Diagnostic validation of vertebral heart score machine learning algorithm for canine lateral chest radiographs

OBJECTIVES

The vertebral heart score is a measurement used to index heart size relative to thoracic vertebra. Vertebral heart score can be a useful tool for identifying and staging heart disease and providing prognostic information. The purpose of this study is to validate the use of a vertebral heart score algorithm compared to manual vertebral heart scoring by three board-certified veterinary cardiologists.

MATERIALS AND METHODS

A convolutional neural network centred around semantic segmentation of relevant anatomical features was developed to predict heart size and vertebral bodies. These predictions were used to calculate the vertebral heart score. An external validation study consisting of 1200 canine lateral radiographs was randomly selected to match the underlying distribution of vertebral heart scores. Three American College of Veterinary Internal Medicine board-certified cardiologists were enrolled to manually score 400 images each using the traditional Buchanan method. Post-scoring, the cardiologists evaluated the algorithm for misaligned anatomic landmarks and overall image quality.

RESULTS

The 95th percentile absolute difference between the cardiologist vertebral heart score and the algorithm vertebral heart score was 1.05 vertebrae (95% confidence interval: 0.97 to 1.20 vertebrae) with a mean bias of -0.09 vertebrae (95% confidence interval: -0.12 to -0.05 vertebrae). In addition, the model was observed to be well calibrated across the predictive range.

CLINICAL SIGNIFICANCE

We have found the performance of the vertebral heart score algorithm comparable to three board-certified cardiologists. While validation of this vertebral heart score algorithm has shown strong performance compared to veterinarians, further external validation in other clinical settings is warranted before use in those settings.

Utility of vertebral left atrial size and vertebral heart size to aid detection of congestive heart failure in dogs with respiratory signs

BACKGROUND

Differentiating cardiogenic vs noncardiogenic causes of respiratory signs can be challenging when echocardiography is unavailable. Radiographic vertebral left atrial size (VLAS) and vertebral heart size (VHS) have been shown to predict echocardiographic left heart size, with VLAS specifically estimating left atrial size.

HYPOTHESIS/OBJECTIVES

Compare the diagnostic accuracy of VLAS and VHS to predict left-sided congestive heart failure (CHF) in dogs presenting with respiratory signs.

ANIMALS

One-hundred fourteen dogs with respiratory signs and radiographic pulmonary abnormalities.

METHODS

Retrospective cross-sectional study. Dogs had to have an echocardiogram and thoracic radiographs obtained within 24 hours. Diagnosis of CHF was confirmed based on the presence of respiratory signs, cardiac disease, LA enlargement, and cardiogenic pulmonary edema.

RESULTS

Fifty-seven dogs had CHF and 57 did not have CHF. Compared to VHS (area under the curve [AUC] 0.85; 95% confidence interval [CI], 0.77-0.91), VLAS was a significantly (P = .03) more accurate predictor of CHF (AUC, 0.92; 95% CI, 0.85-0.96). Optimal cutoff for VLAS was >2.3 vertebrae (sensitivity, 93.0%; specificity, 82.5%). Murmur grade (P = .02) and VLAS (P < .0001) were independently associated with CHF and VHS was not. Increased VHS (54%) was significantly (P = .01) more common than increased VLAS (24%) in dogs without CHF. Results were similar in a subsample of older and smaller dogs.

CONCLUSIONS AND CLINICAL IMPORTANCE

When echocardiography is unavailable, VLAS and murmur grade have clinically utility to aid in differentiating cardiogenic from noncardiogenic respiratory signs. These findings might be especially useful to help rule out CHF in dogs with increased VHS that present with respiratory signs.

The heart to single vertebra ratio: A new objective method for radiographic assessment of cardiac silhouette size in dogs

