Automatic segmentation of cell nuclei in Feulgen-stained histological sections of prostate cancer and quantitative evaluation of segmentation results

Digital image analysis of cell nuclei is useful to obtain quantitative information for the diagnosis and prognosis of cancer. However, the lack of a reliable automatic nuclear segmentation is a limiting factor for high-throughput nuclear image analysis. We have developed a method for automatic segmentation of nuclei in Feulgen-stained histological sections of prostate cancer. A local adaptive thresholding with an object perimeter gradient verification step detected the nuclei and was combined with an active contour model that featured an optimized initialization and worked within a restricted region to improve convergence of the segmentation of each nucleus. The method was tested on 30 randomly selected image frames from three cases, comparing the results from the automatic algorithm to a manual delineation of 924 nuclei. The automatic method segmented a few more nuclei compared to the manual method, and about 73% of the manually segmented nuclei were also segmented by the automatic method. For each nucleus segmented both manually and automatically, the accuracy (i.e., agreement with manual delineation) was estimated. The mean segmentation sensitivity/specificity were 95%/96%. The results from the automatic method were not significantly different from the ground truth provided by manual segmentation. This opens the possibility for large-scale nuclear analysis based on automatic segmentation of nuclei in Feulgen-stained histological sections.     

Neural network-based segmentation and classification system for automated grading of histologic sections of bladder carcinoma

OBJECTIVE: To develop an image analysis system for automated nuclear segmentation and classification of histologic bladder sections employing quantitative nuclear features. STUDY DESIGN: Ninety-two cases were classified into three classes by experienced pathologists according to the WHO grading system: 18 cases as grade 1, 45 as grade 2, and 29 as grade 3. Nuclear segmentation was performed by means of an artificial neural network (ANN)-based pixel classification algorithm, and each case was represented by 36 nuclei features. Automated grading of bladder tumor histologic sections was performed by an ANN classifier implemented in a two-stage hierarchic tree. RESULTS: On average, 95% of the nuclei were correctly detected. At the first stage of the hierarchic tree, classifier performance in discriminating between cases of grade 1 and 2 and cases of grade 3 was 89%. At the second stage, 79% of grade 1 cases were correctly distinguished from grade 2 cases. CONCLUSION: The proposed image analysis system provides the means to reduce subjectivity in grading bladder tumors and may contribute to more accurate diagnosis and prognosis since it relies on nuclear features, the value of which has been confirmed.     

Automated cell nuclear segmentation in color images of hematoxylin and eosin-stained breast biopsy

OBJECTIVE: To develop an automated, reproducible epithelial cell nuclear segmentation method to quantify cytologic features quickly and accurately from breast biopsy. STUDY DESIGN: The method, based on fuzzy c-mean clustering of the hue-band of color images and the watershed transform, was applied to 39 images from 3 histologic types (typical hyperplasia, atypical hyperplasia, and ductal carcinoma in situ [cribriform and solid]). RESULTS: The performance of the segmentation algorithm was evaluated by visually determining the percentage of badly segmented nuclei (approximately 25% for all types), the percentage of nuclei that remained in clumps (4.5-16.7%) and the percentage of missed nuclei (0.4-1.5%) for each image. CONCLUSION: The segmentation algorithm was sensitive in that a small percentage of nuclei were missed. However, the percentage of badly segmented nuclei was on the order of 25%, and the percentage of nuclei that remained in clumps was on the order of 10% of the total number of nuclei in the duct. Even so, > 600 nuclei per duct, on average, were segmented correctly; that was a sufficient number by which to calculate accurate quantitative, cytologic, morphometric measurements of epithelial cell nuclei in stained tissue sections of breast biopsy.     

Quantitative microscopy of human DNA topoisomerase II-alpha expression in transitional cell carcinoma of the bladder

OBJECTIVE: To compare the nuclear size of various grades of transitional cell carcinoma (TCC) stained immunohistochemically with the nuclear enzyme topoisomerase II-alpha (topo II-alpha) in bladder urothelial neoplasms. STUDY DESIGN: Histologic sections from 53 consecutive papillary bladder neoplasms were stained immunohistochemically for topo II-alpha expression. There were 18 (33.9%) urothelial neoplasms of low malignant potential (UNLMP), 18 (33.9%) low grade urothelial carcinoma (LGUCa), and 17 (32%) with high grade urothelial carcinoma (HGUCa). The histologic slides were photographed at 400 x magnification and then projected on a screen, and the area with stained nuclei was measured. RESULTS: The cells and nuclei in HGUCa were significantly larger than in LGUCa (P < .05) and UNLMP (P < .01). CONCLUSION: Calculation of the area fraction of nuclei in TCC of the bladder stained with topo II-alpha is an additional method of establishing the grade of these tumors.     

