RESUMO
Automated karyotyping is of great importance for cytogenetic research, as it speeds up the process for cytogeneticists through incorporating AI-driven automated segmentation and classification techniques. Existing frameworks confront two primary issues: Firstly the necessity for instance-level data annotation with either detection bounding boxes or semantic masks for training, and secondly, its poor robustness particularly when confronted with domain shifts. In this work, we first propose an accurate segmentation framework, namely KaryoXpert. This framework leverages the strengths of both morphology algorithms and deep learning models, allowing for efficient training that breaks the limit for the acquirement of manually labeled ground-truth mask annotations. Additionally, we present an accurate classification model based on metric learning, designed to overcome the challenges posed by inter-class similarity and batch effects. Our framework exhibits state-of-the-art performance with exceptional robustness in both chromosome segmentation and classification. The proposed KaryoXpert framework showcases its capacity for instance-level chromosome segmentation even in the absence of annotated data, offering novel insights into the research for automated chromosome segmentation. The proposed method has been successfully deployed to support clinical karyotype diagnosis.
Assuntos
Cariotipagem , Humanos , Cariotipagem/métodos , Metáfase , Algoritmos , Cromossomos Humanos/genética , Processamento de Imagem Assistida por Computador/métodos , Aprendizado ProfundoRESUMO
In the spring of 2023, 10 to 21-day-old chicks in a broiler duck farm in Shandong Province, China, developed swelling of the head and neck, moist eyes with mucous discharge, difficulty in walking, shrinking of the neck, and loose and disorganized coat. Anatomical observation revealed hemorrhages in the esophageal mucosa, myocardium, and liver, and severe hemorrhages in the trachea with copious inflammatory secretions. Soon after, similar symptoms appeared in a large number of ducks in the flock, which eventually led to the elimination of all the 20,000-odd newly introduced ducklings on the farm, resulting in huge economic losses. We detected duck plague virus in the tissues of liver, spleen and lungs of diseased and dead ducks, and successfully isolated the pathogenic strain, named SD423, by inoculating duck embryos and inoculating duck embryo fibroblasts. We successfully conducted animal regression experiments with the isolated strain, and the experimental animals in the 1 d of age group showed symptoms of swollen eyes and tearing, shrinking of the neck, crouching, and hemorrhage in organs such as the liver and intestines successively from the 3rd d. We sequenced the whole genome of the isolated duck plague strain, and by comparing the homology with the published duck plague virus whole sequences in Genbank, the virus strain obtained in this study had the highest homology with the Chinese virulent strain SD (MN518864.1), with nucleotide (nt) homology of about 99.90% and amino acid (aa) homology of about 99.75%, which indicated that the isolate is a virulent strain. Previously, it was reported that the natural infection of duck plague virus mainly occurs above 30 d of age, but the duck plague virus found in this study can naturally infect ducklings up to 20 d of age, and the mortality rate is as high as 100%. In this study, the pathogenicity test and whole genome sequence analysis of this isolate provided data support and theoretical basis for further research on pathogenicity and virulence-related gene analysis of duck plague virus.