Segmentation of cervical cell images.

A major problem in the automation of cervical cytology screening is the segmentation of cell images. This paper describes various standard segmentation methods plus one which determines a segmentation threshold based on the stability of the perimeter of the cell as the threshold is varied. As well as contour, certain structural information is used to decide upon the threshold which separates cytoplasm from the background. Once the cytoplasm threshold is found, cytoplasm and nucleus are separated by simple clustering into three groups, cytoplasm, folded cytoplasm and nucleus. These techniques have been tested on 1500 cervical cells that belong to one of eight normal classes and five abnormal classes. A minimum Mahalanobis distance classifier was used to compare results. Manually thresholded cells were classified correctly 66.0% of the time for the 13 class problem and 95.2% of the time on the two (normal-abnormal) class problem. The contour tracing technique was 52.9% and 90.0% correct, respectively.

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.

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.

Does interstitial lung edema compress airways and arteries? A morphometric study.

We compared areas and diameters of small airways and arteries in three groups of anesthetized dogs: 1) control (n = 5), 2) hydrostatic edema induced by fluid overload (n = 13), and 3) increased permeability edema induced with alpha-naphthylthiourea (n = 5). We measured pulmonary arterial and wedge pressures in all groups and cardiac output in the hydrostatic edema group. Postmortem, lobes were frozen at functional residual capacity and samples taken for measurements of extravascular lung water (Qwl/dQl) and for light microscopy. We also examined lobes from hydrostatic edema experiments fixed at transpulmonary pressures of 5 and 27 cmH2O. From the histology slides, bronchovascular bundles with respiratory bronchioles (n = 706) and bronchioles (n = 467) were photographed and airway and vessel areas and diameters measured. Alveolar and airway luminal edema were graded. We found that only in hydrostatic edema, pulmonary arterial and wedge pressures increased and vascular resistance fell with fluid infusion. Mean Qwl/dQl values were 3.80 +/- 0.17, 6.81 +/- 0.96, and 9.34 +/- 0.62 (SE) in control, hydrostatic, and increased permeability edema groups, respectively. By quantitative histology, airway and arterial areas and diameters did not decrease in edema and rose with increasing transpulmonary pressure. Variable quantities of air-space edema were seen. We conclude that interstitial edema does not compress small airways or arteries and that other mechanisms, including alveolar and airway luminal edema, may explain reported increases in airway resistance.

Altered immunoregulatory function in long-term renal allograft recipients.

In order to determine immunoregulatory lymphocyte subsets in patients with long surviving renal allografts, a study utilising functionally characterised monoclonal antibodies with analysis by flow cytometry using a Fluorescence Activated Cell Sorter was carried out. Cells from 35 patients with allograft survival from one to 15 years were analysed with monoclonal antibodies for the following markers: inducer-helper, cytotoxic-suppressor, monocytes, Ia, IgG and IgM. Control groups consisted of long-term dialysis patients and a group of normal individuals. The results show a strikingly significant difference between the long-term allograft recipients and the control population in terms of the inversion of the normal inducer-helper to cytotoxic-suppressor ratio. Additionally, there were significant differences in the number of Ia+ cells in long-term transplant patients and dialysis patients compared with normals. No impressive differences were found in the numbers of Ig bearing cells. Thus the long-term allograft state is a clear example of an overabundance of cellular suppression and may explain many of the general phenomena seen in surviving recipients.

Distribution of fluid in bronchovascular bundles with permeability lung edema induced by alpha-naphthylthiourea in dogs. A morphometric study.

We utilized light microscopic morphometry to examine the distribution of fluid in bronchovascular bundles of different sizes. Permeability edema was induced in 10 anesthetized dogs with 27 mg/kg of alpha-naphthylthiourea. Eight dogs served as controls. After moderately severe edema, diagnosed on chest radiographs and with decreasing arterial pO2, lobes were fixed with glutaraldehyde and formaldehyde or by freeze substitution. Postmortem wet weight to dry weight ratios were 7.82 +/- 0.62 (mean +/- SE) in the edematous lungs and 4.38 +/- 0.25 in the controls. Bronchovascular bundles were photographed and grouped as follows: bundles composed of separated arteries and bronchioles, bundles with connected arteries and bronchioles, and bundles with connected arteries and bronchi. The transparencies were projected on a tablet interfaced to a computer and the following areas were determined: T, the total bundle area; V, the vessel (artery) area; B, the airway (bronchiole or bronchus) area; A1, the tight periarterial adventitial sheath area; A2, the loose periarterial interstitial area; and A3, the bronchiolar/bronchial interstitial area. In addition, edema ratios for arteries (A2/V) and airways (A3/B) were calculated. We found that (a) A1 was very small and did not change with edema; (b) A2 in all bundles increased 10-fold with edema (p less than 0.01), whereas A3 increased 2- to 3-fold; (c) A2/V increased 9- to 15-fold in the edematous bundles (p less than 0.01) and (d) A3/B did not change in separated bundles (p greater than 0.05) but was approximately double after edema in the connected bundles with bronchioles and bronchi (p less than 0.01). We conclude that edema in bronchovascular bundles accumulates preferentially in the loose periarterial interstitium and does not appear to accumulate around smaller bronchioles. These data may be explained by anatomical factors and by gradients of interstitial pressure.