Color categorization and the structure of perceptive color

This paper is an attempt to develop a coherent framework for understanding, simulating, and predicting color categories. The process of color categorization can be understood as a structuring of preceding color experience on the basis of statistical distribution of light in observers environment. A proposed computational model of color categorization includes: 1) distribution of R, G, B pixel values representing a sample of 630 color images of natural scenes (analogue of physical light experience); 2) transformation of the R, G, B pixel values into L*u*v* coordinates of the CIELUV color space (analogue of the process of color perception); 3) distribution of the L*u*v* coordinates representing the sample of the color images (analogue of perceived color experience); 4) k-means clustering algorithm of the L*u*v* coordinates representing the sample of the color images (analogue of the process of color categorization); 5) location and order of color clusters (analogue of location and order of color categories). The proposed computational model enables us to predict the location and order of color categories, being consistent with psycholinguistic data.     

Budburst and leaf area expansion measured with a novel mobile camera system and simple color thresholding

Plant phenology relates strongly to primary productivity and the energy that enters into ecological food webs, and thus is vital in understanding ecosystem function and the effects of climate and climate change. The manual collection of phenological data is labor-intensive and not easily scalable, thus the ability to quantify leaf flush and other parameters at many locations requires innovative new methodologies such as the use of visible light digital cameras. Improved imaging performance was obtained by using a cabled, mobile camera system that allowed a repeated image census of branches of Rhododendron occidentale in the understory along a 30 m transect during leaf flush. Automatic division of acquired images into areas of interest (leaves) and background for calculating leaf area was accomplished by thresholding images in different color spaces. Transformation of the color space into the hue, saturation, and luminance (HSL) color space before thresholding resulted in a mean RMS error of 21.2 cm² compared to hand-counts of leaf area. Thresholding in the native red, green, and blue (RGB) color space to isolate leaves resulted in a larger error, as did using algebraic combinations of the color components or color ratios. Relating physiological function to images, as for sap flow for branches of R. occidentale, indicates that the greening and calculated leaf area of a species as detected by imagers requires additional meteorological sensor data for interpretation.     

Histometric features for the grading of prostatic carcinoma

Histometric features for the objective grading of prostatic adenocarcinoma in histologic specimens were analyzed in five cases each of well, moderately and poorly differentiated lesions. Tissue sections from the selected cases were stained by the Feulgen method and digitized by a video-based microphotometer. Twenty total fields were recorded for each grade: ten at high resolution (an image sampling of 0.5 micron per pixel) and ten at low resolution (0.8 micron per pixel), with two fields per case recorded at each resolution. The images were segmented by an automated expert system-guided scene segmentation procedure. The performance of that procedure was measured by comparing the automated counts of nuclei in the segmented fields to the visual counts made by a pathologist in the same fields. For well, moderately and poorly differentiated cases, respectively, the nuclear counts made by the expert system at high resolution were 2.7%, 4.2% and 4.7% higher than the visual counts (as estimated from a total of 6,628 nuclei), but 1.2%, 2.5% and 1.1% lower at low resolution (10,329 nuclei). High-resolution features and tissue textural features were computed for each case. The high-resolution features showed good separation between the three groups of cases. The tissue textural features showed consistent separation between well and moderately differentiated cases. The relaxation of the spatial resolution (to 0.8 micron/pixel spacing) did not affect the selection of features, but led to less separation between the data from different grades. In conclusion, the automated system performed satisfactorily in distinguishing sections of prostatic tumors of varying degrees of differentiation.     

Color image analysis of cervical neoplasia using RGB computer color specification

OBJECTIVE: To establish quantitative color image analysis for cytology, red, green and blue (RGB) color specification was applied to Papanicolaou-stained cervical smears. STUDY DESIGN: Cell samples used in this study was those from 300 cervical specimens. We analyzed the color tone of nuclei and cytoplasm of the squamous cells in the cervical smear by means of computer image analysis. RESULTS: Papanicolaou stained nuclei displayed basophilic blue to purple. When they were hyperchromatic and deeply stained, B and G values decreased in value. The RGB values of cytoplasm and nuclei decreased significantly (P < .01) as their degree of cellular atypia increased. CONCLUSION: Using RGB color specification to analyze Papanicolaou-stained cervical smears, a significant difference was perceived in the nucleus and cytoplasm between different groups of squamous cells, from normal, dysplastic and squamous cell carcinoma. These findings may help to establish automated cytology.     