Vertebral heart size (VHS) is widely determined in clinical practice as an objective method to assess the cardiac silhouette dimensions. However, a key limitation is that it is difficult to determine VHS in dogs with vertebral alterations. This retrospective, method comparison, observer agreement study sought to overcome this limitation by using the heart-to-single vertebra ratio (HSVR), by evaluating the level of agreement between VHS and HSVR, as well as the intra- and inter-observer agreement for HSVR. Three independent observers retrospectively evaluated thoracic radiographs obtained over a set time period. Exclusion criteria were the presence of alterations of the thoracic spine and the inability to clearly outline the cardiac silhouette. The lengths of the vertebral bodies, from the fourth to eighth thoracic vertebra, and VHS were measured on each radiograph. The HSVR was calculated by dividing the sum of the cardiac long and short axes by the length of each vertebral body. Eighty dogs of different breeds were included in the final analysis. Lin's concordance correlation coefficients revealed strong correlations between VHS and HSVR (0.91-0.96), and the Bland-Altman plots showed low bias (0.01-0.2) between the methods. The mean absolute errors indicated low average magnitudes of error (0.11-0.28). The intraclass correlation coefficients showed good to excellent inter-observer (0.87-0.92; P = 0.000) and intra-observer (0.87-0.99; P < .001) agreement. In the authors' opinion, this new method, which is less time consuming and more objective, could offer a valuable alternative to VHS.

Cardiac Radiographic Measurements in Ferrets Using the OsiriX MD Programme

Objectives: To adapt the vertebral heart scale (VHS) for use in ferrets and identify new scales and tools that allow to establish the normal heart size by means of radiography more quickly and effectively.

Methods: Forty healthy pet ferrets (Mustela putorius furo) were used in this prospective study. The measurements were made on right lateral, left lateral, ventrodorsal, and dorsoventral projections, using OsiriX MD medical imaging software, to evaluate sex effect and variance within the different heart scales. Cardiac measurements were also correlated to VHS and the cardiac dimension in the same projection.

Results: Most of the cardiac measurements were significantly different between males and females. The results for the VHS were: right lateral VHS (RL-VHS): 5.52 ± 0.28 v (vertebrae units); left lateral (LL-VHS): 5.55 ± 0.28 v; and dorsoventral VHS (DV-VHS): 6.22 ± 0.34 v for males and RL-VHS: 5.24 ± 0.2 v; LL-VHS: 5.25 ± 0.20 v; and DV-VHS: 5.97 ± 0.35 v for females. Regarding the sternebral heart scale (SHS), the values were: RL-SHS: 5.10 ± 0.20 s (sternebrae units) and LL-SHS: 5.11 ± 0.20 s for males and RL-SHS: 4.67 ± 0.24 s and LL-SHS: 4.67 ± 0.28 s for females. The new measurements based on determining the cardiac area were also marked by clear sexual dimorphism, as shown for the cardiac area-axis (AREA-AXIS): RL-AREA-AXIS: 3.82 ± 0.45 cm²; LL-AREA-AXIS: 3.87 ± 0.41 cm²; ventrodorsal (VD)-AREA-AXIS: 4.59 ± 0.64 cm²; and DV-AREA-AXIS: 4.80 ± 0.50 cm² for males and RL-AREA-AXIS: 2.39 ± 0.23 cm²; LL-AREA-AXIS: 2.41 ± 0.26 cm²; VD-AREA-AXIS: 3.08 ± 0.45 cm²; and DV-AREA-AXIS: 3.06 ± 0.47 cm² for females. The cardiac area open polygon (AREA-POL) values were: RL-AREA-POL: 6.78 ± 0.65 cm²; LL-AREA-POL: 6.88 ± 0.68 cm²; VD-AREA-POL: 7.20 ± 0.91 cm²; and DV-AREA-POL: 7.57 ± 0.88 cm² for males and RL-AREA-POL: 4.28 ± 0.30 cm²; LL-AREA-POL: 4.35 ± 0.35 cm²; VD-AREA-POL: 4.72 ± 0.65 cm²; and DV-AREA-POL: 4.79 ± 0.66 cm² for females, with similar differences noted from various radiographic projections. A good correlation was noted between VHS and SHS, and a very strongly positive correlation existed between cardiac area measurements and cardiac dimensions.

Conclusion: The VHS adapted to ferrets, the SHS, as well as the cardiac area measurements presented in our study are ideal tools for the assessment of cardiac size in ferrets.

