Strategies for rare-event detection: an approach for automated fetal cell detection in maternal blood.

This article explores the feasibility of the use of automated microscopy and image analysis to detect the presence of rare fetal nucleated red blood cells (NRBCs) circulating in maternal blood. The rationales for enrichment and for automated image analysis for "rare-event" detection are reviewed. We also describe the application of automated image analysis to 42 maternal blood samples, using a protocol consisting of one-step enrichment followed by immunocytochemical staining for fetal hemoglobin (HbF) and FISH for X- and Y-chromosomal sequences. Automated image analysis consisted of multimode microscopy and subsequent visual evaluation of image memories containing the selected objects. The FISH results were compared with the results of conventional karyotyping of the chorionic villi. By use of manual screening, 43% of the slides were found to be positive (>=1 NRBC), with a mean number of 11 NRBCs (range 1-40). By automated microscopy, 52% were positive, with on average 17 NRBCs (range 1-111). There was a good correlation between both manual and automated screening, but the NRBC yield from automated image analysis was found to be superior to that from manual screening (P=.0443), particularly when the NRBC count was >15. Seven (64%) of 11 XY fetuses were correctly diagnosed by FISH analysis of automatically detected cells, and all discrepancies were restricted to the lower cell-count range. We believe that automated microscopy and image analysis reduce the screening workload, are more sensitive than manual evaluation, and can be used to detect rare HbF-containing NRBCs in maternal blood.

Prenatal diagnosis of trisomy 13 on fetal cells obtained from maternal blood after minor enrichment.

In a pilot study to establish fetal nucleated red blood cell (NRBC) detection in maternal blood, trisomy 13 was diagnosed by FISH analysis at 11 weeks' gestation. The NRBCs were detected after a single-step ficoll density gradient enrichment. In blood samples taken both before and after CVS, 52 and 80 NRBCs, respectively, were found to be positive for fetal haemoglobin. In 47 per cent of these cells, FISH analysis for X and Y chromosomes confirmed the fetal sex. Moreover, 48 per cent of these NRBCs showed three fluorescent signals for a chromosome 13 probe, which confirmed the diagnosis of trisomy 13, previously detected at CVS karyotyping. This is the first report of non-invasive prenatal diagnosis of trisomy 13, i.e., pre-CVS, in the first trimester. The high number of fetal NRBCs detected indicates a connection with aneuploidy, probably due to early impairment of the feto-maternal barrier.

Development of a preparation and staining method for fetal erythroblasts in maternal blood: simultaneous immunocytochemical staining and FISH analysis.

In order to detect fetal nucleated red blood cells (NRBCs) in maternal blood, a protocol was developed which aimed at producing a reliable staining method for combined immunocytochemical and FISH analysis. The technique had to be suitable for eventual automated screening of slides. Chorionic villi washings, cord blood, and maternal blood samples were used for this study. After a density gradient separation and centrifugal cytology, slides were stained either with 3,3-diaminobenzidin (DAB), a marker for heme, or with antibodies against the gamma-chain of fetal hemoglobin (HbF). FISH analysis for both X- and Y-chromosomes was performed on the same slides. Cytocentrifugation provided a controlled cell density on the slides with good cell morphology. Both the DAB and HbF staining were suitable for manual screening of large numbers of slides. The HbF staining, although supposed to be more specific for fetal NRBCs, appeared to be more sensitive to minor changes in preparation. We were eventually able to combine HbF staining with FISH analysis, and produced a detection efficiency of >85% for both X- and Y-chromosome signals. This preparation protocol simplifies the detection of NRBCs in maternal blood. Immunocytochemical staining and FISH analysis can be performed on the same cell with good image contrast, thus facilitating both manual and automated image analysis. This will facilitate the use of this approach for prenatal diagnosis.

Fetal cell detection in maternal blood: a study in 236 samples using erythroblast morphology, DAB and HbF staining, and FISH analysis.

A protocol to detect fetal nucleated red blood cells (NRBCs) was tested in 217 pregnant women and in 19 nonpregnant controls. All the pregnant women were sampled after chorionic villus sampling (CVS); 20 were also sampled pre-CVS. NRBC recognition was based upon morphology by using staining of hemoglobin with 3,3-diaminobenzidin (DAB) or by immunocytochemical staining for fetal hemoglobin (HbF). This was combined with FISH analysis for both the X- and Y-chromosomes on the same cells. Progressive refinement of the methods increased the number of cases where NRBCs were detected from 53% (DAB) to 75% and 78% for DAB and HbF staining, respectively, with on average 43 NRBCs (range, 1-220). DAB gave a slightly higher yield than HbF in the lower cell count range (<25). In 6 out of 18 controls, NRBCs were detected with DAB, vs. 1 out of 19 (5%) with HbF. FISH analysis in 41 cases resulted in correct sex prediction in 80% (DAB) and 89% (HbF), respectively. Our data demonstrated an increase of cases with NRBCs (30% to 75%), as well as a rise of the mean number of NRBCs (6 to 29 cells), after CVS. We conclude that DAB staining is a straightforward way to screen for the presence of NRBCs in maternal blood, but is not specific for NRBCs of fetal origin. HbF immunophenotyping is a reliable marker for fetal NRBCs, which detected slightly fewer NRBCs than DAB-staining, but improved sex prediction and significantly reduced false-positive results. CVS at 10-13 weeks of gestation causes a significant increase of NRBCs in maternal blood. These data indicate that further refinement of NRBC detection is needed before application of noninvasive prenatal diagnosis using maternal blood is feasible.

Detection of 'rare event' fetal erythroblasts in maternal blood using automated microscopy.

