Análise de imagens do núcleo celular como parâmetro prognóstico de mastocitomas em cães

Abstract

Mast cell tumor (MCT) is one of the most common cutaneous neoplasms in dogs and its biological behavior is variable. Histological grading methods are traditionally used to define the prognosis, but these methods present considerable subjectivity. Automated analysis of cell nucleus images may be an alternative to eliminate subjectivity and increase the accuracy of prognostic evaluation. The objective of this study was to identify which nuclear characteristics are most relevant to determine the prognosis of canine cutaneous MCTs, in addition to evaluating the relationship between histological grading and nuclear analysis. Fifty cases of MCT in dogs with known progression were used. Diagnosis and histological grading were performed under light microscopy, using the classification systems of Patnaik, Ehler and MacEwen (1984) and Kiupel et al. (2011). Digital images in bitmap file were obtained from each slide using a microscope (Leica DM500) coupled to an image capture system (Leica ICC50) with a 100X immersion objective. One hundred neoplastic cell nuclei from each slide were analyzed. The nuclei were interactively segmented using ImageJ® software and pasted into a new blank image and saved as a bitmap file. Only neoplastic and non-overlapping nuclei were segmented. Then, the nuclei were transformed into pixel values between 0 (black) and 255 (white) and analyzed using an algorithm developed in the Scilab® mathematical programming environment. Through the analysis of images of slides stained with hematoxylin and eosin, 29 characteristics related to the size, shape, pleomorphism and texture of the chromatin of the nucleus were evaluated. Canonical discriminant analyses were performed for prognosis and histological grading and showed 91.31%, 94% and 96% accuracy in determining the prognosis and grading according to the system of Patnaik, Ehler and MacEwen (1984) and Kiupel et al. (2011), respectively. The characteristics related to the size and shape of the nucleus were the most relevant for the prognosis. The characteristics related to the chromatin texture demonstrated a lower degree of relevance compared to the variables related to the morphology of the nucleus, both for the assignment of grading and for the prognostic evaluation of canine cutaneous MTCs. The present study demonstrated that the algorithm based on grayscale thresholding for digital image analysis can benefit studies with MCT in dogs. The prognosis can be determined from the analysis of nuclear characteristics, as well as the histological grading for MCT, currently used in veterinary medicine.