Comparison of a Deep Learning Algorithm vs. Humans for Vertebral Heart Scale Measurements in Cats and Dogs Shows a High Degree of Agreement Among Readers

Heart disease is a leading cause of death among cats and dogs. Vertebral heart scale (VHS) is one tool to quantify radiographic cardiac enlargement and to predict the occurrence of congestive heart failure. The aim of this study was to evaluate the performance of artificial intelligence (AI) performing VHS measurements when compared with two board-certified specialists. Ground truth consisted of the average of constituent VHS measurements performed by board-certified specialists. Thirty canine and 30 feline thoracic lateral radiographs were evaluated by each operator, using two different methods for determination of the cardiac short axis on dogs' radiographs: the original approach published by Buchanan and the modified approach proposed by the EPIC trial authors, and only Buchanan's method for cats' radiographs. Overall, the VHS calculated by the AI, radiologist, and cardiologist had a high degree of agreement in both canine and feline patients (intraclass correlation coefficient (ICC) = 0.998). In canine patients, when comparing methods used to calculate VHS by specialists, there was also a high degree of agreement (ICC = 0.999). When evaluating specifically the results of the AI VHS vs. the two specialists' readings, the agreement was excellent for both canine (ICC = 0.998) and feline radiographs (ICC = 0.998). Performance of AI trained to locate VHS reference points agreed with manual calculation by specialists in both cats and dogs. Such a computer-aided technique might be an important asset for veterinarians in general practice to limit interobserver variability and obtain more comparable VHS reading over time.

Breed-specific vertebral heart score, vertebral left atrial size, and radiographic left atrial dimension in Cavalier King Charles Spaniels: Reference interval study

Cavalier King Charles Spaniels (CKCS) are predisposed to developing myxomatous mitral valve disease (MMVD), with radiographs frequently used to screen for evidence of left-sided cardiomegaly secondary to MMVD. Vertebral heart size (VHS), vertebral left atrial size (VLAS), modified VLAS (M-VLAS), and radiographic left atrial dimension (RLAD) are reported as objective measurements of global heart size and left atrial size. Normal VHS in CKCS (10.6 ± 0.5) is reportedly higher than the non-breed-specific value (9.7±0.5). Breed-specific VLAS, M-VLAS, and RLAD cut-offs have not been reported in CKCS. The aim of this prospective reference interval study was to describe the VHS, VLAS, M-VLAS, and RLAD values for 30 clinically healthy adult CKCS. Inclusion criteria were unremarkable physical examination, normal echocardiography, and thoracic radiographs without malposition/abnormalities. There were 22 female and eight male dogs. Ages ranged from 1 to 6 years. The VHS mean value in our sample was 10.08 ± 0.56 (95% range, 9.87-10.29). This was significantly greater than a previously published general canine reference value of 9.7 ± 0.5 and significantly less than a previously published CKCS breed-specific value of 10.6 ± 0.5 (P < 0.01). Mean VLAS, M-VLAS, and the RLAD values in our study were 1.79 ± 0.3 (95% range, 1.68-1.9), 2.23 ± 0.44 (95% range, 2.06-2.39), and 1.2 ± 0.34 (95% range, 1.07-1.33), respectively. These were significantly less than previously published reference interval values (P < 0.001). The VHS, M-VLAS, and the RLAD were not affected by sex, body weight, or BCS; whereas the VLAS was moderately affected by body weight. Findings from this study can be used as background for future thoracic radiographic assessments in CKCS.

The vertebral heart scale on CT is correlated to radiographs in dogs

The vertebral heart scale (VHS) has long been used as an objective standard for evaluation of cardiac silhouette size on thoracic radiographs and plays a key role in the diagnosis as well as the assessment of the severity of canine heart disease. Based on our review of the literature, there has been no published study describing an objective method for measuring overall heart size using computed tomography (CT) in canine patients. The goals of this exploratory prospective method comparison study were to describe a method for objectively evaluating heart size on canine thoracic CT images, determine whether there was a correlation between the VHS when applied to a right lateral thoracic radiograph and a sagittal image of a thoracic CT scan, and determine the effect of cardiac phase on a VHS measured on CT. A method for measuring VHS on canine CT studies was defined. Data were collected on canine patients admitted to The Ohio State Veterinary Medical Center. Each patient received an ECG-gated thoracic CT immediately followed by a right lateral thoracic radiograph. The VHS was measured on non-gated, systolic, and diastolic sagittal CT reformatted images. Our results indicated that cardiac gating may not be necessary when assessing canine heart size on CT. When comparing the VHS on gated and non-gated CT to the VHS on radiographs, there was a moderate to high degree of correlation; however, there was a high degree of variability between CT and radiography.

