Utility of the TracCell system in mapping Papanicolaou-stained cytologic material.

BACKGROUND: Much attention has been directed toward commercial application of automation technology to support both quality and productivity enhancement in cervical cytology screening. The introduction of a fully automated precision microscopy workstation, the AcCell Series 2000, (Accu-Med International, Inc., Chicago, Illinois, U.S.A.) has made it possible to effectively incorporate automated prescreening systems that support the human screener. SYSTEM DESIGN: The TracCell 2000 slide mapping system (AccuMed) is a fully automated, stand-alone prescreening device for Papani-colaou-stained cytologic material, such as cervical cytology specimens. The system locates the areas of the slide that do not require review and maps material in the remaining areas; it finds the optimal focal plane, calculates a preferred routing path for human review and determines speed adjustment relative to material density variation. Through slide bar coding identification, this information is transmitted to the AcCell precision microscopy workstation for presentation of the specimen to the human screener. DISCUSSION: By making it possible to automatically guide the human screener to the material of interest, the TracCell system eliminates the need to review empty space and irrelevant areas of the slide. This has the immediate advantage of improved productivity. In addition, by reducing the need to review neutral background material, the system increases the signal-to-noise ratio, thus contributing to a sustained level of vigilance in the screener. The system also addresses operator fatigue through computer-assisted focus and speed variation. This instrument is for investigational use only. The performance characteristics of the device have not been established.

Cytometric validation of immunocytochemical observations in developing lung cancer.

Video-enhanced optical microscopy objectively confirms that immunocytochemical biomarker supplementation of routine cytology permits separation of (1) "pre"-cancer epithelial cells from JHLP screened individuals who subsequently developed lung cancer from (2) similarly atypical epithelial cells from JHLP participants who did not. The establishment of a bank of sputum cytology specimens along with a data respository of 5-yr follow-up and clinical outcome had permitted banked specimens to be immunostained and scored against a gold standard of subsequent histologic lung cancer. A highly skilled cytopathologist, interpreting immunostained screening sputum specimens and blinded to the clinical outcome, originally was able to correctly predict clinical outcome with 88.7% accuracy compared to this "gold standard" (Tockman et al., J Clin Oncol 1988;6:1685-1693). This study presents an independent confirmation of that earlier immunocytochemical classification through feature extraction of digitally recorded, transmission optical microscope video images of immunostained, moderately atypical sputum cells from the original slides.

Considerations in bringing a cancer biomarker to clinical application.

Specific challenges face our application of emerging biomarkers to early lung cancer detection. These challenges might be considered frontiers to be bridged between established biomedical disciplines, requiring expertise often beyond the range of individual investigators. Cross-disciplinary research already has led to new appreciation of the mechanisms which underlie the phenotypic expression of the transformed cell and places within our grasp the tools which might lead to successful early lung cancer detection. Prior to the successful application of newly described markers, further cross-disciplinary research must (a) refine the selection of biologically appropriate markers, (b) validate such markers against acknowledged disease end points, (c) establish quantitative criteria for marker presence/absence, and (d) confirm marker predictive value in prospective population trials.

A quantitative method for the detection and localization of quantum-limited events from radionuclides in cells and tissue sections by computer-enhanced video microscopy.

Cellular dynamics often involve extremely low concentrations of biologically active substances, which can be radiolabeled and detected, localized and quantitated by autoradiography. The latter may require exposures from a few days to many months. The objective of this research was to demonstrate the feasibility of reducing this long period of data collection by one to two orders of magnitude, while maintaining or improving the spatial resolution and localization in tissues and the quantitative characteristics inherent in autoradiography. A mathematical model describing the complete system was generated using energy partition calculations to estimate photon production via scintillant per H3 beta particle emission and to estimate the subsequent photon capture based upon imaging system parameters and microscope geometry. Calculations showed that, typically, a single tritium beta particle produces a maximum of 5.8 X 10(3) photons. A photon-limited camera and microscope imaging system were selected and optimized in conjunction with a specially developed physical scintillation model. Results showed that the number of detected photoevents increases monotonically with both signal integration time and, independently, with the concentration of the radionuclide. Consequently, this work demonstrates that video microscopy imaging methods can spatially and temporally quantify very low concentrations of radiolabeled substances and can reduce data acquisition times.

Computer-aided microscopy in medicine: introduction to the 15 August 1987 issue of Applied Optics.

