Deep learning-based diagnosis of stifle joint diseases in dogs

In this retrospective, analytical study, we developed a deep learning-based diagnostic model that can be applied to canine stifle joint diseases and compared its accuracy with that achieved by veterinarians to verify its potential as a reliable diagnostic method. A total of 2382 radiographs of the canine stifle joint from cooperative animal hospitals were included in a dataset. Stifle joint regions were extracted from the original images using the faster region-based convolutional neural network (R-CNN) model, and the object detection accuracy was evaluated. Four radiographic findings: patellar deviation, drawer sign, osteophyte formation, and joint effusion, were observed in the stifle joint and used to train a residual network (ResNet) classification model. Implant and growth plate groups were analyzed to compare the classification accuracy against the total dataset. All deep learning-based classification models achieved target accuracies exceeding 80%, which is comparable to or slightly less than those achieved by veterinarians. However, in the case of drawer signs, further research is necessary to improve the low sensitivity of the model. When the implant group was excluded, the classification accuracy significantly improved, indicating that the implant acted as a distraction. These results indicate that deep learning-based diagnoses can be expected to become useful diagnostic models in veterinary medicine.     

Comparative evaluation of six different body regions of the dog using analog and digital radiography

In this study the quality of digital and analog radiography in dogs was compared. For this purpose, three conventional radiographs (varying in exposure) and three digital radiographs (varying in MUSI-contrast [MUSI = MUlti Scale Image Contrast], the main post-processing parameter) of six different body regions of the dog were evaluated (thorax, abdomen, skull, femur, hip joints, elbow). The quality of the radiographs was evaluated by eight veterinary specialists familiar with radiographic images using a questionnaire based on details of each body region significant in obtaining a radiographic diagnosis. In the first part of the study the overall quality of the radiographs was evaluated. Within one region, 89.5% (43/48) chose a digital radiograph as the best image. Divided into analog and digital groups, the digital image with the highest MUSI-contrast was most often considered the best, while the analog image considered the best varied between the one with the medium and the one with the longest exposure time. In the second part of the study, each image was rated for the visibility of specific, diagnostically important details. After summarisation of the scores for each criterion, divided into analog and digital imaging, the digital images were rated considerably superior to conventional images. The results of image comparison revealed that digital radiographs showed better image detail than radiographs taken with the analog technique in all six areas of the body.     

A computer-derived protocol using recursive partitioning to aid in estimating prognosis of horses with abdominal pain in referral hospitals

In order to determine which variables are useful and accurate in estimating prognosis in horses with abdominal pain, data were analyzed from 231 horses presented at a veterinary teaching hospital. Using multiple stepwise discriminant analysis in a recursive partition model, we obtained a decision protocol that identified survivors and nonsurvivors. The prevalence of survivors was 61% in this population. The sensitivity, specificity, and positive and negative predictive values of this model were 71, 83, 87 and 65%, respectively. This decision protocol was validated by Jackknife classification and also by evaluation with a referral population of 100 horses in which the prevalence of survivors was 83%. This led to sensitivity, specificity, and positive and negative predictive values of 83, 78, 94 and 50%, respectively.     

Modelling broilers’ abdominal fat in response to dietary treatments

Neural networks are capable of modelling any complex function and can be used in poultry production. Dietary crude fibre (CF) and exogenous enzymes (exEn) extensively affected abdominal fat (AF) of broilers. Current methods to study AF and its correlation with dietary CF levels and exEn supplements are costly, laborious and time‐consuming. The purpose of this study was to develop an artificial neural network–genetic algorithm (ANN‐GA) to model data on the response of broiler chickens (AF) to CF and exEn from 0 to 42 days of age. A data set containing eight treatments was divided to the train, validation, and test data set of the ANN models. The information about feeding eight diets at two periods [starter (0–21 days of age) and grower (22–42 days of age)] were used to estimate AF of broilers by ANN‐GA. A multilayer feed‐forward neural network with different structures was developed using matlab software, and optimal values for the ANN weights were obtained using the genetic algorithm (GA). Crude fibre, and exEn were used as input variables and AF of broilers was output variable. The best model of ANN‐GA was determined based on the train root mean square error (RMSE). The best selected ANN‐GA showed desirable results, RMSE, 0.1286% and R² coefficient, 0.876 for test data.     

