AN AI-BASED ALGORITHM FOR THE AUTOMATIC EVALUATION OF IMAGE QUALITY IN CANINE THORACIC RADIOGRAPHS

Radiography is the most widely used imaging technique for the evaluation of the canine thorax. Obtaining high-quality images is paramount for a correct interpretation of thoracic radiographs (1). In human medicine specific guidelines on diagnostic image quality were established (2). To date, no specific standards for diagnostic image quality were established in veterinary medicine. The aim of this study was to develop and test an artificial intelligence (AI)-based algorithm, based on ResNet-50 pre-trained on ImageNet, for detecting common technical errors in canine thoracic radiography. The algorithm was trained using a database of 7975 thoracic radiographs (4776 latero- lateral and 3199 sagittal projections) from three different veterinary clinics in Italy. The radiographs were evaluated for image quality by three experienced veterinary diagnostic imagers. The algorithm was designed to classify the images as correct or having one or more of the following errors: rotation (superimposition of opposite ribs), underexposure (evident quantum mottle or lack of detail of pulmonary structures), overexposure, incorrect limb positioning, incorrect neck positioning, blurriness (evident motion artefacts), cut, or presence of foreign objects or medical devices. All the tags, except for “correct”, were not mutually exclusive and therefore a multi-label deep-learning approach was used. The algorithm correctly identified errors in thoracic radiographs with an overall accuracy of 81.5% in latero-lateral and 75.7% in sagittal images. The most accurately identified errors were limb mispositioning and underexposure both in latero-lateral (AUC 0.93 and 0.88, respectively) and sagittal images (AUC 0.88 and 0.92, respectively). The accuracy of the developed model in the classification of technically correct radiographs was fair in latero-lateral (AUC=0.77) and good in sagittal (AUC=0.81) images. The authors conclude that their AI-based algorithm is a promising tool for improving the accuracy of radiographic interpretation by identifying technical errors in canine thoracic radiographs.