New research and development opportunities and advances are evident in the multidisciplinary field of computer-aided microscopy for cell and tissue imaging and analysis. This growth applies to both the development of new and improved scientific methods as well as to the utilization of these techniques to investigate significant biomedical applications. Much of the core and supporting technology in this work involves optical and visual processes. Consequently, progress may be facilitated by stimulating the interaction between the optical scientists/engineers and the applications-oriented investigators. This 15 Aug. 1987 issue of Applied Optics presents a heterogeneous collection of papers in three sections: Scanning microscopy and related optical methodsfor cell and tissue imaging; Image cytometry/histometry systems, methods, and standards; and Biomedical applications of image cytometry and histometry. Since the purposes of this issue are to present a representative international cross-section of current research results and to provide direction and assistance to optical scientists and engineers who are interested in becoming more active in this field, this Introduction also documents relevant peer-reviewed journals and sponsoring societies, recommended didactic reading, and scheduled professional events.

Strategies for automated placement of cells for microscopy.

Computer-software-controlled placement of cells onto microscope slides is advantageous for quantitative optical microscopy. Cell detection, indentification, and placement have been implemented on a software-controlled flow sorter. Computer hardware timing latencies that comprise the dead times of the system are sufficiently short (10 micros) to allow adequate dealy of cells for sorting. Operating system timing latencies (150 micros) are also short enough for low-rate sorting. As a result of the low cost of computers and because formats for cell placement are still controversial, software control is desirable somce it provides the flexibility to implement multiple variations for cell detection, identification, amd sorting.

Centrifugal separation of cells in sputum specimens from patients with undifferentiated carcinoma.

Ethanol-fixed cells in sputum from patients with undifferentiated carcinoma of the lung were separated in aqueous Ficoll using a discontinuous density gradient centrifugation technique. The selective enrichment of small cell undifferentiated (e.g., oat cell) or large cell undifferentiated carcinoma cells was achieved while removing most of the leukocytes (80-90%) and macrophages (65-75%) from specimen fractions containing the greatest relative frequencies of cancer cells. The maximum purity of small cell carcinoma cells (0.04%) occurs in moderate density (rho = 1.121 g/ml) gradient fractions and results in a 2.4-fold enrichment relative to unprocessed specimens. In contrast, the maximum purity of large cell carcinoma cells (0.22%) is obtained in very high density (rho = 1.172 g/ml) gradient fractions and results in a 1.2-fold enrichment in comparison with unprocessed specimens. Microscopic examination of Papanicolaou-stained specimen fractions reveals that these enrichments were achieved while retaining diagnostically significant cytomorphologic and tinctorial features necessary for cancer screening and diagnosis. Peak purity ranges of undifferentiated cancer cells significantly overlap comparable ranges for material from bronchogenic adenocarcinoma and squamous cell carcinoma.

Sputum cytopathology: use and potential in monitoring the workplace environment by screening for biological effects of exposure.

Sputum cytopathologic monitoring detects squamous cell lung cancers at an extremely early stage (x-ray negative). It holds further potential for preventing disease by detecting epithelial alterations which reflect environmental hazards. The addition of sputum cytology screening to screening by chest x-ray film does not significantly reduce mortality from all types of lung cancer, but preliminary analysis of Johns Hopkins Lung Project data suggests that mortality from squamous cell carcinoma is reduced. Quantitative automated cytopathology systems and biochemical/immunological cell markers enhance understanding of these precursors and offer great promise for increasing capacity, accuracy, and usefulness in cytopathology screening of workers. Cytological specimens collected over years of screening workers considered at risk may be important to eventually understanding development and prevention of major occupational diseases.

A quantitation of myelin-associated glycoprotein and myelin basic protein loss in different demyelinating diseases.

The loss of myelin-associated glycoprotein (MAG) and myelin basic protein (MBP) was compared by quantitative immunocytochemistry in demyelinating lesions of measles encephalomyelitis (ME), multiple sclerosis (MS), and progressive multifocal leukoencephalopathy (PML). Serial sections from paraffin-embedded tissue were reacted with antisera for MAG and MBP, and areas of staining loss were compared morphometrically. Lesions in ME showed MAG loss equal to that of MBP, lesions of PML showed MAG loss greater than that of MBP, and MS lesions showed a mixture of patterns. These data demonstrate distinctive patterns of MAG and MBP loss in these three diseases.

Automatic cell identification and enrichment in lung cancer: V. Adenocarcinoma and large cell undifferentiated carcinoma.

