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

Advancements in the field of artificial intelligence (AI) are modest in veterinary medicine relative to their substantial growth in human medicine. However, interest in this field is increasing, and commercially available veterinary AI products are already on the market. In this retrospective, diagnostic accuracy study, the accuracy of a commercially available convolutional neural network AI product (Vetology AI®) is assessed on 56 thoracic radiographic studies of pulmonary nodules and masses, as well as 32 control cases. Positive cases were confirmed to have pulmonary pathology consistent with a nodule/mass either by CT, cytology, or histopathology. The AI software detected pulmonary nodules/masses in 31 of 56 confirmed cases and correctly classified 30 of 32 control cases. The AI model accuracy is 69.3%, balanced accuracy 74.6%, F1-score 0.7, sensitivity 55.4%, and specificity 93.75%. Building on these results, both the current clinical relevance of AI and how veterinarians can be expected to use available commercial products are discussed.

Prevalence of pulmonary nodules in dogs with malignant neoplasia as determined by CT.

In order to estimate the prevalence of pulmonary nodules in dogs with nonpulmonary malignant neoplasia, medical record descriptions of CT findings in dogs diagnosed with nonpulmonary malignant neoplasia were reviewed retrospectively. A total of 536 dogs were sampled from a single hospital. For malignant neoplasms with >10 affected individuals, prevalence of multiple pulmonary nodules at first CT was hemangiosarcoma 24 of 58 (41%), osteosarcoma 14 of 55 (26%), carcinoma 20 of 80 (25%), histiocytic sarcoma five of twenty-one (24%), soft tissue sarcoma 13 of 57 (23%), adenocarcinoma 11 of 60 (18%), melanoma five of thirty-seven (14%), lymphoma 10 of 76 (13%), mast cell tumor two of forty-seven (4%), and squamous cell carcinoma zero of seventeen (0%). A solitary pulmonary nodule was identified at first CT in 33 (6%) dogs. Of these, nine had follow-up CT, including two dogs in which the nodule disappeared, three dogs in which the size of the nodule did not change, and four dogs in which the nodule enlarged and additional pulmonary nodules appeared. Dogs with hemangiosarcoma were most likely to have signs of pulmonary metastasis at first CT, whereas dogs with mast cell tumor were infrequently affected, and no dog with squamous cell carcinoma had signs of pulmonary metastasis. A solitary pulmonary nodule at first CT was an indeterminate finding, potentially unassociated with neoplasia.

Computed tomographic findings in 15 dogs with eosinophilic bronchopneumopathy.

Eosinophilic bronchopneumopathy is a disease characterized by the infiltration of the lung and bronchial mucosa by eosinophils. The aim of the present study was to describe the CT findings in a large series of dogs with confirmed diagnosis of eosinophilic bronchopneumopathy. Computed tomographic scans of 15 dogs with confirmed diagnosis of eosinophilic bronchopneumopathy were evaluated retrospectively by two boarded radiologists who reached a consensus. Abnormalities were identified in 14/15 (93%) dogs, including pulmonary parenchymal abnormalities in 14/15 (93%) dogs, bronchial wall thickening in 13 (87%) dogs, which was considered marked in eight (53%), plugging of the bronchial lumen by mucus/debris in 11 (73%) dogs, and bronchiectasis in nine (60%) dogs. Pulmonary nodules were identified in 5/15 (33%) dogs including one dog with a mass. All dogs with a nodular lung pattern had additional abnormalities. Lymphadenopathy was present in 10 dogs (67%). Lesions associated with eosinophilic bronchopneumopathy are variable and heterogeneous and encompass a wider variety of computed tomographic features than reported previously. Computed tomographic images were abnormal in the majority of affected dogs, hence CT is a useful modality to characterize the nature and distribution of thoracic lesions in dogs with eosinophilic bronchopneumopathy.

Bilateral mandibular pyogranulomatous lymphadenitis and pulmonary nodules in a dog with Bartonella henselae bacteremia.

This report describes a 2-year-old collie dog with pulmonary nodules, visualized by computed tomographic (CT) scan, with evidence of Bartonella henselae bacteremia and pyogranulomatous lymphadenitis. Clinical signs resolved with antimicrobial therapy.