APPROACHES TO INSPECTING COMPUTED TOMOGRAPHIC AND MAGNETIC RESONANCE STUDIES

There is a need to better understand how to optimally inspect large image datasets. The aim of the present study was to complement experimental studies of visual perception by using an online questionnaire to collect opinions of practicing veterinary radiologists about the approaches they use when inspecting clinical computed X‐ray tomography (CT) and/or magnetic resonance (MR) studies, and to test associations between radiologist's approaches and their training, experience, or caseload. Questionnaires were received from 90/454 (20%) American College of Veterinary Radiology (ACVR) Diplomates and 58/156 (37%) European College of Veterinary Diagnostic Imaging (ECVDI) Diplomates, providing 139 complete responses for CT studies and 116 for MR. Questionnaire responses differed for the following variables: specialty college, years since Board Certification, CT and MR caseload, and type of practice. ACVR Diplomates more frequently inspected multiple anatomic structures in CT and MR images before moving on to the next image, and ECVDI Diplomates more frequently inspected a specific anatomic structure through a series, then went back and checked another structure. A significant number of radiologists indicated that they initially ignore the history, adopt relatively rigid search patterns with emphasis on viewing images in a predetermined order with minimal deviation, and arrange series of images to facilitate comparisons between images, such as pre‐ and postcontrast images. Radiologists tended to adopt similar approaches for both CT and MR studies. Findings from this study could be used as foci for teaching novices how to approach large imaging studies, and provide guidance for case‐based assessment of trainees.     

BIBI: Bayesian inference of breed composition

The aim of this paper was to develop statistical models to estimate individual breed composition based on the previously proposed idea of regressing discrete random variables corresponding to counts of reference alleles of biallelic molecular markers located across the genome on the allele frequencies of each marker in the pure (base) breeds. Some of the existing regression‐based methods do not guarantee that estimators of breed composition will lie in the appropriate parameter space, and none of them account for uncertainty about allele frequencies in the pure breeds, that is, uncertainty about the design matrix. To overcome these limitations, we proposed two Bayesian generalized linear models. For each individual, both models assume that the counts of the reference allele at each marker locus follow independent Binomial distributions, use the logit link and pose a Dirichlet prior over the vector of regression coefficients (which corresponds to breed composition). This prior guarantees that point estimators of breed composition such as the posterior mean pertain to the appropriate space. The difference between these models is that model termed BIBI does not account for uncertainty about the design matrix, while model termed BIBI2 accounts for such an uncertainty by assigning independent Beta priors to the entries of this matrix. We implemented these models in a data set from the University of Florida's multibreed Angus‐Brahman population. Posterior means were used as point estimators of breed composition. In addition, the ordinary least squares estimator proposed by Kuehn et al. ( 2011 ) (OLSK) was also computed. BIBI and BIBI2 estimated breed composition more accurately than OLSK, and BIBI2 had a 7.69% improvement in accuracy as compared to BIBI.     

The Acute Patient Physiologic and Laboratory Evaluation (APPLE) Score: A Severity of Illness Stratification System for Hospitalized Dogs

Background: Objective risk stratification models are used routinely in human critical care medicine. Applications include quantitative and objective delineation of illness severity for patients enrolled in clinical research, performance benchmarking, and protocol development for triage and therapeutic management. Objective: To develop an accurate, validated, and user‐friendly model to stratify illness severity by mortality risk in hospitalized dogs. Animals: Eight hundred and ten consecutive intensive care unit (ICU) admissions of dogs at a veterinary teaching hospital. Methods: Prospective census cohort study. Data on 55 management, physiological, and biochemical variables were collected within 24 hours of admission. Data were randomly divided, with 598 patient records used for logistic regression model construction and 212 for model validation. Results: Patient mortality was 18.4%. Ten‐variable and 5‐variable models were developed to provide both a high‐performance model and model maximizing accessibility, while maintaining good performance. The 10‐variable model contained creatinine, WBC count, albumin, SpO₂, total bilirubin, mentation score, respiratory rate, age, lactate, and presence of free fluid in a body cavity. Area under the receiver operator characteristic (AUROC) on the construction data set was 0.93, and on the validation data set was 0.91. The 5‐variable model contained glucose, albumin, mentation score, platelet count, and lactate. AUROC on the construction data set was 0.87, and on the validation data set was 0.85. Conclusions and Clinical Importance: Two models are presented that enable allocation of an accurate and user‐friendly illness severity index for dogs admitted to an ICU. These models operate independent of primary diagnosis, and have been independently validated.     

