Automatic counting of trypanosomatid amastigotes in infected human cells

This article presents an automatic approach to counting amastigotes in human cells infected with Chagas. The approach is divided into four steps: first, morphological pretreatment removes the complex image background; sets are then segmented by unsupervised classification; the infected cells are then preserved using a thresholding process; and, finally, they undergo morphological granulometric processing and are filtered by the average. An experimental protocol was employed to compare the amastigotes nuclei labeled by a professional biochemist with the results obtained by the proposed approach. We observed that using the granulometric sieving conducted with square SE plus average size filtering is the best option to obtain the minor error and the best precision and using the granulometric sieving conducted with rhombus SE without average size filtering represents the best combination for obtaining the best F-measure and recall rates.