The aims of this study were to develop a protocol for the identification and enrichment of cancer cells from sputum obtained from patients with adenocarcinoma of the lung (n = 6) and large-cell undifferentiated carcinoma of the lung (n = 2), and to compare these findings with the results from our previous studies on other cell types from lung cancer. The hypotheses tested were: Cancer cells in sputum can be preserved following flow sorting. Enrichment for cancer cells from acridine orange (AO)-stained specimens can be achieved. Discrimination of cancer cells from noncancer cells is by AO green fluorescence and discrimination of lymphocytes from other cell types is by AO red fluorescence. Cancer cells are consistently enriched in the AO high green and red fluorescence region, although, for a given cell type, maximal enrichment is patient-dependent. Finally, cancer cell enrichment and lymphocyte exclusion can be done simultaneously. Cells from sputum were initially fixed, stained with AO, sorted on a dual parameter flow sorter, and classified into six groups corresponding to two ranges of green and three ranges of red fluorescence intensities. Cells of each region were stained by the method of Papanicolaou and differential counts were performed to determine the relative frequencies (i.e., purities) of leukocytes, macrophages, squamous cells, and cancer cells, in sorted and unsorted (i.e., control) samples. The average purity of leukocytes (81%), macrophages (6%), squamous cells (11%), and cancer cells (2%) varied markedly from sample to sample. However, the largest enrichment values (i.e., ratio of purity of a cell type in a sorted sample to its purity in the unsorted control sample) achieved for cancer cells consistently occurred for each patient sample in the region corresponding to high green and high red fluorescence intensities. Experimentally, a cancer cell average enrichment of sixteen-fold was obtained by this method. Additionally, fluorescence intensity ranges which increased the enrichment for macrophages by cell sorting typically excluded leukocytes and squamous cells, and vice versa. Finally, red fluorescence intensity was the primary discriminatory parameter for all cell types studied, although the additional use of green fluorescence intensity significantly increased cancer cell enrichment rates.(ABSTRACT TRUNCATED AT 400 WORDS)

Measles encephalomyelitis: lack of evidence of viral invasion of the central nervous system and quantitative study of the nature of demyelination.

Measles encephalomyelitis appears to be an immune-mediated parainfectious disorder, but it is unclear whether viral invasion of brain is an obligate step in its development. Immunocytochemical methods were used to search for virus antigen in formalin-fixed, paraffin-embedded central nervous system (CNS) tissues from 10 patients with measles encephalomyelitis and 12 patients who had died of measles without CNS involvement. All the CNS tissues studied were viral antigen negative. Similarly fixed CNS tissues from all of 6 patients with subacute sclerosing panencephalitis were viral antigen positive and served as controls. The pattern of perivenular demyelination was also determined in 4 cases of measles encephalomyelitis using antibodies to myelin associated glycoprotein and myelin basic protein and a Luxol fast blue stain. Areas of demyelination in serial sections were quantitated, and no morphometrical differences were found among tissues stained with the three stains. The data suggest the lack of virus replication in the CNS during encephalomyelitis or fatal measles without CNS symptoms. The pattern of loss of myelin associated glycoprotein and myelin basic protein in regions of perivenular demyelination resembles that reported in experimental allergic encephalomyelitis. This pattern of demyelination has been proposed to result from a primary attack on the myelin sheath rather than from direct involvement of the oligodendroglial cell.

Centrifugal separation of cells in sputum specimens from patients with adenocarcinoma.

A Ficoll density gradient centrifugation technique was used to separate fixed cells from the sputa of patients with bronchogenic adenocarcinoma. The selective enrichment of leukocytes, squamous or cancerous cells was achieved while retaining the diagnostically significant cell morphologic features. This technique is shown to typically eliminate most of the leukocytes (96.9%) macrophages (86.7%), squamous cells (70.0%) and necrotic debris from specimen fractions containing the majority of cancerous cells. The maximum purity of cancer cells (14.6%) occurs in high density (p = 1.35 g/ml) gradient fractions and results in an average 12.2-fold enrichment for cancer cells in contrast to their relative frequency of occurrence in unprocessed specimens. Differential cell count analyses of serial density centrifugation fractions show that this technique produces comparable enrichment rates for material from bronchogenic adenocarcinoma and squamous cell carcinoma.

Microphotometric differentiation of human T and B cells tagged with monospecific immunoadsorbent beads.

A comparative and statistical study was done of the classification of peripheral blood lymphocytes (PBLs) as T cells, B cells and monocytes by various immunologic procedures and computerized microphotometric analysis. PBLs from 15 healthy male subjects were examined by immunofluorescence, E rosettes, and immunoadsorbent beads (IAB) for T cells and B cells, by phase contrast microscopy and as fixed slide preparations. Cells tagged with IAB for T cells and B cells were fixed, stained with Papanicolaou stain and analyzed. Evaluation of immunologic data shows 50% to 57% T cells, 9.7% to 24.1% B cells, 13.4% to 16.1% monocytes and about 20% unmarked cells. Analysis of T cells shows significant correlations between E and T cell-IAB rosettes, but neither rosetting procedure reveals a positive correlation with immunofluorescence-labelled T cells. Comparison of B cells shows an insignificant correlation between B-cell IAB and immunofluorescence. Results showed that permanent slides of rosetted cells can be made without alteration in relative numbers of rosetted cells. Assessment of immunologically tagged cell samples by image analysis correctly classified 80% to 90% of cells as T and B cells. Evaluation of homogeneity of the T and B cell populations shows the existence of four subsets of B cells and five subsets of T cells.

