The integration of artificial intelligence (AI) into chest radiography (CXR) has greatly impacted both human and veterinary medicine, enhancing diagnostic speed, accuracy, and efficiency. In human medicine, AI has been extensively studied, improving the identification of thoracic abnormalities, diagnostic precision in emergencies, and the classification of complex conditions such as tuberculosis, pneumonia, and COVID-19. Deep learning-based models assist radiologists by detecting patterns, generating probability maps, and predicting outcomes like heart failure. However, AI is still supplementary to clinical expertise due to challenges such as data limitations, algorithmic biases, and the need for extensive validation. Ethical concerns and regulatory constraints also hinder full implementation. In veterinary medicine, AI is still in its early stages and is rarely used; however, it has the potential to become a valuable tool for supporting radiologists in the future. However, challenges include smaller datasets, breed variability, and limited research. Addressing these through focused research on species with less phenotypic variability (like cats) and cross-sector collaborations could advance AI in veterinary medicine. Both fields demonstrate AI’s potential to enhance diagnostics but emphasize the ongoing need for human expertise in clinical decision making. Differences in anatomy structure between the two fields must be considered for effective AI adaptation.
In recent years, deep learning has emerged as a promising technique for medical image analysis. However, this application domain is likely to suffer from a limited availability of large public datasets and annotations. A common solution to these challenges in deep learning is the usage of a transfer learning framework, typically with a fine-tuning protocol, where a large-scale source dataset is used to pre-train a model, further fine-tuned on the target dataset. In this paper, we present a systematic study analyzing whether the usage of small-scale in-domain x-ray image datasets may provide any improvement for landmark detection over models pre-trained on large natural image datasets only. We focus on the multi-landmark localization task for three datasets, including chest, head, and hand x-ray images. Our results show that using in-domain source datasets brings marginal or no benefit with respect to an ImageNet out-of-domain pre-training. Our findings can provide an indication for the development of robust landmark detection systems in medical images when no large annotated dataset is available.
Filopodia are thin, finger-like extensions that project from the surface of cells. Present in various cell types, analyzing these structures can yield sign
About Me
PhD Student at MaLGa - DIBRIS in the University of Genoa, Italy.
PhD Student at MaLGa - DIBRIS in the University of Genoa, Italy.