Estudo comparativo entre estereologia e análise de imagem para quantificação rápida e precisa em patologia experimental
Miniatura indisponível
Data
2025
Autores
Título da revista
ISSN da revista
Título do Volume
Editora
Resumo
A análise de imagem representa um avanço na avaliação quantitativa em patologia. Os programas de código aberto, como o QuPath, facilitam o acesso a essas ferramentas, mas a precisão e exatidão podem variar, e inferir estruturas tridimensionais a partir de estruturas bidimensionais pode induzir em erro. Este estudo combinou a estereologia com a análise de imagem para criar uma abordagem precisa e rápida. Foram aproveitados blocos de parafina de estudos de melanoma metastático pulmonar e fibrose cardíaca. A estereologia foi realizada no Visiopharm, e a análise de imagem no QuPath, utilizando o threshold para as metástases pulmonares e o machine learning para a fibrose cardíaca. Não houve diferenças estatísticas entre os métodos (p>0,05). O coeficiente de correlação foi de r=1,000 para volume pulmonar e metástases, e r=0,900 para percentagem metástase/pulmão. Para o volume do miocárdio ventricular foi de r=0,900, r=1,000 para o volume de fibrose e para a percentagem fibrose/ventrículo. Entre lâminas do mesmo animal, observou-se coeficientes de variação elevados indicado uma forte heterogeneidade. Conclui-se que a integração da estereologia com a análise de imagem gera estimativas de volume precisas e rápidas. Além disso, não é adequado confiar apenas em uma única secção bidimensional para representar uma lesão tridimensional.
Image analysis represents a significant advancement in quantitative pathology assessment. Open-source programs, such as QuPath, enhance accessibility to these tools; however, accuracy and precision may vary, and inferring three-dimensional structures from two-dimensional representations can lead to errors. This study combined stereology with image analysis to develop a rapid and accurate approach. Paraffin blocks from studies on pulmonary metastatic melanoma and cardiac fibrosis were used. Stereology was performed using Visiopharm, and image analysis was conducted with QuPath, applying the threshold method for pulmonary metastases and machine learning for cardiac fibrosis. No statistical differences were found between the methods (p>0.05). The correlation coefficient was r=1.000 for lung and metastasis volumes and r=0.900 for the metastasis/lung ratio. For ventricular myocardial volume, the coefficient was r=0.900, and r=1.000 for both fibrosis volume and the fibrosis/ventricle ratio. High coefficients of variation were observed among slides from the same animal, indicating significant heterogeneity. In conclusion, integrating stereology with image analysis provides precise and rapid volume estimations. Although relying solely on a single two-dimensional section to represent a three-dimensional lesion is not appropriate.
Image analysis represents a significant advancement in quantitative pathology assessment. Open-source programs, such as QuPath, enhance accessibility to these tools; however, accuracy and precision may vary, and inferring three-dimensional structures from two-dimensional representations can lead to errors. This study combined stereology with image analysis to develop a rapid and accurate approach. Paraffin blocks from studies on pulmonary metastatic melanoma and cardiac fibrosis were used. Stereology was performed using Visiopharm, and image analysis was conducted with QuPath, applying the threshold method for pulmonary metastases and machine learning for cardiac fibrosis. No statistical differences were found between the methods (p>0.05). The correlation coefficient was r=1.000 for lung and metastasis volumes and r=0.900 for the metastasis/lung ratio. For ventricular myocardial volume, the coefficient was r=0.900, and r=1.000 for both fibrosis volume and the fibrosis/ventricle ratio. High coefficients of variation were observed among slides from the same animal, indicating significant heterogeneity. In conclusion, integrating stereology with image analysis provides precise and rapid volume estimations. Although relying solely on a single two-dimensional section to represent a three-dimensional lesion is not appropriate.
Descrição
Palavras-chave
VETERINARY MEDICINE, IMAGE ANALYSIS, ARTIFICIAL INTELLIGENCE, MACHINE LEARNING, VETERINARY PATHOLOGY, FIBROSIS, PULMONARY FIBROSIS, NEOPLASMS, QUANTITATIVE ANALYSIS, MESTRADO INTEGRADO EM MEDICINA VETERINÁRIA, VETERINÁRIA, MEDICINA VETERINÁRIA, ANÁLISE DA IMAGEM, INTELIGÊNCIA ARTIFICIAL, APRENDIZAGEM COMPUTACIONAL, PATOLOGIA VETERINÁRIA, FIBROSE, FIBROSE PULMONAR, NEOPLASIAS, ANÁLISE QUANTITATIVA