Abstract
Purpose :
To develop an automatic method to count lens epithelial cells and to detect the expression of fluorescent signal of biomarkers in nucleus and cytoplasm of lens epithelial cells in a histological section.
Methods :
An automatic algorithm was developed in Matlab to localize and quantify lens epithelial cells. Active caspase-3 florescent expression was selected as the model florescent signal. Then, an expert observer classified the fluorescent signal in the nuclei as ‘labelled’ or ‘not labelled’ in 18 lenses. The probability of labelling estimated by the expert observer was used as the reference for setting the threshold for classification of nuclei and cytoplasms in the automatic algorithm. Then, ten blinded observers classified the fluorescent signal of active caspase-3 in the nuclei of one lens image and recorded the proportion of labelling. Time consumed by the automatic algorithm and the observers for the classification of nuclei was recorded.
Results :
The threshold for classification of a nucleus as labelled was determined to 4.9 times local background fluorescence intensity. On an average, the manual observers reported higher proportion of nuclear labeling than the automatic algorithm and the classification varied considerably among them. The automatic algorithm did not classify any cytoplasm as labelled. The time consumed for classification was considerably shorter for the automatic algorithm than for the manual observers.
Conclusions :
The presently developed technique enables objective and fast quantification of lens epithelial cells and classification of fluorescent signal of biomarkers in a histological section of rat lens.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.