CT Decision Trees Boost Prediction of Adrenal Tumor Adhesions in Dogs and Cats

Computed Tomography Accuracy for Prediction of Adrenal Tumor Adhesion With Organs and Vessels in Dogs and Cats

Manuela Baldinetti, Chiara Mattei, Pamela Di Donato, Marion Grapperon-Mathis, Melania Moioli, Swan Specchi, Ignazio Drudi, Stefano Nicoli, Filippo Cinti, Federica Foschi, Pascaline Pey

Background

Adrenal tumors in dogs and cats frequently pose surgical challenges due to potential adhesion to adjacent organs and vessels, which increases operative difficulty and risk. While CT is a key preoperative imaging modality, standardized criteria for predicting adrenal tumor adhesion (ATA) have not been described in veterinary medicine. Existing CT literature focuses largely on vascular invasion rather than adhesion, and the lack of validated imaging markers complicates surgical planning. The study aimed to define repeatable CT criteria for organ and vascular adhesions and evaluate their predictive value using a probabilistic classification tree, with surgical findings serving as the reference standard.

Methods

This prospective multicenter study enrolled 30 dogs and 2 cats with adrenal tumors undergoing abdominal CT (pre- and post-contrast) followed by adrenalectomy within two weeks. Five blinded radiologists applied consensus-defined CT adhesion criteria (eight organ criteria, five vessel criteria) after training on sample cases. Surgeons recorded intraoperative findings regarding contact, adhesion, invasion, or absence thereof. CT variables—criteria combinations, tumor side, study quality, and evidence of tumor rupture—were incorporated into a probabilistic classification tree model to assess predictive accuracy. Interobserver agreement was calculated via Kendall and correlation coefficients.

Results

Adhesions were documented surgically in 84.4% of patients, most commonly involving renal vein (50%), caudal vena cava (40.6%), and cranial mesenteric artery (21.9%). A single CT criterion alone lacked predictive capability; however, specific combinations of two or more criteria significantly improved accuracy. For example, renal vein adhesion was predicted with 84.8% accuracy using the combination of no cleavage plane + fat stranding + increased contact surface + mass effect + traction displacement—rising to 93.8% when CT image quality was excellent. For the CVC, a combination of no cleavage plane + increased contact surface + mass effect + traction displacement in right-sided tumors provided 86.2% accuracy. Kidney adhesion prediction reached 90–100% accuracy with combinations of organ criteria, particularly when tumors were right-sided. Interobserver agreement was excellent for CVC and liver, and good to moderate for most other structures.

Limitations

Sample size was modest, particularly for feline subjects (n=2), limiting species-specific conclusions. Heterogeneous CT protocols across institutions introduced variability in image quality, which influenced model performance. No histopathology standard for adhesion exists, and contact without adhesion created an ambiguous diagnostic zone. Some organs (e.g., colon) were not included in surgical reporting, limiting correlation. Low numbers of aortic adhesions prevented generation of a classification tree for that vessel, despite clinically important complications observed.

Conclusions

CT-based prediction of adrenal tumor adhesions is feasible when using combinations of defined imaging criteria rather than single markers. Probabilistic decision trees effectively identified the most predictive variables and yielded strong accuracy for several organs and vessels, improving preoperative planning. Radiologist agreement was high for key vessels, supporting repeatability of the criteria. These findings support incorporating structured CT adhesion assessment into routine evaluation of adrenal tumors, though larger multicenter datasets are needed to strengthen generalizability.

Disclaimer: The summary generated in this email was created by an AI large language model. Therefore errors may occur. Reading the article is the best way to understand the scholarly work. The figure presented here remains the property of the publisher or author and subject to the applicable copyright agreement. It is reproduced here as an educational work. If you have any questions or concerns about the work presented here, reply to this email.