A survey of testicular texture in canine ultrasound images.

Introduction: Computer-based texture analysis provides objective data that can be extracted from medical images, including ultrasound images. One popular methodology involves the generation of a gray-level co-occurrence matrix (GLCM) from the image, and from that matrix, texture fractures can be extracted. Methods: We performed texture analysis on 280 ultrasound testicular images obtained from 70 dogs and explored the resulting texture data, by means of principal component analysis (PCA). Results: Various abnormal lesions were identified subjectively in 35 of the 280 cropped images. In 16 images, pinpoint-to-small, well-defined, hyperechoic foci were identified without acoustic shadowing. These latter images were classified as having "microliths." The remaining 19 images with other lesions and areas of non-homogeneous testicular parenchyma were classified as "other." In the PCA scores plot, most of the images with lesions were clustered. These clustered images represented by those scores had higher values for the texture features entropy, dissimilarity, and contrast, and lower values for the angular second moment and energy in the first principal component. Other data relating to the dogs, including age and history of treatment for prostatomegaly or chemical castration, did not show clustering on the PCA. Discussion: This study illustrates that objective texture analysis in testicular ultrasound correlates to some of the visual features used in subjective interpretation and provides quantitative data for parameters that are highly subjective by human observer analysis. The study demonstrated a potential for texture analysis in prediction models in dogs with testicular abnormalities.

Collimation and Exposure Parameter Influence Image Quality and Potential Radiation Dose to the Eye Lens of Personnel in Computed Radiography of the Canine Pelvis.

Introduction: The purpose of this study was to evaluate the effect of collimation on image quality and radiation dose to the eye lenses of the personnel involved in computed radiography of the canine pelvis. Materials and Methods: A retrospective study of canine pelvic radiographs (N = 54) was undertaken to evaluate the relationship between image quality and the degree of field the collimation used. This was followed by a prospective cadaver study (N = 18) that assessed the effects on image quality and on scattered radiation dose of different collimation field areas and exposure parameters. All radiographs were analyzed for image quality using a Visual Grading Analysis (VGA) with three observers. Finally, the potential scattered radiation dose to the eye lens of personnel restraining a dog for pelvic radiographs was measured. Results: The retrospective study showed a slightly better (statistically non-significant) VGA score for the radiographs with optimal collimation. Spatial and contrast resolution and image sharpness showed the greatest improvement in response to minimizing the collimation field. The prospective study showed slightly better VGA scores (improved image quality) with the optimal collimation. Increasing the exposure factors especially the tube current and exposure time (mAs) resulted in improved low contrast resolution and less noise in the radiographs. The potential eye lens radiation dose increased by 14, 28, and 40% [default exposures, increased the tube peak potential (kVp), increased mAs, respectively] as a result of reduced collimation (increased beam size). Conclusion: The degree of collimation has no statistically significant on image quality in canine pelvic radiology for the range of collimation used but does have an impact on potential radiation dose to personnel in the x-ray room. With regard to radiation safety, increases in kVp are associated with less potential scatter radiation exposure compared to comparable increases in mAs.

Deep transfer learning can be used for the detection of hip joints in pelvis radiographs and the classification of their hip dysplasia status.

Reports of machine learning implementations in veterinary imaging are infrequent but changes in machine learning architecture and access to increased computing power will likely prompt increased interest. This diagnostic accuracy study describes a particular form of machine learning, a deep learning convolution neural network (ConvNet) for hip joint detection and classification of hip dysplasia from ventro-dorsal (VD) pelvis radiographs submitted for hip dysplasia screening. 11,759 pelvis images were available together with their Fédération Cynologique Internationale (FCI) scores. The dataset was dicotomized into images showing no signs of hip dysplasia (FCI grades "A" and "B", the "A-B" group) and hips showing signs of dysplasia (FCI grades "C", "D," and "E", the "C-E" group). In a transfer learning approach, an existing pretrained ConvNet was fine-tuned to provide models to recognize hip joints in VD pelvis images and to classify them according to their FCI score grouping. The results yielded two models. The first was successful in detecting hip joints in the VD pelvis images (intersection over union of 85%). The second yielded a sensitivity of 0.53, a specificity of 0.92, a positive predictive value of 0.91, and a negative predictive value of 0.81 for the classification of detected hip joints as being in the "C-E" group. ConvNets and transfer learning are applicable to veterinary imaging. The models obtained have potential to be a tool to aid in hip screening protocols if hip dysplasia classification performance was improved through access to more data and possibly by model optimization.