The Longitudinal Outcome Of Canine (K9) myxomatous mitral valve disease (LOOK-Mitral registry): Baseline characteristics

INTRODUCTION

The Longitudinal Outcome Of Canine (K9) myxomatous mitral valve disease (MMVD) registry (LOOK-Mitral registry) was established to describe the natural history and predictors of outcome in dogs affected by MMVD. This study was intended to describe the baseline characteristics of dogs in the LOOK-mitral registry.

ANIMALS

Dogs with echocardiographic evidence of MMVD were prospectively enrolled by thirteen referral centers.

RESULTS

A total of 6102 dogs with MMVD were included. The median age was 10 years (1-19 years), and mixed breed was the most common breed (n = 1,360, 22%). Concomitant diseases were reported in 2459 dogs with chronic respiratory diseases occurring most frequently (14%), followed by the presence of azotemia (6%) and orthopedic diseases (5%). Regarding disease severity, 65% of dogs were in ACVIM Stage-B1, 15% in Stage-B2, and 20% in Stage-C. Dogs in Stage-B1 were younger (p<0.001) than dogs in other stages. Murmur intensity, heart rate during physical examination, and radiographic vertebral heart score were positively associated with the stage. Dogs in Stage-C were more likely to have tachypnea (p<0.001), dyspnea (p<0.001), cough (p<0.001), syncopal episodes (p<0.001), and tachyarrhythmias (p<0.001) compared to dogs in Stage-B1 and B2. Echocardiographic indices of left atrial and ventricular size were positively correlated with the ACVIM stage. Interestingly, 4% of dogs that weighed <20 kg had an increased normalized end-systolic left ventricle internal diameter (>1.26).

CONCLUSIONS

This study contributes to a better understanding of the clinical characteristics of dogs affected by MMVD and provides new findings that may be of clinical relevance.

Radiographic quantification of left atrial size in dogs with myxomatous mitral valve disease

Background

In the absence of echocardiography, identification of cardiomegaly via thoracic radiography is a necessary criterion for classification of disease severity in dogs with myxomatous mitral valve disease (MMVD).

Objective

Modified‐vertebral left atrial size (M‐VLAS) facilitates objective radiographic assessment of the left atrium (LA) in 2 dimensions and identifies LA enlargement more accurately than existing methods.

Animals

Sixty‐four dogs with various stages of MMVD and 6 control healthy dogs.

Methods

Retrospective case–control study. Medical records were searched for dogs with varying severity of MMVD. Modified‐vertebral left atrial size, vertebral left atrial size (VLAS), vertebral heart size (VHS), and radiographic left atrial dimension (RLAD) were measured from thoracic radiographs and compared with echocardiographically derived measurements.

Results

Positive correlation to LA/Ao was identified for M‐VLAS (r = 0.77, P < .001), VLAS (r = 0.76, P < .001), RLAD (r = 0.75, P < .001), and VHS (r = 0.67, P < .001). Receiver operating characteristic analyzes provided an area under the curve of 0.97 (95% CI, 0.94‐1.00) for M‐VLAS, which was superior to VHS (0.90, 95% CI, 0.94‐1.00, P = .03) in identifying dogs with LA/Ao ≥1.6. A cut‐off value of ≥3.4 vertebrae using M‐VLAS provided 92.7% sensitivity and 93.1% specificity in predicting LA enlargement.

Conclusions and clinical importance

M‐VLAS, which is superior to VHS, offers an accurate and repeatable way to radiographically identify LA enlargement in dogs with MMVD.

Clinical, computed tomographic and histopathological findings in two cats with pulmonary fibrosis of unknown aetiology

Case series summary

Two cats were presented for further investigation of respiratory signs. One cat had a history of a cough and the other, tachypnoea. In each case, thoracic CT was performed, which revealed a generalised marked reticular pattern in the first cat and focal consolidation of the right caudal lung lobe in the second cat. The first cat was euthanased following completion of the imaging study and a post-mortem examination was performed. The second cat underwent surgical excision of the abnormal lung lobe and survived for 4 years after diagnosis. Histopathology performed on lung tissue removed from each cat was consistent with pulmonary fibrosis.

