Robust automatic coregistration, segmentation, and classification of cell nuclei in multimodal cytopathological microscopic images

Abstract

The paper describes the key component of the Multimodal Cell Analysis approach, a novel cytologic evaluation method for early cancer detection. The approach is based on repeated staining of a cell smear. The correlation of features and data extracted from the different stains, and related to relocated individual cells, may yield a dramatic increase of diagnostic reliability.In order to utilise the technique, fully automatic, adaptive image preprocessing techniques need to be applied, which are described in this article: coregistration of multimodal images, segmentation, and classification of cell nuclei. The presented feasibility study shows both efficiency and robustness of all steps being high regarding medical image material, and it strongly supports clinical application.

Segmentation and Detection of Nuclei in Silver Stained Cell Specimens for Early Cancer Diagnosis

“…Rather than directly applying the silver stain, these specimens were first stained using the so-called Feulgen stain, where the nuclei in general appear very clearly. An accurate segmentation of the nuclei in this staining is obtained as decribed in [17]. After review by an experienced cytopathologist, these segmentations provide the gold standard against which the segmentation results for silver stain obtained by the described algorithm are compared.…”

“…Note that, in contrast to a gold standard created by pure manual segmentation by an expert, this procedure reduces the influences of inter-and intra-individual variability. After de-staining to remove the Feulgen stain and subsequent re-staining with silver nitrate, images of the same cells were acquired and registered [17] to their corresponding Feulgen images. For corresponding nuclei segmentations between the Feulgen reference and the silver stain data, we calculated the non-overlapping area (symmetric difference) per nucleus ΔA = A Feulgen A silver .…”

Overlapping nuclei segmentation using direction‐based flow tracking

“…Also, the coupled parametric active contour [15] solves the overlapping contours in multiple overlapping nuclei. In addition, the determination of knot of the contour [17] and the cooperation between watershed and active contour [14] are also available strategies. On the accurate segmented boundary and the uncomplicated operation from the active contour method and the watershed algorithm, we attempt to combine their advantages together for improving the segmentation ACC.…”

On the use of classification reliability for improving performance of the one-per-class decomposition method

“…Pattern recognition techniques have been widely applied to medical and biological domains, such as mammography [4,10,12], microscope imaging [32,34,40,46] and radiography [43], to name a few. In general, classification applications require to handle both binary and multiclass recognition problems.…”