A comparative cytopathologic study of noninvasive and invasive squamous cell carcinoma of the lung.

Atypical cells and tissue fragments from the sputum of patients with early and advanced stages of squamous cell carcinoma of the bronchus are objectively characterized and quantitatively compared in this paper. Four classes of single-cell features of cytoplasm, nucleus, nucleolus and nuclear-cytoplasmic ratio are analyzed as a function of cell size and tumor development stage. Distinct differences in the cellular patterns are observed which may enhance cytologic discrimination between noninvasive and invasive stages of squamous cell carcinoma of the bronchus. Initial results justify the application of more sensitive measurement techniques (i.e., automated cytology) to an enlarged data base.

Automatic cell identification and enrichment in lung cancer. I. Light scatter and fluorescence parameters.

Two physical parameters were investigated to automatically recognize cells in sputum from human squamous cell carcinoma of the lung and to separate them for preparation by the Papanicolaou methods, for human interactive identification and for automated high resolution image analysis. The two parameters, 0.5-15.0 degrees forward argon-ion laser light scatter to estimate total cell size and 546 nm Acridine orange fluorescence to approximate total cell DNA content, were measured in a flow-through fluorescence activated cell sorting system. Enrichment for neoplastic cells in three cases of squamous cell carcinoma of the lung averaged 7.8-fold over the original sputum when only green fluorescence was used and 10.5-fold using green fluorescence and forward light scatter. The average enrichment for neoplastic cells was 65.6-fold relative to polymorphonuclear deenrichment.

Automatic cell identification and enrichment in lung cancer. II. Acridine orange for cell sorting of sputum.

Fluorescence spectra were obtained from cells from sputum and pleural effusions stained with different fluorescent dyes and fixed by alternate methods. The spectra were referenced to a standard allowing for fluorescence comparisons of unstained and stained cells under various conditions. The metachromasia of acridine orange-stained cells offers nuclear/cytoplasmic differentiation in a single stain; mithramycin and propidium iodide do not. Unstained cells have an appreciable amount of green (546 nm) fluorescence, as does Carbowax in Saccomanno's preservative. Cytoplasm stained with acidine orange also has appreciable green fluorescence. Consequently, cells with much cytoplasm have high total fluorescence. Cytoplasmic fluorescence is negligible with mithramycin or propidium iodide. The metachromasia of acridine orange-stained cells is altered by alcohol and Carbowax levels in fixatives, keeping other factors constant.

Automatic cell identification and enrichment in lung cancer. III. Light scatter and two fluorescence parameters.

Two fluorescence parameters and size are used in a flow through system to enrich sputum specimens for cancer cells. Human cells in sputum which are stained with acridine orange show a characteristic distribution of red and green fluorescence from which cancer cells can be localized. The peak enrichment is obtained by selectively sorting cells with the largest values of red and green fluorescence. Cancer cells located in other distribution regions having smaller fluorescence intensities show progressively diminished nuclear and cytoplasmic tinctorial features by Papanicolaou stain, consistent with the decreased intensity of red and green fluorescence.

Centrifugal separation of cells in sputum.

A centrifugation technique was developed and used to separate fixed cells from the sputa of patients with bronchogenic squamous cell carcinoma and ones with no evidence of cancer. This article presents the relative frequencies of occurrence of five cell types (i.e., leukocytes, macrophages, squamous, columnar and atypical/cancer) in specimen fractions separated from a discontinuous aqueous Ficoll density gradient. These differential counts show that individual cell types may be selectively collected. Atypical and cancer cells are found at high-density gradient fractions (p congruent to 1.138-1.155 g/ml) with a 10-fold enrichment over unprocessed samples.

Markovian analysis of cervical cell images.

Markovian analysis is a method to measure optical texture based on gray-level transition probabilities in digitized images. Experiments are described that investigate that classification performance of parameters generated by Markovian analysis. Results using Markov texture parameters show that the selection of a Markov step size strongly affects classification error rates and the number of parameters required to achieve the maximum correct classification rates. Markov texture parameters are shown to achieve high rates of correct classification in discriminating images of normal from abnormal cervical cell nuclei.