHNA), the blue lines represent the mechanical axis (MA) and the hip joint orientation 509 line (HJOL), the red line is the proximal femoral long axis (PFLA) and the gold line is the 510 transcondylar axis (TCA). (B) Cranial and caudal aspect of the proximal femoral epiphysis. Notice 511 the fitting of the femoral head and the section of the femoral neck (light blue). (C) Medial-lateral 512 and caudal-cranial views of the femoral condyles. Note the sphere fitting of both condyles (light 513 blue spheres) as well as the green vertices that represent the contact area of the TCA. 514 515 Fig. 2. 3D computation performed in a stereolithographic file obtained from a 3D-scanner 516 acquisition (3D aCAD) of a 4-years-old Bernese mountain dog (A). The green line is the femoral 517 head and neck axis (FHNA), the blue lines represent the mechanical axis (MA) and the hip joint 518 orientation line (HJOL), the red line is the proximal femoral long axis (PFLA) and the gold line is 519 the transcondylar axis (TCA). (B) Cranial and caudal aspect of the proximal femoral epiphysis. 520 Notice the presence of red vertices outside of the femoral head fitting which represent parts of the 521 acetabulum excluded from the computation. (C) Medio-lateral and caudal-cranial aspects of the 522 distal femoral epiphysis. TCA, PFLA and MA are visible. 523 524 Fig. 3. Cranio-caudal views of four abnormal femurs after importation on Rhinoceros. (A) Right 525 femur of a 12-years-old German Shepherd severely affected by osteoarthritis (OA) of the femoral 526 head. (B) Right femur of a 10-years-old Pug which had a severe degeneration of the femoral head 527 and neck. (C and D) Left chondrodystrophic femurs affected by mild (C) and severe OA (D) of the 528 distal femoral epiphysis. The dogs were an 8-years-old French Bouledog and a 13 years-old 529 Dachshund. 530 531 Fig. 4. Graphical representation of the regression analysis. Line (A): regression 532 line of the totality of the femurs assessed for each angle. The R2 are >80 % for all three 533 angles. Line (B): regression analysis of group 1 (normal femurs). The R2 are > 93 %, having the 534 FTA measurement as the most accurate angle. Line (C): graphical representation of the 535 regression of group 2 (abnormal). The aLDFA angle was the most accurate (R2> 93 %), while the 536 FTA the most challenging to measure (R2> 82 %). 537 538 Fig. 5. Digital cranio-caudal photograph of the femur specimen of a 12-years-old German 539 Shepherd. (B and C) Cranial and caudal views of the femoral head and neck. The green line is the 540 femoral head and neck axis (FHNA), the blue lines represent the mechanical axis (MA) and the hip 541 joint orientation line (HJOL), the red line is the proximal femoral long axis (PFLA). Observe that 542 the osteophytes fall outside the green sphere and are not considered for fitting of the femoral head. 543 (D) Caudal view of the femoral condyles: the MA and transcondlyar axis (gold line) are drawn. (E) 544 Femoral cranio-caudal view after the 3D computation. 545 546 547 548 549 550