High resolution analysis of cervical cells--a progress report.

This paper presents preliminary results of research toward the development of a high resolution analysis stage for a dual resolution image processing-based prescreening device for cervical cytology. Experiments using both manual and automatic methods for cell segmentation are described. In both cases, 1500 cervical cells were analyzed and classified as normal or abnormal (dysplastic or malignant) using a minimum Mahalanobis distance classifier with eight subclasses of normal cells, and five subclasses of abnormal cells. With manual segmentation, false positive and false negative error rates of 2.98 and 7.73% were obtained. Similar experiments using automatic cell segmentation methods yielded false positive and false negative error rates of 3.90 and 11.56%, respectively. In both cases, independent training and testing data were used.     

Flow microfluorometric system for screening gynecologic cytology specimens using propidium iodide-fluorescein isothiocyanate.

Seventy cervical cytology specimens have been screened by a xero resolution flow analyzer-sorter using propidium iodide and fluorescein isothiocyanate as fluorochromes for nucleus and cytoplasm, respectively. This system shows a 1% sensitivity for detection of abnormal cells using only crude visual data analysis. Screening of clinical specimens was performed on the instrument with a 5.8% false negative rate and a 11.8% false positive rate by comparison with routine visual cytologic evaluation of the same samples.     

Report on a long-term trial of CYBEST Model 2 for prescreening for squamous cell carcinoma of the uterine cervix

The results of a 12 year field trial, on a large patient base, of the CYBEST Model 2, an automated screening system using an image analysis method, are summarized. The cell specimens were stained by a conventional Papanicolaou method. Individual cellular atypism was determined from the sum of the nuclear area, the N/C ratio, the mean nuclear optical density, and the nuclear roundness with an ambiguity function value of 0.4295. CYBEST's final assessment as 'suspicious' or 'normal' was statistically determined from a cumulative histogram of the cellular atypism of a maximum 300 detected cells for each cell population using the Kolmogrov-Smirnov test, and those of unsatisfactory samples were automatically classed as 'rejected', which occurred in 6% of the study. A total of 84 atypical preparations including 17 histologically proven carcinoma patients were encountered during the entire test period from 1977 to 1988, and the overall false negative rate was 1.19% (1/84). Among the results on a total of 42,988 slides during the test period of the last 9 years from 1980 to 1988, the false positives occurred at a rate of 30.7% (12,383/40,307 of non-dysplastic slides) and the false negatives at approximately 2% (1/55 of dysplastic slides). These results are compared with those of other important systems.     

Large performance variation does not affect outcome in the Finnish cervical cancer screening programme

S. Lönnberg, P. Nieminen, L. Kotaniemi‐Talonen, H. Kujari, J. Melkko, G. Granroth, M. Vornanen, T. Pietiläinen, J. Arola, J. Tarkkanen, T. Luostarinen and A. Anttila Large performance variation does not affect outcome in the Finnish cervical cancer screening programme Objective: Cytology screening for prevention of cervical cancer can reduce incidence and mortality by more than 80% in settings with good organization and rigorous quality control. Audit studies are essential for reaching and maintaining a high quality of screening. The aim of this study was to evaluate variation in performance indicators by screening laboratory and assess the impact on the effectiveness of screening as indicated by cervical intraepithelial neoplasia grade 3 and above (CIN3+) rates after a negative screen. Methods: Seven cytology screening laboratories operating during 1990–1999 with a total of 953 610 screening tests performed were included in the study. By linking screening and cancer register files, all cases of CIN3+ diagnosed in the screened population were identified. For 395 CIN3+ cases with a preceding negative screen and 787 controls, a re‐evaluation of smears was undertaken to uncover false negative screening tests. Performance parameters and rates of CIN3+ after a negative screen were analysed for interlaboratory heterogeneity. Results: The rates of follow‐up recommendations and referrals varied by up to 3.6‐ (2.8–10.2%) and 4.0‐fold (0.03–0.12%), respectively. CIN1, CIN2 and CIN3+ screen detection rates differed by up to 8.5‐ (0.02–0.17%), 5.4‐ (0.05–0.25%) and 3.3‐fold (0.05–0.18%). False negative rates determined by re‐evaluation showed up to 2.1‐fold differences (29–62%). Rates of CIN3+ after a negative screen (0.023–0.048%) and as a proportion of total CIN3+ (15–31%) in the screened population were low and did not vary significantly. Conclusions: There were large variations in the sensitivity–specificity trade‐off between laboratories, reflected in all performance indicators as well as in the test validity estimates of the re‐evaluation phase, but not in screening effectiveness. Even though performance variations do not always have an impact on the effectiveness of screening, they lead to variations in cost, treatment and psychological burden, and should be addressed.     

