Summary statement on quantitative cytochemistry (DNA and molecular biology): Task Force 8.

OBJECTIVE: To reach consensus on the application of quantitative cytochemical analysis of chromosomal and DNA aneuploidy in cervical cytopathology. CONCLUSION: The current Pap test has limited specificity to predict cancer and its truly progressive pre-malignant lesions. Infection with human papillomavirus may trigger genetic instability, hyperproliferation and immortalization of the cervical mucosa and cause cervical cancer. Several related molecular markers have been shown to be informative about this neoplastic process. Quantitative analysis of chromosomal and DNA aneuploidy has been shown to be an important tool for identifying (progression to) high grade squamous intraepithelial lesions. A high degree of standardization (material handling, calibration and quality control, measurement and interpretation of results) is required for accurate and reproducible measurements. Areas for further study are presented.

Nuclear texture features for classifying benign vs. dysplastic or malignant squamous epithelium of the larynx.

OBJECTIVE: To search for nuclear features and feature combinations able to assess malignancy and premalignant changes on tissue sections of laryngeal squamous epithelium. STUDY DESIGN: A total of 139 lesions of benign changes (BC) (n = 44), epithelial dysplasias (ED) (n = 50) and invasive laryngeal cancer (LC) (n = 45) were retrieved from archival pathology specimens. The goal of this study was to identify the best features or feature combinations that discriminate BC from LC and also reflect the degree of ED. In order to verify the results on independent data, the groups were split into two separate subgroups, one for training and one for testing. RESULTS: On the test set of slides, the overall correct classification of BC vs. LC cases was 82% using only one feature, fractal2_area. This classification rate could be increased to 91% when a discriminant function based on 10 features was used. However, this gain was not significant. CONCLUSION: Fractal texture features can be used to assess malignancy on tissue sections as an alternative to DNA measurement. In this study feature combinations did not significantly improve classification rates.

Computer-assisted image analysis-derived intermediate endpoints.

The development of prostatic lesions undergoes a slow progression. To establish efficacy of chemopreventive intervention it is therefore necessary to define surrogate endpoint biomarkers. Such biomarkers should be sensitive in their ability to indicate response. They should be objective, ie, the result of measurement, and numerically defined so that a statistical validation of response is possible. They should be able to indicate not only a halt of progression of a lesion, but also a reversal of progression. The spatial and statistical distribution of nuclear chromatin in the secretory and luminal cells in prostatic intraepithelial neoplastic lesions has been shown to be well defined. It can be represented by a set of features. These have been used to define a progression curve along which progression or regression of a lesion can be assessed. One could define a fixed endpoint, or one might choose to accept a statistically significant regression along the progression curve as criterion for chemopreventive efficacy. Expected difficulties could arise from lesion heterogeneity, as it would affect the sampling, and from multifocal lesions of differing progressions. Lesion heterogeneity thus limits the precision with which regression could be detected. These problems might be partially overcome by observations taken in histologically normal appearing regions of the prostate. The nuclear chromatin pattern of secretory cell nuclei measured in such tissue regions from prostates harboring intraepithelial or malignant lesions has been shown to exhibit distinctive changes from the chromatin pattern seen in secretory cell nuclei from prostates free from any such lesions. These changes appear to be expressed in the tissue up to a substantial distance from a lesion. The expression of changes in the nuclear chromatin suggests the existence of an intraepithelial preneoplastic lesion that can be detected by biomarkers, but which is not apparent from visual microscopic inspection. Since chemoprevention might be expected to be most effective at the earliest stages of lesion development, the assessment of such early alterations is seen as highly relevant to efforts to validate the efficacy of chemopreventive intervention.

Specific changes of chromatin structure in nuclei of normal epithelium adjacent to laryngeal squamous cell carcinoma. A preliminary study of 82 cases.