Relevance and novel information

This small case series adds to the existing literature and highlights the heterogeneous clinical course and variable appearance of pulmonary fibrosis on CT of affected cats. These cases provide evidence that pulmonary fibrosis in cats incorporates a wide spectrum of fibrotic lung disease and demonstrates the possibility for prolonged survival following diagnosis where disease is localised and amenable to surgical resection.

Use of cardiac sphericity index and manubrium heart scores to assess radiographic cardiac silhouettes in large-and small-breed dogs with and without cardiac disease

OBJECTIVE

To evaluate the usefulness of manubrium heart scores (MHSs) in distinguishing between dogs with and without cardiac disease.

ANIMALS

184 client-owned dogs with (n = 64) and without (120) cardiac disease.

PROCEDURES

Medical records, including thoracic radiographic images, of dogs that either had echocardiographically confirmed moderate to severe cardiac disease with cardiomegaly (case dogs) or were healthy and without evidence of cardiac disease (control dogs) were retrospectively reviewed. From right lateral thoracic radiographic images, the lengths of the cardiac short and long axes and the manubrium were measured and cardiac sphericity indices (CSIs), MHSs, and vertebral heart scores were determined. Dogs were grouped on the basis of whether they did or did not have cardiac disease (case dogs vs control dogs), breed size (large-breed dogs vs small-breed dogs), and whether cardiac disease affected the right side of the heart (right-sided cardiac disease) or the left side of heart (left-sided cardiac disease). Results were compared across groups.

RESULTS

The mean CSI was greatest in case dogs with right-sided cardiac disease, compared with other groups. In dogs of the present study, a combination of high short-, long-, and overall-MHSs was more suggestive of left-sided cardiac disease, whereas a high short-MHS without high long- and overall-MHSs was more suggestive of right-sided cardiac disease.

CONCLUSIONS AND CLINICAL RELEVANCE

Results indicated that MHSs could be useful, objective values to help assess dogs for potential heart disease, and we recommend that MHSs be added to the diagnostic tools used by veterinarians when screening for heart disease in dogs.

Comparison of fluoroscopic cardiovascular measurements from healthy dogs obtained at end-diastole and end-systole

INTRODUCTION

Quantitative measurements are commonly implemented to objectively evaluate heart size in dogs. However, recent studies suggest that the phase of cardiac cycle can impact vertebral heart score, thereby potentially influencing clinical management. This study used fluoroscopy to assess the impact of the cardiac cycle on quantitative cardiovascular measurements in healthy dogs of various breeds.

ANIMALS, MATERIALS, AND METHODS

This was a prospective study. Multiple cardiac and respiratory cycles were recorded fluoroscopically. Peak inspiratory end-systole and end-diastole frames were captured from 49 dogs in right lateral recumbency. Vertebral heart score (VHS), cardiothoracic ratio (CTR), vertebral left atrial size (VLAS), and caudal vena cava diameter ratio (CVCDR) measurements were performed. Mean cardiac measurements were compared between cardiac cycle phases, and the impact of body condition score (BCS), weight, thoracic conformation, sex, and age was evaluated.

RESULTS

Cardiac cycle had a significant impact on VHS (mean difference: 0.36 ± 0.14 vertebral units between systole and diastole; p < 0.001) and CTR (mean difference: 2.2 ± 1.2% between systole and diastole; p < 0.001). Cardiac cycle had no significant impact on VLAS or CVCDR. Increasing BCS significantly increased variation between systole and diastole in CTR measurements (p = 0.024).

CONCLUSIONS

The cardiac cycle has a significant effect on VHS and CTR but does not impact VLAS or CVCDR. These findings should be taken into consideration during clinical use of these measurements, especially if a patient is being monitored for cardiac changes over time via serial radiographs.