Estimating false positive and false negative error rates in cervical cell classification.

The performance of a cell recognition system on unknown data is often estimated in terms of its error rates on a test set. This paper investigates methods for producing estimates of error rates in cervical cell classification. Classification performance curves calculated using these methods are given for several classification schemes used to classify 1500 cervical cells.     

The confirmation rate of primary hits: a predictive model

HTS data from primary screening are usually analyzed by setting a cutoff for activity, in order to minimize both false-negative and false-positive rates. An alternative approach, based on a calculated probability of being active, is presented here. Given the predicted confirmation rate derived from this probability, the number of primary positives selected for follow-up can be optimized to maximize the number of true positives without picking too many false positives. Typical cutoff-determining methods are more serendipitous in their nature and not easily optimized in an effort to optimize screening efforts. An additional advantage of calculating a probability of being active for each compound screened is that orthogonal mixtures can be deconvoluted without presetting a deconvolution threshold. An important consequence of using the probability of being active with orthogonal mixtures is that individual compound screening results can be recorded irrespective of whether the assays were performed on single compounds or on cocktails.     

Diagnostic schemes for fine needle aspirates of breast masses

A comparison was made of four statistically based schemes for classifying epithelial cells from 243 fine needle aspirates of breast masses as benign or malignant. Two schemes were computer-generated decision trees and two were user generated. Eleven cytologic characteristics described in the literature as being useful in distinguishing benign from malignant breast aspirates were assessed on a scale of 1 to 10, with 1 being closest to that described as benign and 10 to that described as malignant. The original computer-generated dichotomous decision tree gave 6 false negatives and 12 false positives on the data set; another tree generated from the current data improved performance slightly, with 5 false negatives and 10 false positives. Maximum diagnostic overlap occurred at the cut-point of the original dichotomous tree. The insertion of a third node evaluating additional parameters resulted in one false negative and seven false positives. This performance was matched by summing the scores of the eight characteristics that individually were most effective in separating benign from malignant. We conclude that, while statistically designed, computer-generated dichotomous decision trees identify a starting sequence for applying cytologic characteristics to distinguish between benign and malignant breast aspirates, modifications based on human expert knowledge may result in schemes that improve diagnostic performance.     

The effects of alignment error and alignment filtering on the sitewise detection of positive selection

When detecting positive selection in proteins, the prevalence of errors resulting from misalignment and the ability of alignment filters to mitigate such errors are not well understood, but filters are commonly applied to try to avoid false positive results. Focusing on the sitewise detection of positive selection across a wide range of divergence levels and indel rates, we performed simulation experiments to quantify the false positives and false negatives introduced by alignment error and the ability of alignment filters to improve performance. We found that some aligners led to many false positives, whereas others resulted in very few. False negatives were a problem for all aligners, increasing with sequence divergence. Of the aligners tested, PRANK's codon-based alignments consistently performed the best and ClustalW performed the worst. Of the filters tested, GUIDANCE performed the best and Gblocks performed the worst. Although some filters showed good ability to reduce the error rates from ClustalW and MAFFT alignments, none were found to substantially improve the performance of PRANK alignments under most conditions. Our results revealed distinct trends in error rates and power levels for aligners and filters within a biologically plausible parameter space. With the best aligner, a low false positive rate was maintained even with extremely divergent indel-prone sequences. Controls using the true alignment and an optimal filtering method suggested that performance improvements could be gained by improving aligners or filters to reduce the prevalence of false negatives, especially at higher divergence levels and indel rates.     