The aim of this study was to confirm the existence of specific nuclear texture feature alterations of histologically normal epithelial borders nearby invasive laryngeal cancer (NC). Paraffin sections of NC and of chronic inflammations unrelated to cancer (CI) were analysed for nuclear texture and for integrated optical density (IOD-index) and were compared to normal epithelium of patients without evidence of cancer (NE). Several discriminant functions based on nuclear texture features were trained to separate different subgroups. As the most important result, specific nuclear texture feature shifts were only found in NC with high-density lymphocytic stroma infiltrate (NC+). Classification of nuclei of NE versus NC+ was correct in 70%. The same classifier was correct in only 58% when nuclei of NE were classified versus CI. We also found lower values of IOD-Index within the NC+ group when compared to NE (p < 0.001).

Detection and localization of early lung cancer by fluorescence bronchoscopy.

BACKGROUND: Curative therapy is available for patients with Stage 0 lung carcinoma, with a >90% 5-year survival rate. Promising chemopreventive agents also are under investigation currently to reduce the risk of lung carcinoma in high risk populations. However, preinvasive bronchial lesions (moderate to severe dysplasia and carcinoma in situ) are very small and thin. They are difficult to localize by conventional white-light bronchoscopy. Fluorescence bronchoscopy is a new diagnostic tool for the detection of these preinvasive lesions. METHODS: The data on the use of fluorescence bronchoscopy to detect and localize preinvasive lesions in current heavy smokers and in former smokers at the British Columbia Cancer Agency as well as the worldwide experience cited in MEDLINE, Index Medicus, and Deutsches Institut fur Medizinische Dokumentation und Information (Cologne, Germany) comparing white-light and fluorescence bronchoscopy using the lung imaging fluorescence endoscope (LIFE)-Lung device (Xillix Technologies Corp., Richmond, British Columbia, Canada) were reviewed. RESULTS: Among current heavy smokers and former smokers with sputum atypia, the prevalence of carcinoma in situ was 1.6%. Moderate or severe dysplasia was found in another 19%. The preinvasive lesions were found to be small: 55% measured < or = 1.5 mm in greatest dimension. Over 1000 cases have been reported in the literature between 1994 and 1999. Overall, 40% of the preinvasive lesions were detected by white-light bronchoscopy alone. The addition of fluorescence bronchoscopy increased the detection rate to an average of 80%. CONCLUSIONS: Preinvasive lesions, especially dysplastic lesions, are small. They are difficult to detect and localize by white-light bronchoscopy. Fluorescence bronchoscopy improves the detection rate. It is an important part of the armamentarium in the overall management of early lung cancer.

An automated image cytometry system for monitoring DNA ploidy and other cell features of radiotherapy and chemotherapy patients.

DNA content and distribution in cell nuclei were studied in samples of fine-needle aspiration (FNA) from 27 locally advanced breast and head and neck cancers in two going randomized trials that compared accelerated fractionation to standard fractionation radiation in locally advanced breast cancer and head and neck cancer. Two image cytometry methods were compared: a new, fully automated DNA image cytometry system (AIC) and a conventional image cytometry (CIC) system with manual selection, focusing, and segmentation of cells. The results of both techniques were compared on the basis of DNA histogram parameters including DNA index (DI), mean DNA values (MDV), and Auer's DNA histogram patterns. An excellent correlation was achieved between the two imaging techniques in terms of DI (r=0.985, p<0.001) and MDV (r=0.951, p<0.001) as well as between Auer's histogram patterns, where both methods agreed completely. It was concluded in these analyses that the two image cytometry methods were equivalent. However, the AIC offered an advantage by scanning samples in a fully automated way, which represented significant time saving for cytopathologists working with the system, as well as a larger number of cells used in the automated analysis. With the automated image cytometer, 500 relevant cells were collected and analyzed in about 10 minutes, where with the interactive (manual) method, it took typically an hour to collect and analyze only about 250 cells. Seventeen samples were sufficient for flow analysis. Image cytometry and flow cytometry showed good agreement in DI determination; however, three cases reported as diploid by flow cytometry were found to be aneuploid by image cytometry techniques.

Fluorescent location of lung tumour cells.

Cells were collected on glass slides by touching tumour surfaces (A) and normal regions (B) of the lung. The slides were stained with a nuclear stain and a fluorescent probe for a tumour associated cell surface protein. The (B) slides from the normal regions lacked fluorescent epithelial cells. The tumour slides (A) contained typical tumour cells and dyskaryotic cells which exhibited cell surface fluorescence.

