Automated recognition of cell images in high grade malignant lymphoma and reactive follicular hyperplasia.

In the past, the comparison of results of studies on malignant lymphomas has been biased by the use of different classifications of the diseases and to an even greater extent by subjective interpretation in the classification of the tumour cells. To overcome these short-comings, we have developed cytometric features specifically for automated recognition of cell images from high grade malignant non-Hodgkin's lymphomas (NHL) and reactive lymphoid lesions. This study used a colour TV-microscope system, high resolution scanning (13.3 pixel/microns), and image processing to study a total of 3600 lymphoid cells from 15 high grade malignant NHL and three tonsils. Sixteen out of 64 features, especially developed for image analysis in cytological preparations, have been evaluated. Because of a considerable overlap of all the single features, no feature on its own allows reliable discrimination. But, multivariant analysis of suitable feature combinations resulted in reliable identification and discrimination of the most frequently occurring cell types. We show that the lymphocytes, centrocytes, centroblasts, immunoblasts and lymphoblasts, as they are defined by subjective morphological criteria in the Kiel-classification of malignant NHL, also form distinctive subpopulations on the basis of their objective mathematical cell features. Furthermore, we have shown that there are distinctive differences between the lymphoma cells and their benign counterparts derived from reactive lymphoid lesions.

On-line transmission electron microscopic image analysis of chromatin texture for differentiation of thyroid gland tumors.

Nuclei of the cells from the thyroid gland were analyzed in a transmission electron microscope by direct TV scanning and on-line image processing. The method uses the advantages of a visual-perception model to detect structures in noisy and low-contrast images. The features analyzed include area, a form factor and texture parameters from the second derivative stage. Three tumor-free thyroid tissues, three follicular adenomas, three follicular carcinomas and three papillary carcinomas were studied. The computer-aided cytophotometric method showed that the most significant differences were the statistics of the chromatin texture features of homogeneity and regularity. These findings document the possibility of an automated differentiation of tumors at the ultrastructural level.

Segmentation of stained blood cell images measured at high scanning density with high magnification and high numerical aperture optics.

In hematological morphology, it is necessary to resolve and analyze the smallest possible cellular details appearing in the light microscope. A prerequisite for computer-aided analysis of subtle morphological features is measuring the cells at a high scanning density with high magnification and high numerical aperture optics. Contrary to visual observations, the information content in a measured picture can be increased by setting the condensor's numerical aperture (NA) greater than the objective's NA. The complexity and heterogeneity of such cell images necessitate a new segmentation method that conserves the morphological information required in the subsequent image analysis, feature extraction, and cell classification. In our segmentation strategy, characteristic color difference thresholds for each nucleus and cytoplasm are combined with geometric operations, probability functions, and a cell model. All thresholds are repeatedly recalculated during the successive improvements of the image masks. None of the thresholds are fixed. This strategy segments blood cell images containing touching cells and large variations in staining, texture, size, and shape. Biological inconsistencies in the calculated cell masks are eliminated by comparing each mask with the cell model criteria integrated into the entire segmentation process. All 20,000 leukocyte images from 120 smears in our leukemia project were segmented with this method.

Leukemia-related morphological features in blast cells.

This paper investigates the use of image-processing methods to detect leukemia-related morphological differences in mononuclear blast cells. Routinely prepared Pappenheim-stained blood smears were scanned in a high-resolution color TV-microscope system. Eleven blast-cell classes (OMSBC, T-ALL, OMS, ALL, LBL, IBL, AUL, AML, AMOL, AMMOL, and CML) were analyzed with the nonparametric statistical software program "Classification and Regression Trees" (CART). This paper documents the initial statistical evaluation of 62 leukemia-related morphological features that directly measure and analyze the cell-related quantifiable differences occurring in the various blast cells. The 62 cell image features include both common cytophotometric features, and new texture and color features developed for this project. This study found that each leukemia specimen contains a dominant class of blasts that correlates with the specific leukemia, plus a distribution of blasts from related diseases. The present data suggest the existence of a distribution fingerprint pattern for each leukemia.

Computer-based cytophotometric classification of thyroid tumors in imprints.

This study of 14 follicular adenomas, 10 papillary carcinomas and 11 follicular carcinomas of the human thyroid gland demonstrates the possibility of a cytological tumor classification using digital picture processing. Routinely prepared, HE-stained imprints of surgical specimens were scanned under a light microscope at high resolution with a colour TV camera. The cell nuclei were segmented and analysed with an image processing system. The computer-aided cytophotometric methods detected the most significant differences in the chromatin texture with a criteria variance of texture line distances and texture points per texture knots. Using these criteria benign and malignant tumor types could be successfully differentiated.