Hypothesis testing in high-throughput screening for drug discovery

Following the success of small-molecule high-throughput screening (HTS) in drug discovery, other large-scale screening techniques are currently revolutionizing the biological sciences. Powerful new statistical tools have been developed to analyze the vast amounts of data in DNA chip studies, but have not yet found their way into compound screening. In HTS, characterization of single-point hit lists is often done only in retrospect after the results of confirmation experiments are available. However, for prioritization, for optimal use of resources, for quality control, and for comparison of screens it would be extremely valuable to predict the rates of false positives and false negatives directly from the primary screening results. Making full use of the available information about compounds and controls contained in HTS results and replicated pilot runs, the Z score and from it the p value can be estimated for each measurement. Based on this consideration, we have applied the concept of p-value distribution analysis (PVDA), which was originally developed for gene expression studies, to HTS data. PVDA allowed prediction of all relevant error rates as well as the rate of true inactives, and excellent agreement with confirmation experiments was found.     

Geographical patterns in prediction errors of species distribution models

Aim To describe and explain geographical patterns of false absence and false presence prediction errors that occur when describing current plant species ranges with species distribution models. Location Europe. Methods We calibrated species distribution models (generalized linear models) using a set of climatic variables and gridded distribution data for 1065 vascular plant species from the Atlas Florae Europaeae. We used randomly selected subsets for each species with a constant prevalence of 0.5, modelled the distribution 1000 times, calculated weighted averages of the model parameters and used these to predict the current distribution in Europe. Using a threshold of 0.5, we derived presence/absence maps. Comparing observed and modelled species distribution, we calculated the false absence rates, i.e. species wrongly modelled as absent, and the false presence rates, i.e. species wrongly modelled as present, on a 50 × 50 km grid. Subsequently, we related both error rates to species range properties, land use and topographic variability within grid cells by means of simultaneous autoregressive models to correct for spatial autocorrelation. Results Grid‐cell‐specific error rates were not evenly distributed across Europe. The mean false absence rate was 0.16 ± 0.12 (standard deviation) and the mean false presence rate was 0.22 ± 0.13. False absence rates were highest in central Spain, the Alps and parts of south‐eastern Europe, while false presence rates were highest in northern Spain, France, Italy and south‐eastern Europe. False absence rates were high when range edges of species accumulated within a grid cell and when the intensity of human land use was high. False presence rates were positively associated with relative occurrence area and accumulation of range edges. Main conclusions Predictions for various species are not only accompanied by species‐specific but also by grid‐cell‐specific errors. The latter are associated with characteristics of the grid cells but also with range characteristics of occurring species. Uncertainties of predictive species distribution models are not equally distributed in space, and we would recommend accompanying maps of predicted distributions with a graphical representation of predictive performance.     

Goblet-like cells in atrophic vaginal smears and their histologic correlation. Possible confusion with endocervical cells.

The occurrence and origin of goblet-like cells seen between clusters of parabasal cells in atrophic vaginal smears were investigated. The goblet-like cells were cytologically identified in the vaginal smears from 23 (19.2%) of 120 patients whose smears showed an atrophic pattern, but without any inflammatory, dysplastic or malignant changes. Histologically, these cells were found in sections from 6 (18.8%) of 32 elderly women with atrophic vaginal epithelium. The goblet-like cells were situated among the squamous cells of the upper layer of the atrophic squamous epithelium from the vagina to the portio. These goblet-like cells in atrophic smears were initially misinterpreted as endocervical cells, which are regarded as a marker of smear adequacy in the cytologic screening for cancer of the uterine cervix. The correct interpretation of these goblet-like cells in smears from postmenopausal and elderly women is thus obviously important in assessing the adequacy of the sample for the detection of abnormal cells.     

