Development, validation, and application of an algorithm for canine adrenal gland volume estimation

Introduction: Using computed tomography (CT) for quantifying adrenal volume could aid identification and management of adrenal disease. This study determined anatomic features associated with CT volume of adrenals of dogs, and validated an algorithm to estimate subject-specific adrenal volume. Application of the algorithm to clinical cases with known adrenal pathology is reported. Material and Methods: Dimensions and volume of the left (VLA) and right (VRA) adrenals, first lumbar vertebrae (L1), and aorta, from CT images of 150 dogs undergoing abdominal scans for reasons other than adrenal disease were used to model adrenal volume. Using cross validation, multiple linear regression established best-fit explanatory subsets and predictive models. Additional data from 29 dogs were used for external validation and agreement testing of estimated and measured volumes. The algorithm was retrospectively applied to 46 dogs that underwent adrenalectomy and histologic examination of excised glands. Linear measurements of caudal pole thickness and lesion diameter were also recorded. Results: Mean normal left and right adrenal volumes were 1.22cm3 (range 0.23 to 2.84cm3) and 1.11cm3 (0.14 to 2.97cm3), respectively, and were highly correlated (r=0.88). Bodyweight (W) was the best single explanatory variable for variance in each adrenal and improved with addition of L1 height (HL1). Algorithms after cross validation and adjustment for bias were: VLA= 0.154 + (0.02 x W) + (0.07 x HL1) (adjR2=0.62); and VRA= -0.116 + (0.02 x W) + (0.06 x HL1) (adjR2=0.64). Application of the algorithm to adrenals with known pathology identified enlargement in all dogs compared to inconsistent assessment using linear measurements. Conclusion: Left and right adrenal volume is associated with subject bodyweight and L1 height, and application of a validated algorithm consistently identified enlargement in adrenals with confirmed pathology