| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Molecular Oncology, Markers, Clinical Correlates |
Departments of 1 Cancer Biology and 2 Adult Oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston, Massachusetts; 3 Synta Pharmaceuticals Corp., Lexington, Massachusetts; and 4 Vaytek, Incorporated, Fairfield, Iowa
 |
ABSTRACT |
|---|
 Top ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Experimental Design: We recently demonstrated proof of concept using a fluorescence-based automated microscope system, REIS, for OTC detection from the PB. For this study, the prototype of the system was adopted for high-throughput and high-content cellular analysis.
Results: The performance of the improved REIS was examined using normal blood (n = 10), normal blood added to cancer cells (n = 20), and blood samples obtained from cancer patients (n = 80). Data from the screening of 80 clinical slides from breast and lung cancer patients, by manual microscopy and by the REIS, revealed that as many as 14 of 35 positive slides (40%) were missed by manual screening but positively identified by REIS. In addition, REIS-assisted scanning reliably and reproducibly quantified the total number of cells analyzed in the assay and categorized positive cells based on their marker expression profile.
Conclusions: REIS-assisted analysis provides excellent sensitivity and reproducibility for OTC detection. This approach may enable an improved method for screening of PB samples and for obtaining novel information about disease staging and about risk evaluation in cancer patients.
|
INTRODUCTION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Because tumor cell dissemination into the bone marrow takes place via blood circulation, it should be possible to detect these "cells in transit" in a peripheral blood sample. A simple blood test would, in contrast to bone marrow aspiration, allow for the analysis to be frequently repeated and thus used for staging at diagnosis as well as for therapy monitoring and long-term patient management. Using sensitive molecular and cell-based methods, OTCs have been successfully detected and isolated from blood in various solid malignancies, such as breast, prostate, colorectal, pancreatic, ovarian, and head and neck cancers, as well as malignant melanoma (12, 13, 14, 15, 16, 17, 18) . However, the frequency of positive samples in studies of different cancers varies considerably, ranging between 95% in untreated breast cancer (12) and 12% in ovarian cancer (13) . These variations may reflect the behavior of different tumor types but also illustrate the limited sensitivity of available assays. In a comparison of 15 studies of peripheral blood samples of malignant melanoma patients, positivity rates ranged from 6 to 93% (16) . The frequency of OTCs in peripheral blood specimens of solid malignancies has yet to be determined. Reported numbers of cell-based assays range from as low as 0.2 of 107 in head and neck cancer (17) to as high as 150 of 106 in ovarian cancer (13) .
Identifying and enumerating very rare tumor cells by microscopy is highly laborious, and the accuracy and sensitivity of the analysis is potentially impacted by the fatigue of the reviewer. Automated microscopy systems have been developed to reduce the subjectivity of manual microscopic interpretations, improving the sensitivity and inter-laboratory consistency of the OTC detection. Automated systems are either based on brightfield or fluorescence microscopy.
Brightfield microscopy analyzes slides that are labeled with immunoenzymatic techniques. Such preparations are usually stable for long periods of time (years), and an analysis can be performed on archived material. However, in contrast to immunofluorescent techniques, they can be influenced by endogenous enzyme activity that can cause high background and nonspecific staining. Fluorescently labeled specimens are subject to photobleaching, caused by light illumination and/or oxidation by room air and must therefore be especially protected and analyzed within weeks. This main disadvantage of fluorescence microscopy is probably the reason why most studies of occult tumor cell detection use brightfield microscopy. However, fluorescence-based analysis has one important advantage over brightfield techniques, i.e., the potential of multimarker assessment of rare cancer cells. This capability is of particular importance if one considers the reported heterogeneity of OTCs (19 , 20) .
Two commercial systems based on brightfield microscopy and single marker detection, the ACIS by ChromaVision Medical Systems, Inc. (San Juan Capistrano, CA; Ref. 21 ) and the MDS by Applied Imaging Corp. (Santa Clara, CA; Ref. 22 ) were recently introduced for the detection of OTC in bone marrow. Both systems were reported to provide excellent sensitivity and reproducibility for disseminated tumor cell detection. Two automated microscope systems based on fluorescence imaging, the Laser Scanning Cytometer (Compucyte Corp., Cambridge, MA) and the Metafer 3.0 automatic image analysis system (MetaSystems GmbH, Altlussheim, Germany), have been demonstrated to detect small numbers of tumor cells in spiked blood samples (23 , 24) and in the bone marrow of neuroblastoma patients (25) . Both systems showed a high level of sensitivity, although data on the reproducibility or day-to-day performance of the OTC detection assays were not presented in those studies.