AN OPTIMIZED PROTOCOL FOR MULTISLICE COMPUTED TOMOGRAPHY OF THE CANINE BRAIN

Computed tomography (CT) is commonly used in veterinary practice to evaluate dogs with suspected brain disease, however contrast resolution limitations and artifacts may reduce visualization of clinically important anatomic features. The purpose of this study was to develop an optimized CT protocol for evaluating the canine brain. The head of a 5‐year‐old Springer Spaniel with no neurological signs was imaged immediately following euthanasia using a 4‐slice CT scanner and 282 protocols. Each protocol used a fixed tube voltage of 120 kVp and 10 cm display field of view. Other acquisition and reconstruction parameters were varied. For each protocol, four selected images of the brain were reconstructed, anonymized and saved in DICOM format. Three board‐certified veterinary radiologists independently reviewed each of the four images for each protocol and recorded a numerical quality score for each image. The protocol yielding the lowest total numerical score was defined as the optimal protocol. There was overall agreement that the optimal protocol was the one with the following parameters: sequential mode, 300 mAs, 1 mm slice thickness, 1 s tube rotation time, medium image reconstruction algorithm and applied beam hardening correction. Sequential imaging provided optimal image resolution. The thin‐sliced images provided a small blur due to partial volume artifacts. A high tube current resulted in a relatively low noise level. Use of a medium frequency image reconstruction algorithm provided optimal contrast resolution for brain tissue. Use of a proprietary beam hardening correction filter (Posterior Fossa Optimization) markedly reduced beam‐hardening artifact.     

If you build it,they will learn: A review of models in veterinary surgical education

Surgical skills are learned through deliberate practice, and veterinary educators are increasingly turning to models for teaching and assessing surgical skills. This review article sought to compile and review the literature specific to veterinary surgical skills models, and to discuss the themes of fidelity, educational outcomes, and validity evidence. Several literature searches using broad terms such as “veterinary surgery model,” “veterinary surgical model,” and “veterinary surgical simulator” were performed using PubMed, CAB abstracts, and Google scholar. All articles describing a model created and utilized for veterinary surgical training were included. Other review articles were used as a source for additional models. Commercially available models were found using review articles, internet browser searches, and communication with veterinary clinical skills educators. There has been an explosion of growth in the variety of small animal surgical task trainers published in the last several decades. These models teach orthopedic surgery, ligation and suturing, open celiotomy and abdominal surgery, sterilization surgeries, and minimally invasive surgeries. Some models were published with accompanying rubrics for learner assessment; these rubrics have been noted where present. Research in veterinary surgical models is expanding and becoming an area of focus for academic institutions. However, there is room for growth in the collection of validity evidence and in development of models for teaching large animal surgery, training surgical residents, and providing continuing education to practitioners. This review can assist with evaluation of current surgical models and trends, and provide a platform for additional studies and development of best practices.     

Use of infrared spectroscopy for diagnosis of traumatic arthritis in horses

OBJECTIVE: To evaluate use of infrared spectroscopy for diagnosis of traumatic arthritis in horses. ANIMALS: 48 horses with traumatic arthritis and 5 clinically and radiographically normal horses. PROCEDURES: Synovial fluid samples were collected from 77 joints in 48 horses with traumatic arthritis. Paired samples (affected and control joints) from 29 horses and independent samples from an affected (n = 12) or control (7) joint from 19 horses were collected for model calibration. A second set of 20 normal validation samples was collected from 5 clinically and radiographically normal horses. Fourier transform infrared spectra of synovial fluids were acquired and manipulated, and data from affected joints were compared with controls to identify spectroscopic features that differed significantly between groups. A classification model that used linear discriminant analysis was developed. Performance of the model was determined by use of the 2 validation datasets. RESULTS: A classification model based on 3 infrared regions classified spectra from the calibration dataset with overall accuracy of 97% (sensitivity, 93%; specificity, 100%). The model, with cost-adjusted prior probabilities of 0.60:0.40, yielded overall accuracy of 89% (sensitivity, 83%; specificity, 100%) for the first validation sample dataset and 100% correct classification of the second set of independent normal control joints. CONCLUSIONS AND CLINICAL RELEVANCE: The infrared spectroscopic patterns of fluid from joints with traumatic arthritis differed significantly from the corresponding patterns for controls. These alterations in absorption patterns may be used via an appropriate classification algorithm to differentiate the spectra of affected joints from those of controls.     