The dynamics of laser-induced changes in human skin autofluorescence--experimental measurements and theoretical modeling.

To study the temporal dynamics of human skin autofluorescence photobleaching, we measured the autofluorescence spectral changes of skin in vivo during continuous exposure to 442 nm (He-Cd) laser light. Integral intensities were calculated for various spectral wavelength bands and plotted as a function of time. Mathematical analysis of the time function revealed a double-exponential photobleaching process: I(t) = a exp (-t/tau 1) + b exp(-t/tau 2) + c, in which tau 1 and tau 2 differed by an order of magnitude. A hypothesis for the mechanism of the double-exponential photobleaching dynamics was proposed and evaluated using Monte Carlo modeling of light propagation in the skin and autofluorescence escape from skin. By combining the fluorophore microdistributions, Monte Carlo simulation results and the variation in fluorescence decrease parameters (a, b, c, tau 1, tau 2) with increasing exposure intensities a biophysical explanation for the double-exponential photobleaching function was elucidated. The fast decrease term corresponds to laser-induced photobleaching in the stratum corneum, while the slow decrease term represents fluorophore changes in the dermis. The measured autofluorescence photobleaching dynamics can be used to determine the fractional contributions of different skin layers to the total autofluorescence signal measured in vivo.

Automated fluorescence microscopic measurement of apoptosis frequency following ionizing radiation exposure in cultured mammalian cells.

PURPOSE: To develop and assess an automated image cytometric method of apoptotic cell classification for use under conditions in which apoptosis is a rare event (e.g. fibroblastoid cell lines or low-dose irradiation). METHOD: Image acquisition software was adapted to gather double-stained cell images from slides prepared using cell fixation and staining methods that emphasized apoptotic morphology. Chinese hamster ovary cells (CHO) were classified individually by discriminant analysis of morphological and nuclear texture features calculated for each image. Discriminant functions were constructed from a manually classified set of over 60000 cell images categorized as 'normal', 'apoptotic', 'cell doublets' or 'debris' and all subsequent cell images collected were classified using these functions. RESULTS: Application of this technique resulted in a 99.8% accuracy in classification of the normal cell population, and 81.7% classification accuracy for apoptotic cells. This method was then applied to study the time course of the apoptotic response of CHO cells following X-irradiation. Following irradiation with 5 Gy no increase above control levels of apoptosis was noted until 18 h post-irradiation, which corresponded with the release of the G2 block as determined by DNA-content analysis. Apoptotic frequency increased to a peak level of 12.1 +/- 4.6% at 42 h post-irradiation. CONCLUSIONS: Automated image cytometry provides an efficient and consistent method of apoptosis measurement. This study represents the first detailed characterization of the time course and the role of cell division in CHO cell apoptosis.

Sputum cells from lung tumor patients carry a cell surface marker not found in normal sputum.

Cells were collected from sites of known lung tumours and corresponding control areas of these lungs. Fluorescent staining demonstrated that the tumour cells and epithelial cells (cytologically these cells appeared normal) both possessed a receptor for these fluorescent probes. Fluorescent labelling of sputum cells from these tumour patients also resulted in fluorescent labelling of these "cyto logically normal" epithelial cells. No such fluorescent epithelial cells were observed in sputum samples collected from control subjects or in cells collected from the control areas of the tumour patients' lungs. We conclude that a cell surface protein receptor is expressed in lung tumour-associated epithelial cells but is absent from control sputum epithelial cells.

Malignancy associated changes in bronchial epithelial cells and clinical application as a biomarker.

