Global FAST for Patient Monitoring and Staging in Dogs and Cats

Deep learning models for interpretation of point of care ultrasound in military working dogs

Introduction

Military working dogs (MWDs) are essential for military operations in a wide range of missions. With this pivotal role, MWDs can become casualties requiring specialized veterinary care that may not always be available far forward on the battlefield. Some injuries such as pneumothorax, hemothorax, or abdominal hemorrhage can be diagnosed using point of care ultrasound (POCUS) such as the Global FAST® exam. This presents a unique opportunity for artificial intelligence (AI) to aid in the interpretation of ultrasound images. In this article, deep learning classification neural networks were developed for POCUS assessment in MWDs.

Methods

Images were collected in five MWDs under general anesthesia or deep sedation for all scan points in the Global FAST® exam. For representative injuries, a cadaver model was used from which positive and negative injury images were captured. A total of 327 ultrasound clips were captured and split across scan points for training three different AI network architectures: MobileNetV2, DarkNet-19, and ShrapML. Gradient class activation mapping (GradCAM) overlays were generated for representative images to better explain AI predictions.

Results

Performance of AI models reached over 82% accuracy for all scan points. The model with the highest performance was trained with the MobileNetV2 network for the cystocolic scan point achieving 99.8% accuracy. Across all trained networks the diaphragmatic hepatorenal scan point had the best overall performance. However, GradCAM overlays showed that the models with highest accuracy, like MobileNetV2, were not always identifying relevant features. Conversely, the GradCAM heatmaps for ShrapML show general agreement with regions most indicative of fluid accumulation.

Discussion

Overall, the AI models developed can automate POCUS predictions in MWDs. Preliminarily, ShrapML had the strongest performance and prediction rate paired with accurately tracking fluid accumulation sites, making it the most suitable option for eventual real-time deployment with ultrasound systems. Further integration of this technology with imaging technologies will expand use of POCUS-based triage of MWDs.

Successful management of pulmonary edema secondary to accidental electrocution in a young dog

BACKGROUND

Human records describe pulmonary edema as a life-threatening complication of electric shock. Successful management requires prompt recognition and intensive care. However, in companion animals, electrocutions are rarely reported, even though domestic environments are full of electrical devices and there is always the possibility of accidental injury. Therefore, it is important for veterinarians to know more about this condition in order to achieve successful patient outcomes.

CASE PRESENTATION

A 3-month-old male Labrador Retriever was presented with a history of transient loss of consciousness after chewing on a household electrical cord. On admission, the puppy showed an orthopneic position with moderate respiratory distress. Supplemental oxygen via nasal catheter was provided, but the patient showed marked worsening of respiratory status. Point-of-care ultrasound exams suggested neurogenic pulmonary edema due to electrical shock close to the central nervous system and increased B-lines without evidence of cardiac abnormalities. Mechanical ventilation of the patient was initiated using volume-controlled mode with a tidal volume of 9 to 15 ml/kg until reaching an end-tidal carbon dioxide ≤ 40 mm Hg, followed by a stepwise lung-recruitment maneuver in pressure-controlled mode with increases of the peak inspiratory pressure (15 to 20 cm H2O) and positive end-expiratory pressure (3 to 10 cm H2O) for 30 min, and return to volume-controlled mode with a tidal volume of 15 ml/kg until reaching a peripheral oxygen saturation ≥ 96%. Weaning from the ventilator was achieved in six hours, and the patient was discharged two days after admission without neurological or respiratory deficits.

CONCLUSIONS

We present a rather unusual case of a neurogenic pulmonary edema subsequent to accidental electrocution in a dog. Timely diagnosis by ultrasound and mechanical ventilation settings are described. Our case highlights that pulmonary edema should be considered a potentially life-threatening complication of electrical shock in small animal emergency and critical care medicine.

Performance of Different Echocardiographic Measurements of Left Atrial Size in Dogs by Observers with Different Levels of Experience

Simple Summary

Assessing enlargement of the left atrium (one of the four cardiac chambers) is extremely important for gaining information about dogs’ heart disease, their prognosis, and directing treatment. However, people with different levels of experience may be required to make this assessment, and we don’t know how observers with different experiences perform in making this assessment. In this study, five observers with different levels of experience evaluated the left atrium of 36 dogs in a blinded fashion (i.e., unaware of each other measurements, or of the identity of the dog), compared to two cardiologists. We then used statistical analysis to evaluate repeatability, reproducibility, accuracy of the measurements, and the capacity of correctly identifying left atrial enlargement. We found that the measurements performed by observers with more experience where more similar to the cardiologists’ measurements, and that combining more than one technique for measuring the left atrium can improve accuracy of the identification of left atrial enlargement.

Abstract

Assessment of left atrial (LA) sizes in dogs informs clinical staging, risk assessment, treatment decisions, and prognosis. The objective of this study was to assess the diagnostic performance of observers with different levels of experience measuring the LA with three different techniques. Echocardiographic images from 36 dogs with different degrees of left atrial enlargement (LAE) were retrospectively retrieved, anonymized and measured in a blinded fashion by a veterinary student, a first-year cardiology resident, a third-year cardiology resident, and two board-certified veterinary cardiologists. The LA to aortic root ratio (LA:Ao), LA antero-postero diameter indexed to body weight (LAiAPD) and left atrial area were measured. Inter- and intra-observer intraclass correlation coefficients (ICCs) were calculated for all three variables. Bland–Altman plots and accuracy in identification of LAE were calculated for the three least experienced observers using LA:Ao and LAiAPD. Intra- and interobserver ICCs were greater than 0.9 for every variable. The observer with least experience had significant positive bias and a tendency to overestimate larger measurements using LA:Ao, but not using LAiAPD. The accuracy of identification of LAE also increased with the increasing level of experience and was higher for LAiAPD compared to LA:Ao. Combining both methods for identification of LAE, further increased accuracy.