Recently, we reported the development and clinical use of a fluorescence-based rare event detection system, rare event imaging system (REIS), designed for both the detection and further characterization of OTCs (26 , 27) . By introducing new hard- and software, the system was markedly improved for ease of use, faster scanning time with minimal hands-on time, and wider flexibility regarding markers to be detected. As an added feature, we introduced an automated scanning algorithm for samples that are simultaneously labeled with multiple markers. Using our second generation REIS, results are described for the analysis of normal blood added to cancer cells and blood specimens from 80 cancer patients. System performance is characterized by assay specificity, sensitivity, and reproducitility, as well as by comparison to manual microscopy. In separate studies to be published elsewhere, the REIS has been used for the routine analysis of blood samples from patients with small cell lung cancer and breast cancer.5
|
MATERIALS AND METHODS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
For the sensitivity, specificity, and reproducibility testing of automated microscopy, several sets of spiked samples were prepared from normal human blood and breast carcinoma cells (cell lines SK-BR-3 and MCF-7; American Type Culture Collection, Rockville, MD). Blood specimens were added to 0.1–10 carcinoma cells/106 peripheral blood mononuclear cells. For the clinical testing, slides from a large collection of blood samples from small cell lung and breast cancer patients at Dana-Farber Cancer Institute were used. Small cell lung cancer samples were drawn at diagnosis (before treatment) from patients with limited or extended stage disease. Breast cancer samples were collected from patients with metastatic disease.
Immunocytochemical Staining
Blood specimens were immunostained essentially as described previously (26
, 27)
. Briefly, for the single labeling of cytokeratin, cells on slides were fixed in ice-cold methanol for 5 min., rinsed in PBS, and blocked with 20% human AB serum (Nabi Diagnostics, Boca Raton, FL) in PBS at 37°C for 20 min. Slides were then incubated at 37°C for 1 h either with a monoclonal antipan cytokeratin mixture (clones C-11, PCK-26, CY-90, KS-1A3, and A53-B/A2; Sigma, St. Louis, MO), which recognizes human cytokeratins 1, 4, 5, 6, 8, 10, 13, 18, and 19 in immunoblotting, or slides were incubated with rabbit polyclonal anticytokeratin antibodies (Biomedical Technologies, Stoughton, MA) that recognize class I and II cytokeratins. Subsequently, slides were washed in PBS, incubated with AlexaFluor488-conjugated antimouse antibodies (Molecular Probes, Eugene, OR) or Rhodamine-Red-X-conjugated antirabbit antibodies (Jackson ImmunoResearch, West Grove, PA) at 37°C for 30 min, counterstained with 0.5 µg/ml 4',6-diamidino-2-phenylindole (DAPI, Molecular Probes) in PBS at room temperature for 10 min, and mounted in ProLong mounting medium (Molecular Probes). Processed slides were stored at room temperature and analyzed microscopically within a month.
The specimens from cancer patients were stained with polyclonal anticytokeratin antibodies (lung cancer) or monoclonal anticytokeratin antibodies (breast cancer; see results in Fig. 1
and Tables 2
3
4
). The specimens from healthy blood donors were labeled with monoclonal antipan cytokeratin followed by antimouse rhodamine-conjugated antibodies (see results in Table 1
).
|
|
|
|
|
and Fig. 2
). The cells were fixed in 2% paraformaldehyde in PBS (pH 7.4) at room temperature for 20 min., washed in PBS, and blocked with 20% human AB serum in PBS at 37°C for 20 min. Subsequently, primary polyclonal rabbit antibodies directed against HER-2/neu were applied at 37°C for 1 h (DakoCorporation, Glostrup, Denmark), followed by incubation with Rhodamine-Red-X-conjugated antirabbit secondary antibody (Jackson ImmunoResearch) at 37°C for 30 min. Cells were then washed, fixed in ice-cold methanol for 5 min., blocked with 20% human AB-serum, and incubated with the antipan cytokeratin antibody (Sigma) at 37°C for 1 h. Secondary AlexaFluor488-conjugated antibody was applied at 37°C for 30 min., followed by counterstaining of the nuclei with 0.5 µg/ml DAPI in PBS. Doubly labeled cells were mounted in ProLong mounting medium. Slides were stored at 4°C and analyzed microscopically within a week.
|
|
Automated Screening.
