Ewert Bengtsson

Combining intensity, edge and shape information for 2D and 3D segmentation of cell nuclei in tissue sections

We present a region‐based segmentation method in which seeds representing both object and background pixels are created by combining morphological filtering of both the original image and the gradient magnitude of the image. The seeds are then used as starting points for watershed segmentation of the gradient magnitude image. The fully automatic seeding is done in a generous fashion, so that at least one seed will be set in each foreground object. If more than one seed is placed in a single object, the watershed segmentation will lead to an initial over‐segmentation, i.e. a boundary is created where there is no strong edge. Thus, the result of the initial segmentation is further refined by merging based on the gradient magnitude along the boundary separating neighbouring objects. This step also makes it easy to remove objects with poor contrast. As a final step, clusters of nuclei are separated, based on the shape of the cluster. The number of input parameters to the full segmentation procedure is only five. These parameters can be set manually using a test image and thereafter be used on a large number of images created under similar imaging conditions. This automated system was verified by comparison with manual counts from the same image fields. About 90% correct segmentation was achieved for two‐ as well as three‐dimensional images.

A feature set for cytometry on digitized microscopic images.

Feature extraction is a crucial step in most cytometry studies. In this paper a systematic approach to feature extraction is presented. The feature sets that have been developed and used for quantitative cytology at the Laboratory for Biomedical Image Analysis of the GSF as well as at the Center for Image Analysis in Uppsala over the last 25 years are described and illustrated. The feature sets described are divided into morphometric, densitometric, textural and structural features. The latter group is used to describe the eu‐ and hetero‐chromatin in a way complementing the textural methods. The main goal of the paper is to bring attention to the need of a common and well defined description of features used in cyto‐ and histometrical studies. The application of the sets of features is shown in an overview of projects from different fields. Finally some rules of thumb for the design of studies in this field are proposed. Colour figures can be viewed on http://www.esacp.org/acp/2003/25‐1/rodenacker.htm.

Algorithms for cytoplasm segmentation of fluorescence labelled cells

Automatic cell segmentation has various applications in cytometry, and while the nucleus is often very distinct and easy to identify, the cytoplasm provides a lot more challenge. A new combination of image analysis algorithms for segmentation of cells imaged by fluorescence microscopy is presented. The algorithm consists of an image pre‐processing step, a general segmentation and merging step followed by a segmentation quality measurement. The quality measurement consists of a statistical analysis of a number of shape descriptive features. Objects that have features that differ to that of correctly segmented single cells can be further processed by a splitting step. By statistical analysis we therefore get a feedback system for separation of clustered cells. After the segmentation is completed, the quality of the final segmentation is evaluated. By training the algorithm on a representative set of training images, the algorithm is made fully automatic for subsequent images created under similar conditions. Automatic cytoplasm segmentation was tested on CHO‐cells stained with calcein. The fully automatic method showed between 89% and 97% correct segmentation as compared to manual segmentation.

Paraffin section storage and immunohistochemistry - Effects of time, temperature, fixation, and retrieval protocol with emphasis on p53 protein and MIB1 antigen

It has been observed that immunoreactivity in paraffin sections decreased during storage. In this study, stored paraffin sections from both biopsy material and cultured cells were assessed for changes in immunoreactivity, using color-based image analysis to quantitate extent and intensity of the stainings. For seven of the 11 antibodies studied, storage at 20 degrees C for 16 weeks reduced the extent of immunostaining compared with that of freshly cut sections. Furthermore, increased storage temperatures resulted in a progressive loss of immunoreactivity. After 2 weeks of storage, at both 4 degrees C and 20 degrees C, p53 protein- and MIB1-antigen expression was significantly reduced regarding extent and intensity. The extent of the immunoreactivity reduced more for p53 protein than for MIB1 antigen, but the intensity did not. Boric acid was used for antigen retrieval on sections stored for 12 weeks at 20 degrees C. For both p53 protein and MIB1 antigen, this resulted in an extent and intensity of immunostaining equal to or higher than (MIB1) that obtained in freshly cut sections, using citrate buffer. Staining of cultured cells confirmed the results from biopsy material on the influence of storage temperature. Fixation time only marginally influenced the storage-related decrease in immunoreactivity. In conclusion, storage of paraffin sections leads to a varying degree of decreased immunoreactivity for several antibodies. The degree is at least partly dependent on storage time and temperature but not fixation time. However, this may be compensated for by optimizing the antigen retrieval protocol.