Lower-order effects adjustment in quantitative traits model-based multifactor dimensionality reduction

Identifying gene-gene interactions or gene-environment interactions in studies of human complex diseases remains a big challenge in genetic epidemiology. An additional challenge, often forgotten, is to account for important lower-order genetic effects. These may hamper the identification of genuine epistasis. If lower-order genetic effects contribute to the genetic variance of a trait, identified statistical interactions may simply be due to a signal boost of these effects. In this study, we restrict attention to quantitative traits and bi-allelic SNPs as genetic markers. Moreover, our interaction study focuses on 2-way SNP-SNP interactions. Via simulations, we assess the performance of different corrective measures for lower-order genetic effects in Model-Based Multifactor Dimensionality Reduction epistasis detection, using additive and co-dominant coding schemes. Performance is evaluated in terms of power and familywise error rate. Our simulations indicate that empirical power estimates are reduced with correction of lower-order effects, likewise familywise error rates. Easy-to-use automatic SNP selection procedures, SNP selection based on "top" findings, or SNP selection based on p-value criterion for interesting main effects result in reduced power but also almost zero false positive rates. Always accounting for main effects in the SNP-SNP pair under investigation during Model-Based Multifactor Dimensionality Reduction analysis adequately controls false positive epistasis findings. This is particularly true when adopting a co-dominant corrective coding scheme. In conclusion, automatic search procedures to identify lower-order effects to correct for during epistasis screening should be avoided. The same is true for procedures that adjust for lower-order effects prior to Model-Based Multifactor Dimensionality Reduction and involve using residuals as the new trait. We advocate using "on-the-fly" lower-order effects adjusting when screening for SNP-SNP interactions using Model-Based Multifactor Dimensionality Reduction analysis.     

Maximum-likelihood estimation of allelic dropout and false allele error rates from microsatellite genotypes in the absence of reference data

The importance of quantifying and accounting for stochastic genotyping errors when analyzing microsatellite data is increasingly being recognized. This awareness is motivating the development of data analysis methods that not only take errors into consideration but also recognize the difference between two distinct classes of error, allelic dropout and false alleles. Currently methods to estimate rates of allelic dropout and false alleles depend upon the availability of error-free reference genotypes or reliable pedigree data, which are often not available. We have developed a maximum-likelihood-based method for estimating these error rates from a single replication of a sample of genotypes. Simulations show it to be both accurate and robust to modest violations of its underlying assumptions. We have applied the method to estimating error rates in two microsatellite data sets. It is implemented in a computer program, Pedant, which estimates allelic dropout and false allele error rates with 95% confidence regions from microsatellite genotype data and performs power analysis. Pedant is freely available at http://www.stats.gla.ac.uk/ approximately paulj/pedant.html.     

Field test and experimental use of CYBEST model 2 for practical gynecologic mass screening

CYBEST is an automated cytologic screening system for uterine cancer utilizing a pattern-recognition image-analysis system. The prototype was developed in 1972 following fundamental studies of feature extraction, feature evaluation using ambiguity differential functions and segmentation of cell and nuclear images. Model 2 was developed in 1974 with an improved mechanism and function. The parameters employed are nuclear size, nuclear optical density, N/C ratio and nuclear shape. The data of field tests using 220 samples containing three cases of dysplasia, 110 cases of carcinoma and 107 nonmalignant cases were as follows: two false-negative cases (1.8%), 13 false positives (12.1%) and one reject (0.9%). This system was experimentally tested for practical mass population screening with 1,829 cases including 17 atypical cases (four epidermoid carcinomas). The data were as follows: no false-negative cases and 581 false-positive cases (32.1%). Of the latter, 311 cases (17.2%) were pathologic samples, such as severe cervicitis, senile colpitis, Trichomonas infestation, etc., and the remaining 270 cases (14.9%) were within physiologic limits, corresponding to true false-positive samples.     