Color image analysis for quantifying renal tumor angiogenesis

OBJECTIVE: To segment and quantify microvessels in renal tumor angiogenesis based on a color image analysis method and to improve the accuracy and reproducibility of quantifying microvessel density. STUDY DESIGN: The segmentation task was based on a supervised learning scheme. First, 12 color features (RGB, HSI, I1I2I3 and L*a*b*) were extracted from a training set. The feature selection procedure selected I2L*S features as the best color feature vector. Then we segmented microvessels using the discriminant function made using the minimum error rate classification rule of Bayesian decision theory. In the quantification step, after applying a connected component-labeling algorithm, microvessels with discontinuities were connected and touching microvessels separated. We tested the proposed method on 23 images. RESULTS: The results were evaluated by comparing them with manual quantification of the same images. The comparison revealed that our computerized microvessel counting correlated highly with manual counting by an expert (r = 0.95754). The association between the number of microvessels after the initial segmentation and manual quantification was also assessed using Pearson's correlation coefficient (r = 0.71187). The results indicate that our method is better than conventional computerized image analysis methods. CONCLUSION: Our method correlated highly with quantification by an expert and could become a way to improve the accuracy, feasibility and reproducibility of quantifying microvessel density. We anticipate that it will become a useful diagnostic tool for angiogenesis studies.     

Image processing for accurate cell recognition and count on histologic slides

OBJECTIVE: To design an automatic system for recognition and count of two different cell families on histologic slides. STUDY DESIGN: The segmentation strategy uses color information on the image. The morphologic operations and Support Vector Machine approaches are used for each color to obtain precise segmentation of the image into separate cells for recognition. RESULTS: A large set of histologic slides of bone marrow was assessed byour system and the results compared to the score of a human expert. The results are in good agreement. The difference is within acceptable limits (below 10%). CONCLUSION: The automatic system of cell recognition and extraction is accurate and provides a useful tool for cell recognition and count on histologic slides.     

Mathematical morphologic segmentation dedicated to quantitative immunohistochemistry

OBJECTIVE: To develop automatic segmentation sequences for fully automated quantitative immunohistochemistry of cancer cell nuclei by image analysis. STUDY DESIGN: The study focused on the automated delineation of cancer cell lobules and nuclei, taking breast carcinoma as an example. A hierarchic segmentation was developed, employing mainly the chaining of mathematical morphology operators. The proposed sequence was tested on 22 images of various situations, collected from 18 different cases of breast carcinoma. A quality control procedure was applied, comparing the automated method with manual outlining of cancer cell foci and with manual pricking of cancer cell nuclei. RESULTS: Good concordance was found between automated and manual segmentation procedures (90% for cancer cell clumps, 97% for cancer cell nuclei on average), but the rate of false positive nuclei (small regions labeled as nuclei by the segmentation procedure) could be relatively high (11% on average, with a maximum of 35%) and can result in underestimation of the immunostaining ratio. CONCLUSION: This study examined a preliminary approach to automated immunoquantification, limited to automated segmentation without any color characterization. The automated hierarchic segmentation presented here leads to good discrimination of cancer cell nuclei at the chosen magnification.     

Fully automated measurements by light microscopy of tissue sections using a cellular array computer

Software was developed for the acquisition, segmentation and analysis of microscopic OD-images on a VICOM digital image processor, extended with a VISIOMORPH morphoprocessor board. The delineation algorithms for peroxisomes, lysosomes, and nuclei in liver, kidney, and adrenal gland sections start by thresholding the difference between the original image and a low pass filtered version. The resulting binary mask is then processed by morphological operations in order to produce an object overlay. The efficiency of the programs is evaluated by comparing delineated objects at different OD-levels, created by varying the stain or by multiplying the original pixel values with constant factors. Manual delineation on some images is also used as a reference. More complex algorithms are used for the delineation of muscle fibres in ATP-ase-stained sections and immunocytochemically labelled cells in monolayer preparations. Muscle images from parallel sections with different stainings are matched with a coordinate transform, enabling the transfer of the object mask from a single delineated image to the unprocessed images and thus obtain all necessary information for fibre classification. After segmentation, the OD-images and their object overlays are fed into a data extraction program, measuring for each delineated object user-selected features. Data are sent to a VAX for statistical interpretation.     