PREDICTING BREAST CANCER RECURRENCE

The prognostic value of 435 cytochemical, cytometrical, morphological, epidemiological, and clinical variables was analyzed in a prospective study of 179 breast cancer patients followed for five years after mastectomy. A variable reduction was obtained by first selecting variables correlated with recurrence rate in direct (Students t test) or correlation analysis with consideration of the type of variable analyzed (nominal, interval, ordinal). The 20 variables most strongly correlated with recurrence were analyzed by logistic stepwise regression analysis in order to find out what combination of variables had most discriminatory power in predicting recurrence. It was found that axillary metastization as such was correlated with a combination of variables describing mitotic frequency, size of primary tumor and differentiation of the primary tumor (average cluster size in fine‐needle biopsies). It was also found that there was a strong time dependency in the predictive power of the variables, so that different variable combinations predicted the recurrence rate during the first 2.5 year period (size of axillary metastases and primary tumor, number of lymphocytes around the tumor, mitotic frequency, and degree of differentiation) compared with the second 2.5 year period (variance of DNA content among tumor cell nuclei, number of lymphocytes around the tumor, occurrence of multiple tumors in the operated breast and occurrence of breast cancer among relatives). While other factors previously shown to be correlated with risk of recurrence were also found to be positively correlated here, they were neither as highly predictive as, nor did they increase the predictive value of the above mentioned combined variables. The current study strongly emphasizes that, at the present time, studies of recurrence prediction in human breast cancer should be based on an optimal combination of a number of variables which, independently, influence the prognosis. Further, the current study indicates that prerequisite methods for predicting breast cancer recurrence exist today.

Image analysis for automatic segmentation of cytoplasms and classification of Rac1 activation

Rac1 is a GTP‐binding molecule involved in a wide range of cellular processes. Using digital image analysis, agonist‐induced translocation of green fluorescent protein (GFP) Rac1 to the cellular membrane can be estimated quantitatively for individual cells.

CBA—an atlas-based software tool used to facilitate the interpretation of neuroimaging data

CBA, a software tool used to improve quantification and evaluation of neuroimaging data has been developed. It uses a detailed 3-dimensional brain atlas that can be adapted to fit the brain of an individual patient represented by a series of displayed images. Anatomical information from the atlas can then be introduced into the images. If the patient has been imaged in different modalities, adaptation of the atlas to the different images will provide the transformation that brings the images into registration. CBA can thus be used as a tool for fusing multimodality information from the same patient. Furthermore, by applying the inverse atlas transformation, images from a patient can be transformed to conform to the anatomy of the atlas brain. This anatomical standardization, where the atlas brain itself serves as the anatomy standard, brings data from different individuals into a compatible form providing possibilities to perform individual-group and group-by-group comparisons between patients and normal controls.

Principal component analysis of dynamic positron emission tomography images

Multivariate image analysis can be used to analyse multivariate medical images. The purpose could be to visualize or classify structures in the image. One common multivariate image analysis technique which can be used for visualization purposes is principal component analysis (PCA). The present work concerns visualization of organs and structures with different kinetics in a dynamic sequence utilizing PCA. When applying PCA on positron emission tomography (PET) images, the result is initially not satisfactory. It is illustrated that one major explanation for the behaviour of PCA when applied to PET images is that it is a data-driven technique which cannot separate signals from high noise levels. With a better understanding of the PCA, gained with a strategy of examining the image data set, the transformations, and the results using visualization tools, a surprisingly easily understood methodology can be derived. The proposed methodology can enhance clinically interesting information in a dynamic PET imaging sequence in the first few principal component images and thus should be able to aid in the identification of structures for further analysis.

A new method for segmentation of colour images applied to immunohistochemically stained cell nuclei

A new method for segmenting images of immunohistochemically stained cell nuclei is presented. The aim is to distinguish between cell nuclei with a positive staining reaction and other cell nuclei, and to make it possible to quantify the reaction. First, a new supervised algorithm for creating a pixel classifier is applied to an image that is typical for the sample. The training phase of the classifier is very user friendly since only a few typical pixels for each class need to be selected. The classifier is robust in that it is non‐parametric and has a built‐in metric that adapts to the colour space. After the training the classifier can be applied to all images from the same staining session. Then, all pixels classified as belonging to nuclei of cells are grouped into individual nuclei through a watershed segmentation and connected component labelling algorithm. This algorithm also separates touching nuclei. Finally, the nuclei are classified according to their fraction of positive pixels.

Automatic Quantification of Immunohistochemically Stained Cell Nuclei Based on Standard Reference Cells

A fully automatic method for quantification of images of immunohistochemically stained cell nuclei by computing area proportions, is presented. Agarose embedded cultured fibroblasts were fixed, paraffin embedded and sectioned at 4 µm. They were then stained together with 4 µm sections of the test specimen obtained from bladder cancer material. A colour based classifier is automatically computed from the control cells. The method was tested on formalin fixed paraffin embedded tissue section material, stained with monoclonal antibodies against the Ki67 antigen and cyclin A protein. Ki67 staining results in a detailed nuclear texture with pronounced nucleoli and cyclin A staining is obtained in a more homogeneously distributed pattern. However, different staining patterns did not seem to influence labelling index quantification, and the sensitivity to variations in light conditions and choice of areas within the control population was low. Thus, the technique represents a robust and reproducible quantification method. In tests measuring proportions of stained area an average standard deviation of about 1.5% for the same field was achieved when classified with classifiers created from different control samples.