Comparison of carcass composition among Japanese Black, Holstein and their crossbred steers fattening on farm

Carcass composition among Japanese Black, Holstein and their crossbreed (F1), by Holstein mated with Japanese Black sires, was compared by using carcass components of fattening steers collected at eight carcass processing factories. The least squares means of 14 retail cuts, total weights of retail cuts, trimmed fat and bones, and six carcass traits measured at 6th?7th rib section were used for the comparison and examined the expression of heterosis. The comparison was performed at the same level of left‐side carcass weight and under 10 mm fat cover on each retail cut. Japanese Black had the heaviest total weight of retail cuts, followed by F1 and then Holstein. By contrast, non‐edible parts, trimmed fat and bones, were heavier in the order of Holstein, F1 and Japanese Black. Of the retail cuts, so called higher‐priced cuts, such as Shoulder‐Chuck‐Roll, Tender‐Loin, Rib‐Chuck‐Roll, Strip‐Loin and Sirloin‐Butt surrounding the spinal column, were heavier in Japanese Black than Holstein. Although F1 showed intermediate values for almost all traits, Tender‐Loin, Rib‐Chuck‐Roll and Strip‐Loin of F1 were similar to those of Japanese Black rather than the mid‐parent averages. It reflects that at least the average parent heterosis of 2.3–4.3% might be expressed in these higher‐priced cuts. The classification of breeds was carried out by the canonical discriminant analysis using 20 carcass measurements. By the canonical variates, breeds can be effectively discriminated with misdiscrimination rate of 11.6%. When the comparison is made at the carcass weight constant, Japanese Black has characteristics for producing superior carcass quality as well as meat quality.     

Duty hours restriction for our surgical trainees: An ethical obligation or a bad idea?

To ensure patient safety and protect the well‐being of interns and residents, the Accreditation Council for Graduate Medical Education (ACGME) issued guidelines in 2003 limiting the working hours of physician trainees. Although many supported the goals of the ACGME, institutions struggled to restructure their programs and hire staff required by this unfunded mandate. Numerous studies have analyzed the effects of duty hours restrictions on patient outcomes and physician training over the past 15 years. Most agree that duty hours restrictions improved well‐being of house officers, but these improvements came at the expense of continuity, and patient hand‐offs led to medical errors. Effects on resident training are program specific, with duty hours restrictions having the most deleterious effects on surgical disciplines. Because veterinary specialists assume a similar role in providing 24‐hour patient care, interns and residents face work‐related stress as a result of extended working hours, on‐call duty, and an increasingly complex caseload. The North Carolina State Veterinary Hospital is staffed by approximately 100 house officers representing almost every veterinary specialty group. We surveyed departing house officers regarding their quality of life and training experience. Sixty‐six percent of interns and residents reported that they do not have time to take care of personal needs, and 57%‐62% felt neutral or dissatisfied with their mental and physical well‐being. Most trainees believed that decreased duty hours would improve learning, but 42% believed that decreased caseload would be detrimental to training. Veterinary educators must consider post‐DVM veterinary training guidelines that maintain patient care with a good learning environment for interns and residents.     

Focused cardiac ultrasound examination in the emergency and critical care horse: Training for non‐specialist veterinarians and evaluation of proficiency

BackgroundFocused cardiac ultrasound examination (FoCUS) is rapidly emerging for point‐of‐care cardiac assessment using hand‐carried ultrasound (HCU) devices. A specific FoCUS protocol for horses and adequate training guidelines currently are not available.HypothesisTo gain knowledge about the training necessary to become proficient in performing FoCUS using a HCU device.AnimalsThree healthy Warmblood horses were used for practical training of veterinarians and veterinary students in equine focused cardiac ultrasound (eFoCUS).MethodsProspective educational study. An eFoCUS protocol and 1‐day training course were developed. Pre‐ and post‐course written tests were administered to participants to evaluate proficiency in knowledge of echocardiography and echocardiographic pathology. A post‐course practical examination involved performing eFoCUS and storage of representative images. Images were evaluated using an image quality score and compared between participants with some practical experience and participants with no practical experience.ResultsParticipants'' knowledge of echocardiography increased significantly. Recognition of echocardiographic pathology pre‐course ranged from 40% to 90% (mean score, 65.7%) and post‐course from 85% to 100% (mean score, 92%). Eighteen of 21 participants were proficient in performing eFoCUS with a median image quality score of 79% (range, 42%‐95%). Image quality did not differ between participants with some practical experience compared to participants with no experience.Conclusions and Clinical ImportanceVeterinarians and veterinary students independent of previous experience can become proficient in performing eFoCUS after completion of a 1‐day training course.     

INVITED REVIEW—IMAGE REGISTRATION IN VETERINARY RADIATION ONCOLOGY: INDICATIONS,IMPLICATIONS, AND FUTURE ADVANCES

The field of veterinary radiation therapy (RT) has gained substantial momentum in recent decades with significant advances in conformal treatment planning, image‐guided radiation therapy (IGRT), and intensity‐modulated (IMRT) techniques. At the root of these advancements lie improvements in tumor imaging, image alignment (registration), target volume delineation, and identification of critical structures. Image registration has been widely used to combine information from multimodality images such as computerized tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET) to improve the accuracy of radiation delivery and reliably identify tumor‐bearing areas. Many different techniques have been applied in image registration. This review provides an overview of medical image registration in RT and its applications in veterinary oncology. A summary of the most commonly used approaches in human and veterinary medicine is presented along with their current use in IGRT and adaptive radiation therapy (ART). It is important to realize that registration does not guarantee that target volumes, such as the gross tumor volume (GTV), are correctly identified on the image being registered, as limitations unique to registration algorithms exist. Research involving novel registration frameworks for automatic segmentation of tumor volumes is ongoing and comparative oncology programs offer a unique opportunity to test the efficacy of proposed algorithms.     

