| Resumo: | The observation in microscopy of histological sections allows us to evaluate structural differences, in pancreatic cells, between rats with normal glucose tolerance and with glucose intolerance (pre-diabetic) situation. Nevertheless, this pre-diabetic condition implies subtle changes in islets of Langerhans structure. This and the normal variability among sampled cells makes difficult the task of identifying glucose intolerance (pre-diabetic situation) with a low level of error. This paper presents preliminary results in the processing of histological pancreas images with the goal of identifying pre-diabetic situation in Wistar rats. The immediate goal of this work is to evaluate the performance of a classifier based in a morphometric measurement of the histological images and to assess the potential for image based automatic processing and classification. A set of 90 images, were used (58 from rats with normal glucose tolerance, and 32 from pre-diabetic ones). These images were segmented manually using ImageJ. This segmentation and area measurements have been speedup by the application of ImageJ macros which were defined for this purpose. The ratio, between the area of -cells and the islets of Langerhans , was used has the indicator of the prediabetic situation. Considering this feature, a receiver operating characteristic analysis has been performed. True positive rate, vs. false positive rate shows the predicted performance of a binary classifier as its discrimination threshold is varied.
|
|---|