Estimation of sampling errors in a high-resolution TV microscope image-processing system.

The basic postulate of this paper is that the commonly accepted sampling density of 2-4 pixels/micron in a high-resolution TV microscope system is too low to digitize exactly and analyze the complex cellular detail found in stained cell images. Depending on the specific microscope system, the required sampling density is much higher, lying between 15 and 30 pixels/micron. This sampling density is derived from the aliasing error, the resolution loss, and computational limitations. The mathematical and optical methods and equipment used to obtain these results are described in detail.

[Colorimetric and cytophotometric study of a modified Papanicolaou staining method in cervical cells]. / Kolorimetrische und zytophotometrische Untersuchung einer modifizierten Papanicolaou-Färbemethode in Zervikalzellen.

The staining patterns of a modified Papanicolaou staining method and of its components are compared with those from the common Papanicolaou method. The staining results are investigated by the CIE-DIN-standardized color analysis. The medical investigator is able to discriminate more colors than this cytophotometric method, but this method is superior in the exact determination of colors. For neither of the both Papanicolaou staining methods it is possible to indicate one or two wavelengths for cytophotometric measurement using narrow-band filters which allows an optimal automated separation of nuclei, cytoplasms and background of all cell types. Keeping to this usual color analyzing method there is a better evaluation of the common Papanicolaou staining patterns. More saturated colors in the cyanophilic cytoplasms with the Papanicolaou staining patterns and in the eosinophilic cytoplasms with the investigated modified Papanicolaou staining patterns would be desirable. Since the resulting color patterns are similar from both stains, the modified Papanicolaou method has to prefer if it proves to be more reproducible.

Computer analysis of chromatin arrangement and nuclear texture in follicular thyroid tumours.

The differentiation of the thyroid glands follicular neoplasias into adenomas and carcinomas is currently done using the histological criteria recommended by WHO. This pilot study of 10 human follicular carcinomas and 10 folliculars adenomas demonstrates the possibility of a cytological classification using digital picture processing of high resolution cell images. Giemsa stained paraplast sections were scanned with a Colour-TV-camera, different channels were used with respect to staining and analyzing methods and computed with an image processing system. The computer aided cytophotometric methods detected significant differences in the chromatin arrangement and structure.

Characterization of coronavirus-infected tissue cultures by computer-aided cytophotometry.

Coronavirus-infected cells were cytophotometrically measured and analyze during a replication cycle. Distinct cytoplasmic differences were observed in images scanned at 260 and 280 nm; these differences could be attributed to specific virologic changes using virologic tests. The application of computer-aided cytophotometry may prove to be a valuable approach in the analysis of infected cells.

Computer-aided image analysis for the differentiation of mononuclear cells in peripheral blood smears from leukemic patients.

Different leukemias originating from the lymphatic system and from the myeloid tissue of the bone marrow were selected to develop and test algorithms to cytophotometrically define leukemic cells. Two of these cases were of low-grade, one of intermediate-grade and one of high-grade malignancy. The malignant cell types of each leukemic form could be identified and characterized by assessing the chromatin texture of the nucleus in the leukemic cells. With this method, normal peripheral white blood cells could be differentiated from leukemic ones. The results correlate directly with established hematologic parameters as they are described in the literature. The investigation demonstrated the usefulness of computer methods in the differentiation of closely related cell structures.

Computer aided analysis of chromatin network and basophil color for differentiation of mononuclear peripheral blood cells.

Computer aided differentiation of plasmoblasts, Pfeiffer cells, immunoblasts, lymphocytes and centrocytes is achieved with the parameters of chromatin network arrangement and structure, and multispectral cytoplasm color. The digital methods involve: (a) segmenting the nuclear image into topographic sections and analyzing the optical density distribution from the chromatin in these sections; (b) determining the nuclear structure with a 7 x 7 median filter, gradient filter and contour following algorithms; and (c) clustering two-dimensional chromatic data from panoptically stained cellular components. The parameters reported here are a subset of those needed for the automated diagnosis of many hematologic diseases especially the leukemias.

Bone marrow cell scene segmentation by computer-aided color cytophotometry.

Computer scene segmentation of touching cell images in bone marrow, on the basis of color information, is achieved using digitized scans at three different wavelengths of light. With trivariate histograms and Euler's coordinate transformation, it is possible cytophotometrically to isolate, on the basis of chromatic differences, individual heterogeneous cells located in cell groups. The ability of the described computer methods to isolate correctly the touching cell images is determined by visual comparison of the cells as seen in the microscope and the computer-generated displays of the scanned and segmented scenes.