Morphometry of the distribution of hydrostatic pulmonary oedema in dogs.

Light microscopic morphometry was utilized to examine the distribution of fluid in the interstitium around arteries, veins and within bronchovascular bundles in hydrostatic oedema, comparing it with previous control and permeability oedema experiments. Pulmonary artery wedge pressure was raised with fluid overload and an aortic balloon in five anaesthetized dogs to produce oedema (wet weight to dry weight ratios of 11.66 +/- 0.84). Lung lobes were fixed by freeze-substitution at 20 mmHg airway pressure. Photomicrographs of arteries, veins and bronchovascular bundles were taken, and areas were digitized to obtain the following: for arteries and veins, an oedema ratio=perivascular oedema cuff area/vessel area; for bronchovascular bundles, T=total bundle area, A1=interstitial area around airways, B=airway (respiratory bronchiole, bronchiole, or bronchus) area, A2=periarterial interstitium, V=artery area. From these, oedema ratios were calculated as A1/B and A2/V. We found that the oedema ratios were greater (P less than 0.01) for arteries (1.18, n=675) than veins (0.56, n=263), and were greater for the larger vessels; A1 rose significantly (P less than 0.01) only in bronchovascular bundles with bronchioles and bronchi, not in those with respiratory bronchioles; A2 increased from three- to 25-fold (P less than 0.01) in all bundles; A1/B only increased in bundles with bronchi while A2/V increased two- to six-fold in all bundles with oedema compared with controls. We conclude that these preferential patterns of distribution resemble those reported in permeability oedema, and may shed light on mechanisms of accumulation, and on the physiological effects of oedema on airways and vessels of the lung.

Mineral dusts in lungs with scar or scar cancer.

Five lungs with small scars and five lungs with small scar associated cancers, were studied by light and scanning electron microscopy and x-ray energy dispersive spectrometry. Six hundred particles were photographed and their physical and chemical properties analyzed from scar, cancer, or normal alveolar tissue on carbon planchet-mounted, deparaffinized and low temperature-ashed sections. Amosite/crocidolite fibers were accumulated only in one cancerous lung. All other lungs shared similar types of mineral particles. The lungs with noncancer scars, however, showed an increase in the ratio of aluminum and calcium salts (non-silicates), while the lungs with scar cancers had a higher ratio of silicates. These patterns of particle distribution were similar in different areas of the same lung, despite the fact mechanism is unclear, these results are consistent with the possibility that the pattern of mineral particle distribution in a lung may influence the formation of cancer in a scar.

Evaluation of single-cell classification schemes for computer classification of cervical cells.

Three single-cell classification schemes were evaluated and compared with the performance of a cytotechnologist in classifying single cells from routinely prepared cervical smears. All of the single-cell classification schemes were found to approach the performance of the cytotechnologist in distinguishing normal squamous cells from significantly dysplastic or malignant squamous cells. For distinguishing non-squamous cell types from dysplastic or malignant squamous cells, however, all three schemes fell far short of the cytotechnologist's performance in classifying the same cells. Current research aimed at overcoming these difficulties is also described.

Quantitative morphology of permeability lung edema in dogs induced by alpha-naphthylthiourea.

In pulmonary edema, fluid accumulates first in the interstitium, then in the alveoli. However, the relative amounts of interstitial fluid around arteries and veins of different sizes are unknown; in addition, the effects of fixation on the light microscopic quantitation of edema are unclear. To answer these questions, we induced permeability pulmonary edema in seven anesthetized dogs with 27 mg/kg of alpha-naphthylthiourea. Pulmonary artery and wedge pressures were measured. After moderately severe edema, diagnosed by chest x-ray and falling arterial pO2, lobes were fixed by airways instillation or vascular perfusion with glutaraldehyde and formaldehyde or were frozen with liquid nitrogen. With light microscopy, the edema surrounding arteries and veins of different sizes was measured using a computer equipped with a digitizing tablet and expressed as the edema ratio = area of perivascular edema/area of vessel, or as an absolute area of edema. Alveolar edema was graded semiquantitatively, and wet weight to dry weight ratios were calculated. Two control dogs were also studied. During the induction of edema, pulmonary artery and wedge pressures did not change significantly. Mean wet weight to dry weight ratios were 9.3 +/- 1.1. We found that the edema ratio was greater (p less than 0.01) for arteries (2.75, n = 1305) than for veins (1.40, n = 900). The edema ratio was greater for vessels more than 400 micron than less than 400 micron (p less than 0.01) and greater in the instillation- and perfusion-fixed lobes than in the frozen lobes (p less than 0.01). Similar results were obtained for the absolute areas of periarterial and perivenous edema. Less alveolar edema was seen in the lobes fixed by instillation (p less than 0.01). We conclude that, in permeability edema induced by alpha-naphthylthiourea, the fluid accumulates preferentially around arteries compared with veins and around larger compared with smaller vessels. Airways instillation and vascular perfusion fixation appeared to increase interstitial fluid cuffs compared to freezing.