Effect of expiratory phase for radiographic detection of left heart enlargement in dogs with mitral regurgitation

Radiography is a standard diagnostic test for characterizing left heart enlargement in dogs however limited information is available on the effects of respiratory phases. This prospective and retrospective method comparison study investigated the respiratory effect on the size and shape of the left heart in dogs to determine the usefulness of expiratory radiographs to detect enlargements in the left atrium (LA) and left ventricle (LV). Thoracic radiographs taken at full inspiration and expiration were evaluated in 20 normal beagles and 100 dogs diagnosed with mitral regurgitation (MR). Vertebral heart score (VHS), vertebral left atrial size, elevation of the carina, and dorsal bulging of LA on lateral view and lateral bulging of the left auricular appendage and LV on ventrodorsal view were assessed. In normal dogs, there were no significant differences in the evaluative factors between inspiration and expiration. In dogs with MR, VHS did not change according to respiration. However, bulging of the LA, left auricular appendage, and LV had sharp margin during expiration compared with inspiration. The expiratory radiographic finding of LA bulging had a higher correlation with the LA to aorta ratio compared with LA bulging in the inspiratory radiography. Using a LA to aorta echocardiographic ratio greater than 1.5 as the gold standard, the radiographic sensitivity for LA enlargement was higher during expiration than inspiration. These findings of our study indicated that expiratory radiography can be helpful to support the detection of left heart enlargement, although it can overestimate LA enlargement in dogs with MR.

Correlação entre os achados radiográficos e ecocardiográficos sugestivos de aumento cardíaco em cães: 104 casos

RESUMO Radiografias torácicas e ecocardiogramas de 104 caninos foram avaliados e correlacionados quanto ao aumento das câmaras cardíacas. Os achados radiográficos foram correlacionados estatisticamente a fim de se estabelecer a acurácia do exame radiográfico na detecção do aumento cardíaco em comparação ao ecocardiográfico - padrão-ouro não invasivo. A correlação entre os achados radiográficos indicativos de aumento cardíaco e os índices ecocardiográficos mostrou-se fraca, significativa somente para VHS versus relação átrio esquerdo/aorta (r=0,3136), eixo curto versus relação átrio esquerdo/aorta (r=0,3813) e eixo curto versus velocidade da onda E (r=0,2021). A acurácia da radiografia na determinação subjetiva de aumento das câmaras cardíacas foi razoável, variando entre 72,1% e 80,8%. Em contrapartida, o VHS apresentou baixa acurácia (50,9%) na detecção de cardiomegalia.

Radiographic assessment of the cardiac silhouette in clinically normal large- and small-breed dogs

OBJECTIVE To determine manubrium heart scores (MHSs) from measurements of cardiac short-axis length (cSAL) and long-axis length (cLAL) relative to the corresponding manubrium length (ML) on thoracic radiographic views of dogs and assess correlation of MHSs with vertebral heart scores (VHSs). ANIMALS 120 clinically normal large-breed dogs (LBDs) and small-breed dogs (SBDs). PROCEDURES On right lateral views (RLVs) and ventrodorsal views (VDVs) for each dog, cSAL and cLAL were measured and expressed as a ratio; the cSAL:ML ratio (short-MHS), cLAL:ML ratio (long-MHS), and cSAL-and-cLAL:ML ratio (overall-MHS) were also calculated. The VHS was determined from the RLV. Correlation of VHS with MHS was assessed. RESULTS On RLVs and VDVs, mean cSAL:cLAL ratios were 0.77 (SD, 0.05) and 0.72 (SD, 0.05), respectively, in 60 LBDs and 0.81 (SD, 0.06) and 0.78 (SD, 0.06), respectively, in 60 SBDs. In LBDs, mean short-MHS, long-MHS, and overall-MHS were 2.1 (SD, 0.22), 2.7 (SD, 0.24), and 4.8 (SD, 0.5), respectively, on RLVs and 2.3 (SD, 0.26), 3.2 (SD, 0.34), and 5.4 (SD, 0.6), respectively, on VDVs. In SBDs, mean short-MHS, long-MHS, and overall-MHS were 2.4 (SD, 0.39), 2.9 (SD, 0.50), and 5.3 (SD, 0.83), respectively, on RLVs and 2.5 (SD, 0.44), 3.2 (SD, 0.51), and 5.8 (SD, 0.92), respectively, on VDVs. Mean VHSs were 10.73 (SD, 0.52) and 10.27 (SD, 0.81) in LBDs and SBDs, respectively. A significant correlation was identified between VHS and each MHS in LBDs. CONCLUSIONS AND CLINICAL RELEVANCE In the dogs evaluated, radiographic cardiac dimensions and MHSs were correlated. Validity of the MHS for cardiac dimension assessment in other healthy dogs and dogs with cardiac disease warrants investigation.