Evaluating artificial intelligence algorithms for use in veterinary radiology

Abstract

Artificial intelligence is increasingly being used for applications in veterinary radiology, including detection of abnormalities and automated measurements. Unlike human radiology, there is no formal regulation or validation of AI algorithms for veterinary medicine and both general practitioner and specialist veterinarians must rely on their own judgment when deciding whether or not to incorporate AI algorithms to aid their clinical decision-making. The benefits and challenges to developing clinically useful and diagnostically accurate AI algorithms are discussed. Considerations for the development of AI research projects are also addressed. A framework is suggested to help veterinarians, in both research and clinical practice contexts, assess AI algorithms for veterinary radiology.

Performance of a commercially available artificial intelligence software for the detection of confirmed pulmonary nodules and masses in canine thoracic radiography

“…Further limitation includes that all cases tested were from a single institution, which may cause bias in the type or quality of images submitted for evaluation 17,22 . These limitations highlight the importance of establishing open‐access multi‐institutional data sources so that later investigations can be even more robust 17 …”

“…Performance in discriminating a pulmonary nodular pattern was assessed by the diagnostic imaging resident (L.P.), in consultation with a statistician (E.K.K. ), using commercially available software (Microsoft Excel, Version 2307, Microsoft Corporation, 2018) via sensitivity, specificity, predictive values, accuracy, balanced accuracy, and F1 score 22 . Factors such as age, weight, and breed were not included in the statistical analysis given that the software was purportedly trained to recognize and interpret varied canine breeds of different ages.…”

“…), using commercially available software (Microsoft Excel, Version 2307, Microsoft Corporation, 2018) via sensitivity, specificity, predictive values, accuracy, balanced accuracy, and F1 score. 22 Factors such as age, weight, and breed were not included in the statistical analysis given that the software was purportedly trained to recognize and interpret varied canine breeds of different ages. The nodule/mass-positive cases that were correctly classified as having the lesions and incorrectly classified as not having lesions were then compared post hoc on multiple factors including location, conspicuity, and size of the lesion.…”

“…Further limitation includes that all cases tested were from a single institution, which may cause bias in the type or quality of images submitted for evaluation. 17,22 These limitations highlight the impor-…”

Computer vision model for the detection of canine pododermatitis and neoplasia of the paw

“…Artificial intelligence in veterinary medicine is an emerging field that has mostly been applied to large animals, clinical and anatomical pathology and radiology. [25][26][27][28][29][30][31][32][33][34][35][36] The use of AI for skin disease diagnosis in dogs is rare. [37][38][39] Previous studies used image classification to label entire images, while the current approach uses object detection to not only locate, but also label individual objects within an image.…”