In small animal practice, patients often present with urinary lithiasis, and prediction of urolith composition is essential to determine the appropriate treatment. Through abdominal radiographs, the composition of mineral radiopaque uroliths can be determined by considering many different factors
this can be complex and, as such, tailor-made for the use of artificial intelligence (AI). The Minnesota Urolith Center partnered with Hill's Pet Nutrition to develop a deep learning AI algorithm (CALCurad) within a smartphone application called the MN Urolith Application that allows for the preliminary assessment of urolith composition. The algorithm provides the probability of a urolith being composed of struvite from an image taken of an abdominal radiograph. This pilot study evaluates the accuracy of the CALCurad in the context of clinical practice. A sample population of 139 dogs was considered, and the results obtained by the CALCurad were compared with the results obtained by infrared spectroscopy analysis. Agreement between the application and quantitative analyses was 81.3%. These results suggest that the CALCurad can effectively be used to predict urolith composition in dogs, helping the clinician to decide between medical and surgical management of the patient. The use of the CALCurad is an example of the usefulness of AI in helping veterinarians make clinical decisions in patient care.