Automatic detection of hypoxia in renal tissue stained with HIF-1alpha

Abstract

Objective: The objective of this study was the identification of the stain HIF-alpha using the Image Cytometry, and to help to count the positive cells (with HIF-alpha) and the negative cells (without HIF-alpha) from the same sample.

Method: 17 images of renal tissues from male rats of Winstar lineage; overall, there were 12.587 objects (cells) in the images for analysis. The acquired images were then analyzed through the free softwares CellProfiler (version 2.1.1) and CellProfiler Analyst (version 2.0). In the software CellProfiler Anlyst, there was a separation with the classes of the object, using a classifier, and the classes were: 1) class with HIF-alpha and 2) class without HIF-alpha.

Results: With the data obtained through Score All, it was possible to calculate the percentage of cells that had HIF-alpha; out of 12.587 objects of the sample, 6.773 (54%) had HIF-alpha and 5.814 (46%) did not have HIF-alpha. Data of sensibility 0.90, specificity 0.84 and standard deviation 0.10 and 0.12.

Conclusion: The research shows that the free software CellProfiler, through the light microscope, was able to identify the stains, perform the machine's learning, and subsequently count and separate cells from distinct classes (with and without the stain of HIF-alpha).

Keywords: HIF 1α; Hypoxia; Image analysis; Machine learning.