Three-dimensional image processing for morphometric analysis of epithelium sections.

The reproducible classification of poorly differentiated abnormal epithelium specimens is still a diagnostic problem. The computer-aided method described here improves the differentiation between benign and malignant epithelium specimens. Hematoxylin and eosin-stained sections of normal squamous epithelium, dysplasia, carcinoma in situ, and carcinoma were scanned in a TV microscope system and analyzed by means of image processing methods on a DEC 5000/200 workstation. From the 15-20 microns thick histological sections, 3-5 focus positions in steps of 1-4 microns were scanned. The segmentation of the cell nuclei was performed automatically by color analysis and geometric operations. For each nucleus the best focus level was selected and at this level the center of the cell was calculated. Graph theoretical methods were applied to analyze the morphometry of the epithelium specimens. The minimal spanning tree was computed in the three-dimensional (3D) space of the sections with the selected centers of the nuclei as vertices. The best feature found for discrimination of the specimens is the average length of all edges in a tree. In the two-dimensional (2D) analysis we had to accept an error probability of about 20% in differentiation of dysplasia and carcinoma. In contrast to this we differentiated normal squamous epithelium, dysplasia, and carcinoma with a correct classification rate of 100% in the 3D analysis.     

Technical note: quantitative measures of iris color using high resolution photographs

Our understanding of the genetic architecture of iris color is still limited. This is partly related to difficulties associated with obtaining quantitative measurements of eye color. Here we introduce a new automated method for measuring iris color using high resolution photographs. This method extracts color measurements in the CIE 1976 L*a*b* (CIELAB) color space from a 256 by 256 pixel square sampled from the 9:00 meridian of the iris. Color is defined across three dimensions: L* (the lightness coordinate), a* (the red-green coordinate), and b* (the blue-yellow coordinate). We applied this method to a sample of individuals of diverse ancestry (East Asian, European and South Asian) that was genotyped for the HERC2 rs12913832 polymorphism, which is strongly associated with blue eye color. We identified substantial variation in the CIELAB color space, not only in the European sample, but also in the East Asian and South Asian samples. As expected, rs12913832 was significantly associated with quantitative iris color measurements in subjects of European ancestry. However, this SNP was also strongly associated with iris color in the South Asian sample, although there were no participants with blue irides in this sample. The usefulness of this method is not restricted only to the study of iris pigmentation. High-resolution pictures of the iris will also make it possible to study the genetic variation involved in iris textural patterns, which show substantial heritability in human populations.     

Classification of lung carcinoma by means of digital nuclear image analysis

An investigation was performed of the maximum discriminating efficiency for each subgroup of digital nuclear image features and of the overall classification of nuclei from three types of human lung carcinomas in histologic sections: adenocarcinoma, small-cell carcinoma and squamous-cell carcinoma. The results indicate that, for each subgroup of features, the nuclei of the small-cell carcinomas are generally "correctly" classified in a higher percentage (80% to 100%) than are the nuclei of the adenocarcinomas (46% to 74%) and squamous-cell carcinomas (29% to 68%). The discriminant analysis for the overall classification selected features from most of the subgroups, suggesting that it is useful to perform nuclear image analysis with many subgroups having different properties. The overall classifications for the nuclei of the adenocarcinomas, small-cell carcinomas and squamous-cell carcinomas were, respectively, 81.4%, 93.2% and 74.7%. Before this technique can be applied to histopathologic diagnosis, a larger number of unselected lung carcinomas must be evaluated.     

Colorieetric Tests of Nuclear Staining

The materials of these investigations were the mucosal nuclei of formalin-paraffin sections (stained by hematoxylin and eosin) from the human uterus in carcinoma, non-malignant hyperplasia and in health. By a scale made of water color pigments placed near the magnified images of nuclei projected on a screen, it was demonstrated mathematically that, for a given specimen, nuclear staining intensity varies inversely with nuclear volume. The most probable explanation is the greater separation of chromatic particles in the larger nuclei.     

Colorieetric Tests of Nuclear Staining

The materials of these investigations were the mucosal nuclei of formalin-paraffin sections (stained by hematoxylin and eosin) from the human uterus in carcinoma, non-malignant hyperplasia and in health. By a scale made of water color pigments placed near the magnified images of nuclei projected on a screen, it was demonstrated mathematically that, for a given specimen, nuclear staining intensity varies inversely with nuclear volume. The most probable explanation is the greater separation of chromatic particles in the larger nuclei.