The use of an automated image cytometer for screening and quantitative assessment of cervical lesions in the British Columbia Cervical Smear Screening Programme

The use of an automated image cytometer for screeing and quantitative assessment of cervical lesions in the British Columbia Cervical Smear Screening Programme
The development of an automated device to screen cervical cytology slides for the detection of pre-invasive lesions of the cervix has been the goal of many individuals for over 30 years. The increasing sophistication of the technology of automation and increasingly powerful computer technology have enabled a number of these systems to reach the stage at which they have become a practical reality. The Department of Cancer Imaging at the British Columbia Cancer Agency has developed such a device over the past few years. This study reports the preliminary results of a trial to determine the reliability of the device for the screening and quantitative assessment of cervical cells. A training set of over 1000 cervical slides was used to train the image cytometer. A test set of 1030 slides was screened by the image cytometer and in the Cytology Screening Laboratory. At the 50% sample split the sensitivity of the image cytometer was 95% for severe dysplasia and 90% for moderate dysplasia, compared with a sensitivity of 90% for both of these lesions using conventional screening. A combination of nuclear texture features was found which can be used for the quantitative assessment of both abnormal cells and apparently normal intermediate cells.     

Support vector machines in HTS data mining: Type I MetAPs inhibition study

This article reports a successful application of support vector machines (SVMs) in mining high-throughput screening (HTS) data of a type I methionine aminopeptidases (MetAPs) inhibition study. A library with 43,736 small organic molecules was used in the study, and 1355 compounds in the library with 40% or higher inhibition activity were considered as active. The data set was randomly split into a training set and a test set (3:1 ratio). The authors were able to rank compounds in the test set using their decision values predicted by SVM models that were built on the training set. They defined a novel score PT50, the percentage of the test set needed to be screened to recover 50% of the actives, to measure the performance of the models. With carefully selected parameters, SVM models increased the hit rates significantly, and 50% of the active compounds could be recovered by screening just 7% of the test set. The authors found that the size of the training set played a significant role in the performance of the models. A training set with 10,000 member compounds is likely the minimum size required to build a model with reasonable predictive power.     

Automated multiple ligand screening by frontal affinity chromatography-mass spectrometry (FAC-MS)

High-throughput screening (HTS) efforts to discover "hits" typically rely on the large-scale parallel screening of individual compounds with attempts to screen mixtures of compounds typically and, unfortunately, giving rise to false positives and false negatives due to the nature of the HTS readout (% inhibition/activation above a defined threshold) that makes deconvolution virtually intractable. Bioaffinity screening methods have emerged as an alternative or orthogonal method to classic HTS. One of these methods, frontal affinity chromatography coupled to mass spectrometry detection (FAC-MS), although still a relatively new technique, is turning out to be a viable screening tool. However, to push FAC-MS more to the forefront as a moderate primary HTS system (or a secondary screening assay), automation needs to be addressed. An automated FAC-MS system is described using 2 columns containing immobilized hERbeta, whereby while 1 column is being regenerated, the other is being used. The authors are extrapolating that in a continuous 24-h operation, the number of ligands screened could potentially approach 10,000. In addition, preliminary structure-activity relationship binding information (typically not seen in early primary HTS) can be obtained by observing the rank order of the library members in the various mixtures.     

Flow cytometric prescreening of cervical smears.

One-parameter (nuclear DNA) and two-parameter (nuclear DNA and protein or cellular light scatter) measurements of cervical smears were performed using an ICP 11 and a cytofluorograf 4800 respectively. A total of about 1000 cases was analyzed. For the estimation of nuclear DNA alone two fluorochromes were tested (ethidium bromide (EB) and mithramycin (MMC)) combined with three different methods of cell preparation. For the two-parameter measurements cells were double stained with EB and fluorescein isothiocyanate (FITC). Red fluorescence (EB) versus green fluorescence (FITC) or red fluorescence versus scatter were recorded. A computer analysis of the one-parameter histograms was performed using discriminant analysis and the results were compared with the cytodiagnosis of microscopic specimens stained with the Papanicolaou technique. The error rates of the flow cytometric (FCM) data were as follows: (a) standard EB staining, 11% false negative, 26% false positive, 6% unsatisfactory results; (b) pepsination of vital cells and EB staining, 12% false negative, 14% false positive and 4% unsatisfactory results; (c) MMC staining, 10% false negative, 65% false positive and 5% unsatisfactory results. Our two-parameter measurements prove that, as confirmed by cell sorting, red fluorescence versus scatter allows separation of at least three subpopulations in most analyzed samples: (a) anucleated cells; (b) leukocytes; and (c) intermediate and superficial cells.     