Fingerprinting of Doppler audio signals from the common carotid artery.

Audio fingerprinting involves extraction of quantitative frequency descriptors that can be used for indexing, search and retrieval of audio signals in sound recognition software. We propose a similar approach with medical ultrasonographic Doppler audio signals. Power Doppler periodograms were generated from 84 ultrasonographic Doppler signals from the common carotid arteries in 22 dogs. Frequency features were extracted from each periodogram and included in a principal component analysis (PCA). From this 10 audio samples were pairwise classified as being either similar or dissimilar. These pairings were compared to a similar classification based on standard quantitative parameters used in medical ultrasound and to classification performed by a panel of listeners. The ranking of sound files according to degree of similarity differed between the frequency and conventional classification methods. The panel of listeners had an 88% agreement with the classification based on quantitative frequency features. These findings were significantly different from the score expected by chance (p < 0.001). The results indicate that the proposed frequency based classification has a perceptual relevance for human listeners and that the method is feasible. Audio fingerprinting of medical Doppler signals is potentially useful for indexing and search for similar and dissimilar audio samples in a dataset.

Pulmonary vasculature in dogs assessed by three-dimensional fractal analysis and chemometrics.

Fractal analysis of canine pulmonary vessels could allow quantification of their space-filling properties. Aims of this prospective, analytical, cross-sectional study were to describe methods for reconstructing three dimensional pulmonary arterial vascular trees from computed tomographic pulmonary angiogram, applying fractal analyses of these vascular trees in dogs with and without diseases that are known to predispose to thromboembolism, and testing the hypothesis that diseased dogs would have a different fractal dimension than healthy dogs. A total of 34 dogs were sampled. Based on computed tomographic pulmonary angiograms findings, dogs were divided in three groups: diseased with pulmonary thromboembolism (n = 7), diseased but without pulmonary thromboembolism (n = 21), and healthy (n = 6). An observer who was aware of group status created three-dimensional pulmonary artery vascular trees for each dog using a semiautomated segmentation technique. Vascular three-dimensional reconstructions were then evaluated using fractal analysis. Fractal dimensions were analyzed, by group, using analysis of variance and principal component analysis. Fractal dimensions were significantly different among the three groups taken together (P = 0.001), but not between the diseased dogs alone (P = 0.203). The principal component analysis showed a tendency of separation between healthy control and diseased groups, but not between groups of dogs with and without pulmonary thromboembolism. Findings indicated that computed tomographic pulmonary angiogram images can be used to reconstruct three-dimensional pulmonary arterial vascular trees in dogs and that fractal analysis of these three-dimensional vascular trees is a feasible method for quantifying the spatial relationships of pulmonary arteries. These methods could be applied in further research studies on pulmonary and vascular diseases in dogs.

Ultrasonographic predictors of response of European eels (Anguilla anguilla) to hormonal treatment for induction of ovarian development.

OBJECTIVE To examine ultrasonographic predictors of ovarian development in European eels (Anguilla anguilla) undergoing hormonal treatment for assisted reproduction. ANIMALS 83 female European eels. PROCEDURES Eels received weekly IM injections of salmon pituitary extract (first injection = week 1). Ultrasonography of the ovaries was performed twice during hormonal treatment (weeks 7 and 11). Eels were identified on the basis of body weight as having an adequate response by weeks 14 to 20 or an inadequate response after injections for 21 weeks. Eels were euthanized at the end of the experiment and classified by use of ovarian histologic examination. Ovarian cross-sectional area and size of eel (ie, length (3) ) were used to classify eels (fast responder, slow responder, or nonresponder) and to calculate an ultrasonographic-derived gonadosomatic index. Gray-level co-occurrence matrices were calculated from ovarian images, and 22 texture features were calculated from these matrices. RESULTS The ultrasonographic-derived gonadosomatic index differed significantly between fast responders and slow responders or nonresponders at both weeks 7 and 11. Principal component analysis revealed a pattern of separation between the groups, and partial least squares discriminant analysis revealed signals in the ovarian texture that discriminated females that responded to treatment from those that did not. CONCLUSIONS AND CLINICAL RELEVANCE Ovarian texture information in addition to morphometric variables can enhance ultrasonographic applications for assisted reproduction of eels and potentially other fish species. This was a novel, nonlethal method for classifying reproductive response of eels and the first objective texture analysis performed on ultrasonographic images of the gonads of fish.