This paper describes the use of automated microscopy to detect fetal erythroblasts in maternal blood. The technology is based on the following approach: (1) the use of centrifugal cytology for the preparation of monolayers; (2) simultaneous staining of fetal hemoglobin (immunoalkaline phosphatase) and chromosome sequences (FISH); (3) multi-mode microscopy to detect rare events; (4) visual evaluation of image memories containing detected objects. Model systems show that fetal cells in frequencies as low as 1 in a million cells can be detected easily (manually or by automated microscopy). Algorithms for automated cell selection were developed for a test set of 6 patients. Optimization of hardware and software routines will make analysis of several million cells in approximately 1 h feasible.

Automated screening for cytomegalovirus infected cells using image analysis. Comparison of two immunoenzymatic staining methods with respect to colour segmentation.

The detection of human cytomegalovirus (HCMV) infected poly-morphonuclear leukocytes (PMNLs) for early finding of the pp65 antigen using automated image analysis has been improved. The routinely used immunoenzyme peroxidase (PO) labelling has been replaced by alkaline phosphatase (AP) using new fuchsin as substrate. The number of automatically detected false positive objects due to incomplete inactivation of endogenous peroxidase and strong variations in counterstain intensity could be reduced by 81% using this AP staining method. The number of detected truly positive cells with both staining methods was not significantly different. Furthermore, a new image analysis system providing processing of colour images was evaluated. Since plain differences in absorption wavelength are required for colour segmentation, the red immuno-staining was combined with a green counterstain using methyl green. Screening of AP- instead of PO-stained slides in combination with colour segmentation resulted in a further reduction of the number of falsely detected alarms from 61% to 11%. Consequently, the sensitivity of the automated detection was improved. For AP staining detection of cells in frequencies of approximately one to one million was demonstrated. Screening for CMV-positive, alkaline phosphatase labelled cells using an image analysis system with colour segmentation resulted in a reduced false alarm rate, a better visual interpretation of the images and subsequently an increase in the sensitivity of the automated screening process.

Detection of immunocytochemically stained rare events using image analysis.

The detection of rare-event cells circulating in peripheral blood using automated image analysis was evaluated using a model system consisting of cells from a breast cancer cell line (SKBR3) seeded in a mononuclear cell suspension. Slides of cells with optimal morphology were prepared according to an optimized preparation procedure based on centrifugal cytology in combination with formalin fixation. SKBR3 cells were immunocytochemically stained for cytokeratin using the cam 5.2 monoclonal antibody and labelled with alkaline phosphatase using CAS-red as substrate. Because, for optimal segmentation of cell images, plain differences in absorption wavelength are required, the red immunostaining was combined with a green nuclear counter-staining based on ethyl green. Slides were automatically screened for cytokeratin-positive SKBR3 cells resulting in a lowest detectable frequency of one positive cell per 1.87 x 10(6) negative cells. A comparison between manual screening and automated screening for cytokeratin-positive cells showed a high level of correlation (0.9998). For the definition of the total number of objects per slide, two counting procedures were evaluated. Results were close to the visual score with a coefficient of variation of 0.47% for the counting procedure used in this study. It is concluded that optimization of preparation and staining procedures for the detection of rare-event cells using automated image analysis results in optimal image contrast and, consequently, in an increase in sensitivity for detecting rare events.

Prognostic significance of cell DNA content in early-stage ovarian cancer (FIGO stages I and II/A) by means of automatic image cytometry.

Paraffin-embedded material from 69 patients with epithelial ovarian cancer FIGO stages I and II/A (including 21 patients with borderline carcinoma) was studied with automatic DNA image cytometry. Univariate analysis indicated a significant difference in survival based on the presence of nuclei with high DNA content (higher than 5 C). A group of patients with less than 0.2% cells with high DNA content had a 6-year survival of 87%, whereas in a group of patients with more than 0.2% of such cells, 6-year survival was 49%. This parameter remained significant when used in a group of stage I/a and I/b patients. Statistical analysis of diploid vs. non-diploid tumors also showed significant difference in survival. Separate analysis of 48 invasive ovarian cancers indicated that ploidy, the percentage of cells with high DNA content and tumor stage (stage I/a + b vs. stages I/c + II/a) reached significance for survival, whereas grading did not. In addition, comparison of clinical stage, grading, ploidy and the percentage of cells exceeding 5 C with a threshold at 0.2% by means of a multivariate analysis (Cox regression model) showed that only the percentage of cells exceeding 5 C remained statistically significant.

Interobserver variability in the cytological diagnosis of 1500 Papanicolaou stained cervical monolayer specimens.

Cervical specimens from 1500 patients were prepared by means of a centrifugation procedure to obtain monolayer specimens suitable for automated screening using a machine. After staining according to Papanicolaou, each specimen was diagnosed by four independent cytologists from two different institutes. Within each institute, noncorresponding screening results were discussed to arrive at a conclusion diagnosis. After discussion of the discrepancies between the two centers, the conclusion diagnoses were combined to one final cytological diagnosis for each specimen. This final diagnosis is to be used as a reference diagnosis to evaluate machine classification as obtained by the AUTOPLAN/MIAC system. This system is presently being tested both in Leiden and in Frankfurt for its accuracy of detecting abnormal lesions in cervical specimens. The used diagnostic procedure resulted in a negative reference diagnosis for 1217 of the 1500 specimens; 170 specimens were diagnosed CIN I or II (mild or moderate dysplasia) and 113 specimens had a positive reference diagnosis (CIN III or invasive carcinoma). Based on these three diagnostic classes, the agreement between the four independent cytologists and the reference diagnosis varied between 93.60% and 96.60%, whereas 95.33% of all 6000 diagnoses correlated with the reference diagnosis.