Evaluation of HPV Infection and Smoking Status Impacts on Cell Proliferation in Epithelial Layers of Cervical Neoplasia

Accurate cervical intra-epithelial neoplasia (CIN) lesion grading is needed for effective patient management. We applied computer-assisted scanning and analytic approaches to immuno-stained CIN lesion sections to more accurately delineate disease states and decipher cell proliferation impacts from HPV and smoking within individual epithelial layers. A patient cohort undergoing cervical screening was identified (n = 196) and biopsies of varying disease grades and with intact basement membranes and epithelial layers were obtained (n = 261). Specimens were sectioned, stained (Mib1), and scanned using a high-resolution imaging system. We achieved semi-automated delineation of proliferation status and epithelial cell layers using Otsu segmentation, manual image review, Voronoi tessellation, and immuno-staining. Data were interrogated against known status for HPV infection, smoking, and disease grade. We observed increased cell proliferation and decreased epithelial thickness with increased disease grade (when analyzing the epithelium at full thickness). Analysis within individual cell layers showed a ≥50% increase in cell proliferation for CIN2 vs. CIN1 lesions in higher epithelial layers (with minimal differences seen in basal/parabasal layers). Higher rates of proliferation for HPV-positive vs. -negative cases were seen in epithelial layers beyond the basal/parabasal layers in normal and CIN1 tissues. Comparing smokers vs. non-smokers, we observed increased cell proliferation in parabasal (low and high grade lesions) and basal layers (high grade only). In sum, we report CIN grade-specific differences in cell proliferation within individual epithelial layers. We also show HPV and smoking impacts on cell layer-specific proliferation. Our findings yield insight into CIN progression biology and demonstrate that rigorous, semi-automated imaging of histopathological specimens may be applied to improve disease grading accuracy.     

Evaluation of eluents from separations of CD34+ cells from human cord blood using a commercial, immunomagnetic cell separation system

Human CD34+ cells from cord blood were separated in a two-step process using a commercial, immunomagnetic cell retention system. The performance of the system was evaluated by analyzing a number of eluents from the separations with a number of analytical techniques. In addition to cell counts and flow cytometry analysis, a new experimental technique that is undergoing development, cell tracking velocimetry (CTV), was used. CTV measures the degree to which a cell is immunomagnetically labeled, known as the magnetophoretic mobility, of a population of cells on a cell-by-cell basis and presents the results in the form of a histogram similar to flow cytometry data. The average recovery and purity of CD34+ cells from 10 separations was 52% and 60%, respectively. CTV analysis indicated that the mean magnetophoretic mobility of the positively enriched CD34 cells was 9.64 x 10(-5) mm3/T-A-s, while the mean mobility from negative eluents was -2.02 x 10(-6) mm3/T-A-s, very similar to the mobility of unlabeled cells. Within the positive eluents, the range of magnetophoretic mobility was approximately 50-fold, representing a plausible 50-fold range in surface CD34 antigen expression. CTV analysis also indicated that in some separations, positive cells were not retained by the immunomagnetic cell retention system. Finally, preliminary studies indicate that monocytes might be a primary cause in the lower purities and recoveries seen in this study. It is suggested that the monocytes phagocytose the magnetic nanobeads and become sufficiently magnetized to be retained within the Miltenyi column, reducing the purity of the positive eluent.