Transitioning to digital radiography

Objective – To describe the different forms of digital radiography (DR), image file formats, supporting equipment and services required for DR, storage of digital images, and teleradiology. Background – Purchasing a DR system is a major investment for a veterinary practice. Types of DR systems include computed radiography, charge coupled devices, and direct or indirect DR. Comparison of workflow for analog and DR is presented. Summary – On the surface, switching to DR involves the purchase of DR acquisition hardware. The X‐ray machine, table and grids used in analog radiography are the same for DR. Realistically, a considerable infrastructure supports the image acquisition hardware. This infrastructure includes monitors, computer workstations, a robust computer network and internet connection, a plan for storage and back up of images, and service contracts. Advantages of DR compared with analog radiography include improved image quality (when used properly), ease of use (more forgiving to the errors of radiographic technique), speed of making a complete study (important for critically ill patients), fewer repeat radiographs, less time looking for imaging studies, less physical storage space, and the ability to easily send images for consultation. Conclusions – With an understanding of the infrastructure requirements, capabilities and limitations of DR, an informed veterinary practice should be better able to make a sound decision about transitioning to DR.     

Survey of the prevalence and methodology of quality assurance for B‐mode ultrasound image quality among veterinary sonographers

Image quality in B‐mode ultrasound is important as it reflects the diagnostic accuracy and diagnostic information provided during clinical scanning. Quality assurance programs for B‐mode ultrasound systems/components are comprised of initial quality acceptance testing and subsequent regularly scheduled quality control testing. The importance of quality assurance programs for B‐mode ultrasound image quality using ultrasound phantoms is well documented in the human medical and medical physics literature. The purpose of this prospective, cross‐sectional, survey study was to determine the prevalence and methodology of quality acceptance testing and quality control testing of image quality for ultrasound system/components among veterinary sonographers. An online electronic survey was sent to 1497 members of veterinary imaging organizations: the American College of Veterinary Radiology, the Veterinary Ultrasound Society, and the European Association of Veterinary Diagnostic Imaging, and a total of 167 responses were received. The results showed that the percentages of veterinary sonographers performing quality acceptance testing and quality control testing are 42% (64/151; 95% confidence interval 34–52%) and 26% (40/156: 95% confidence interval 19–33%) respectively. Of the respondents who claimed to have quality acceptance testing or quality control testing of image quality in place for their ultrasound system/components, 0% have performed quality acceptance testing or quality control testing correctly (quality acceptance testing 95% confidence interval: 0–6%, quality control testing 95% confidence interval: 0–11%). Further education and guidelines are recommended for veterinary sonographers in the area of quality acceptance testing and quality control testing for B‐mode ultrasound equipment/components.     

The relationship between the size of caudolateral curvilinear osteophyte of the canine femoral neck and the radiographic view

Caudolateral curvilinear osteophyte (CCO), an osteophyte at the site of joint capsule attachment on the caudal aspect of the femoral neck, has been advocated as a radiographic criterion for coxofemoral subluxation. The correlation between the presence of CCO on radiographs (radiographic-CCO), the size of the CCO (CCO index) on three-dimensional computed tomographic (CT) images, and hip evaluation using transverse CT images was assessed in 22 Border Collies. CCOs were detected on the radiographs and CT images of 32% and 100% femurs, respectively. The CCO index correlated significantly with radiographic-CCO, but a large CCO index did not necessarily imply that the CCO was visible on radiographs. Hence, radiographic-CCO findings should be used cautiously in hip evaluation of Border Collies.     

Teaching veterinary students using shelter animals

Veterinary teaching hospitals (VTHs) are experiencing case-load trends that have negatively affected efforts to prepare students for entry-level veterinary practice, particularly in the area of technical skills training. This article examines the clinical training available to veterinary students through a variety of collaborative shelter models. Benefits and potential problems related to initiating a collaborative shelter clinical training program are reviewed. Collaborative efforts between animal shelters and veterinary schools can provide crucial opportunities for outreach teaching initiatives, particularly for teaching medical and surgical skills.