A total of 74 bronchial brushing specimens, 24 from patients with advanced stage cancer, eight from patients with CIS, 31 from patients with atypical metaplasia and 11 from normal subjects were examined for the existence of malignancy associated changes (MAC). Conventional fiberoptic bronchoscopy and fluorescence endoscopy was carried out on every case. Each case was classified according to the highest grade of abnormality diagnosed by bronchial biopsy of the suspect areas. During the endoscopy examination, a bronchial brushing specimen was obtained from a visually normal area very remote from the abnormal area as possible such as the opposite lung or another lobe. The bronchial brushing specimens were fixed, mounted and stained by a DNA specific method and approximately 1500 images of individual nuclei per case were captured by an automated high resolution image cytometry. For each of these images, more than 100 nuclear features such as size, shape and chromatin spatial organization were calculated. Discriminant function analysis revealed nuclear features which differentiated between normal bronchial cell nuclei from the normal subjects and ostensively normal nuclei (MAC cell nuclei) from the lung cancer patients. The best discrimination was achieved when the frequency of individual cells expressing MAC was 50% or greater. With this threshold, 75% of the patients with invasive cancer and CIS were correctly classified. Fifty percent of those with severe or moderate atypia and 35% with mild atypia were also MAC positive. The frequency of cells expressing MAC also increased as the degree of abnormality of the groups increased. MAC may be a useful criterium to determine biological behavior of the intra-epithelial (pre-invasive) neoplasia.

DNA-cytometry of progressive and regressive cervical intraepithelial neoplasia.

A retrospective analysis was performed on archival cervical smears from a group of 56 women with cervical intraepithelial neoplasia (CIN), who had received follow-up by cytology only. Automated image cytometry of Feulgen-stained DNA was used to determine the differences between progressive and regressive lesions. The first group of 30 smears was from women who had developed cancer after initial smears with dysplastic changes (progressive group). The second group of 26 smears with dysplastic changes had shown regression to normal (regressive group). The goal of the study was to determine if differences in cytometric features existed between the progressive and regressive groups. CIN categories I, II and III were represented in both groups, and measurements were pooled across diagnostic categories. Images of up to 700 intermediate cells were obtained from each slide, and cells were scanned exhaustively for the detection of diagnostic cells. Discriminant function analysis was performed for both intermediate and diagnostic cells. The most significant differences between the groups were found for diagnostic cells, with a cell classification accuracy of 82%. Intermediate cells could be classified with 60% accuracy. Cytometric features which afforded the best discrimination were characteristic of the chromatin organization in diagnostic cells (nuclear texture). Slide classification was performed by thresholding the number of cells which exhibited progression associated changes (PAC) in chromatin configuration, with an accuracy of 93 and 73% for diagnostic and intermediate cells, respectively. These results indicate that regardless of the extent of nuclear atypia as reflected in the CIN category, features of chromatin organization can potentially be used to predict the malignant or progressive potential of CIN lesions.

The use of an automated image cytometer for screening 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.

Sputum screening by quantitative microscopy: a reexamination of a portion of the National Cancer Institute Cooperative Early Lung Cancer Study.

OBJECTIVE: To investigate the hypothesis that image cytometry of sputum specimens can detect squamous carcinoma without requiring visually abnormal cells. DESIGN: The sensitivity and specificity of image cytometry were evaluated in a case-control study. MATERIAL AND METHODS: Seventy-three sputum slides from the Mayo portion of the National Cancer Institute Cooperative Early Lung Cancer Study were restained by a modified Feulgen method. We examined 40 slides from 9 patients in whom squamous carcinoma developed and 33 slides from 11 patients in whom no cancer developed during a follow-up of at least 5 years. Images of normal epithelial nuclei were collected by using an automated image cytometer. Discriminant analysis was used to determine differences in DNA distribution of normal nuclei in sputum specimens from noncancer patients versus normal nuclei in sputum samples from patients in whom carcinoma developed. RESULTS: By using features based on DNA distribution, 74% correct classification of nuclei was possible without human review of the material and without the use of visually abnormal nuclei. A receiver operating characteristic curve demonstrated sensitivities and specificities, including 40% sensitivity and 90% specificity. CONCLUSION: Although this study was limited to 20-year-old slides and squamous cell carcinoma, automated image cytometry detected a substantial proportion of patients with squamous cell cancer without using visually abnormal nuclei.

Reconstruction of in vivo skin autofluorescence spectrum from microscopic properties by Monte Carlo simulation.