Lymphadénite pyogranulomateuse mandibulaire latérale et nodules pulmonaires chez un chien atteint de bactériémie àBartonella henselae. Ce rapport décrit un chien Collie âgé de 2 ans atteint de nodules pulmonaires, visualisés par tomodensitométrie, avec des signes de bactériémie à Bartonella henselae et de lymphadénite pyogranulomateuse. Les signes cliniques se sont résorbés avec un traitement antimicrobien.(Traduit par Isabelle Vallières).

CT and PET-CT of a dog with multiple pulmonary adenocarcinoma.

A 10-year-old, intact female Yorkshire terrier had multiple pulmonary nodules on thoracic radiography and ultrasonography with no lesions elsewhere. Computed tomography (CT) and positron emission tomography and computed tomography (PET-CT) using 18F-fluorodeoxyglucose (FDG) were performed to identify metastasis and undetected primary tumors. On CT examination, pulmonary nodules had a hypoattenuating center with thin peripheral enhancement, suggesting ischemic or necrotizing lesion. In PET-CT at 47 min after intravenous injection of 11.1 MBq/kg of FDG, the maximum standardized uptake value of each pulmonary nodule was about from 3.8 to 6.4. There were no abnormal lesions except for four pulmonary nodules on the CT and PET-CT. Primary lung tumor was tentatively diagnosed, and palliative therapy using 2 mg/kg tramadol and 2.2 mg/kg carprofen twice per day was applied. After the dog's euthanasia due to deteriorated clinical signs and poor prognosis, undifferentiated pulmonary adenocarcinoma was diagnosed through histopathologic and immunochemistry examination. To the best of the authors' knowledge, this is the first study of CT and PET-CT features of canine pulmonary adenocarcinoma. In this case, multiple pulmonary adenocarcinoma could be determined on the basis of FDG PET-CT through screening the obvious distant metastasis and/or lymph node invasions and excluding unknown primary tumors.

Comparison of examination of thoracic radiographs and thoracic computed tomography in dogs with appendicular osteosarcoma.

Appendicular osteosarcoma (OSA) is a highly metastatic tumour in dogs. The aim of the study was to compare thoracic radiographs with thoracic computed tomography (CT) in the staging of canine appendicular OSA. In all, 39 canine patients histologically diagnosed with OSA were reviewed in the retrospective study. All dogs underwent radiographic examination as well as CT examination of the thoracic cavity. Pulmonary nodules were detected radiographically in two cases (5%), whereas the CT imaging showed that pulmonary nodules were evident in 11 cases (28%, P = 0.024). There was an improved detection of small pulmonary nodules in the lung parenchyma with CT (P = 0.021). The number of nodules in CT examination had a significant negative influence on survival time (P = 0.005). However, whether nodules were present in CT or not did not influence overall survival (P = 0.368). CT examination was superior to thoracic radiography in the screening and detection of pulmonary nodules in dogs with OSA.

Comparison of single-slice computed tomography protocols for detection of pulmonary nodules in dogs.

Two dogs (4 and 38 kg) with radiographic evidence of pulmonary nodules were evaluated using single-slice, helical computed tomography (CT). Each thorax was scanned using 12 combinations of examination parameters that included slice collimation width (3 and 5mm for the small dog and 5 and 7mm for the large dog), pitch (1, 1.5, and 2), and reconstruction interval (0.5 and 1). Sensitivity, specificity, and accuracy for nodule detection were evaluated for each protocol by three different observers, their results being compared with a consensus evaluation of images acquired with the protocol providing the best theoretic resolution (narrow collimation, pitch of 1, reconstruction interval of 0.5). For all observers, sensitivity and accuracy were significantly increased when using a protocol with narrow collimation (P < 0.0001-0.005 and P = 0.0003-0.005, respectively). Pitch and reconstruction interval did not significantly influence the accuracy, sensitivity, or specificity for at least two of the observers. Additionally, nodule size (< 3mm vs. > 3mm) did not significantly affect nodule detection. Interobserver repeatability was variable among protocols (K = 0.32-0.78), highlighting the fact that nodule detection may be more dependent on the observer than on the choice of the CT protocol. For single-slice CT, the results of this study suggest that narrow collimation (3-5 mm, depending on the animal's size), a pitch of 2 and a reconstruction interval of 1 should be used in dogs for the detection of pulmonary nodules.