The in vivo skin autofluorescence spectrum was reconstructed by Monte Carlo simulation using microscopic fluorophore distributions and intrinsic fluorescence spectra measured from excised skin tissue sections as well as employing published skin tissue optical parameters. The theoretical modeling took into account the light-tissue interactions of scattering, absorption, and regeneration of fluorescence photons. The modification of the intrinsic spectra by tissue optical properties to generate the in vivo spectrum observed at the tissue surface can be represented by a fluorescence detection efficiency function (eta) which equals the integral of the product of the excitation light distribution inside the tissue and the fluorescence escape efficiency. Comparison of the reconstructed in vivo spectrum with the measured spectra showed good agreement, outside of the blood absorption bands, suggesting that (i) the theoretical modeling, (ii) the skin optical parameters used, and (iii) the measured microscopic morphology and spectral data are consistent. The divergence which exists over the strong blood absorption wavelength band (530-600 nm) suggests that the effect of blood contents on in vivo tissue optical properties deserves further investigations.

Opening the black box: the relationship between neural networks and linear discriminant functions.

Over the last ten years feed-forward neural networks have become a popular tool for statistical decision making. During this time, they have been applied in many fields, including cytological classification. Neural networks are often treated as a black box, whose inner workings are concealed from the researcher. This is unfortunate, since the inner workings of a neural network can be understood in a manner similar to that of a linear discriminant function, which is the standard tool that researchers use for decision making. This paper discusses feed-forward neural networks and some methods to improve their performance for classification problems. Their relationship to discriminant functions will be examined for a simple two-dimensional classification problem.

Detection of malignancy associated changes in cervical cell nuclei using feed-forward neural networks.

Normal cells in the presence of a precancerous lesion undergo subtle changes of their DNA distribution when observed by visible microscopy. These changes have been termed Malignancy Associated Changes (MACs). Using statistical models such as neural networks and discriminant functions it is possible to design classifiers that can separate these objects from truly normal cells. The correct classification rate using feed-forward neural networks is compared to linear discriminant analysis when applied to detecting MACs. Classifiers were designed using 53 nuclear features calculated from images for each of 25,360 normal appearing cells taken from 344 slides diagnosed as normal or containing severe dysplasia. A linear discriminant function achieved a correct classification rate of 61.6% on the test data while neural networks scored as high as 72.5% on a cell-by-cell basis. The cell classifiers were applied to a library of 93,494 cells from 395 slides, and the results were jackknifed using a single slide feature. The discriminant function achieved a correct classification rate of 67.6% while the neural networks managed as high as 76.2%.

A simple clinical method for the preparation of improved cervical smears-approximating to monolayers.

Conventionally prepared cervical smears contain multilayers of cells deposited within strands of mucin. The present study is concerned with chemical reduction of disulphide bonds in the mucin leading to depolymerisation prior to forming a smear in the conventional manner. The resultant distribution of cells on the slide is similar to that obtained by machines designed to produce monolayers of cells. These monolayers have been developed for use in automated analysis of cervical smears and sputum samples. This new technique does not interfere with conventional PAP analysis of dyskaryotic cells nor does it interfere with the fluorescent location of such cells of cytological interest.

Molecular follow-up of a preinvasive bronchial lesion treated by 13-cis-retinoic acid.

We report the result of the follow-up molecular analysis of a bronchial carcinoma in situ treated by 13-cis-retinoic acid, which relapsed 9 months after cessation of drug therapy. Loss of heterozygosity at 3p21 and 9p22 genomic sequences were assessed by polymerase chain reaction (PCR) after microdissection of the dysplastic epithelia. Despite a transient regression of the lesion to a lower grade with treatment, molecular analysis showed the persistence of a 3p and 9p deletion in all the bronchial biopsies taken in the same area during a 1 year follow-up, preceding by 9 months the recurrence of the carcinoma in situ. Our findings suggest that molecular follow-up analysis can help to assess the persistence of a malignant clone within a bronchial epithelium that displays a more benign phenotype under retinoid treatment on follow-up. Molecular analysis may be of great importance to evaluate the effects of chemoprevention and to determine the duration of such intervention in responder patients.