All slides were coded and scanned with the improved REIS by independent operators. The basic components of our original REIS (26
, 27)
remained the same, although major hard- and software improvements were introduced. The improved REIS consists of an automated epifluorescence microscope (Nikon Eclipse 1000; Nikon, Tokyo, Japan) with a computer-controlled stage (Ludl Electronic Products, Hawthorne, NY), a feedback stabilized (model 1962; Opti Quip, Highland Mills, NY) 100-W mercury light source, automated objective turret, automated filter cube changer with five positions, automated shutter, a high resolution black and white cooled CCD-camera (Sensicam; Cooke Corp., Auburn Hills, MI), and a personal computer with the image analysis software Image-Pro Plus 4.5 (Media Cybernetics, Carlsbad, CA). For rare event detection, the REIS makes use of proprietary software allowing for fast and highly sensitive fluorescent color detection and the analysis of a variety of morphometric features. For the detection of fluorescing cells, the slide is first scanned at low magnification using a Nikon Plan Apo 4x objective (numerical aperture = 0.2) and one or more user-defined filter set. After thresholding for fluorescence intensity, objects of interests are identified using predefined ranges of image analysis parameters from the Image-Pro Plus menu, such as mean density, area, and roundness. These finding parameters may be adjusted to obtain optimal sensitivity and specificity in sample types using different fluorescent dyes and filter sets.
In this study, the system’s finding parameters were set to a high sensitivity to detect all tumor cells in both clinical and spiked samples, including cell clumps and weakly or heterogeneously stained cells. In a second scan, the system returns to the objects originally identified in the first scan, and autofocused images are automatically taken at a user-defined higher magnification using Nikon S Fluor objective series (10, 20, 40, or 60x). These images are presented in an image gallery for review and classification by a pathologist or other laboratory professional. The X-Y coordinates of each object are stored, and a "revisit" capability allows the user to double-click on images of interest and return to the proper location on the specimen slide for further manual microscopic review. In this mode, it is possible to navigate across the slide, adjust focus, change microscope objectives/filters, and take additional pictures. During the review process, the user can accept images of true positive cells and delete images of debris or other false positives. Only the accepted images will remain in the final data file, shown on the report, and uploaded to the database. The relocation feature can also be used with archived data files, allowing the slide to be processed for additional markers or for in situ hybridization studies, and reanalyzed.
Total Cell Count.
To accurately measure the frequency of OTCs in the peripheral blood, a total cell count algorithm was implemented in the REIS. The total cell count is performed on the 4x DAPI channel images using a dynamic, intensity-based histogram analysis thresholding method. When the total cell count feature is activated, DAPI channel images are automatically obtained at every field of view on the slide. Image analysis algorithms were designed to exclude dirt particles from the total cell count and to split clumps or touching cells into individual objects. The precision of the total cell count algorithm was validated on low- and high-density slide preparations by comparing results with manual counts (results not shown). Because we observed <10% slide-to-slide variation of total cell counts between samples from the same patient, the total cell count is only performed on one slide per patient sample. The total number of cells is calculated by multiplying this value with the number of slides analyzed.
Multicount Feature.
In addition to the DAPI channel (identification of nuclei), the REIS is equipped with a three color proprietary detection and discrimination algorithm that quantifies single-, double-, or triple-labeled cells and displays the cell counts and images accordingly. The number of channels can be flexibly extended if necessary.
Data Presentation, Storage, and Management.
The pertinent information for each sample is entered via a client program to a PostgreSQL database. This information is later associated with the scanning results obtained by the REIS. The scanning results include morphological measurements, total cell counts, and automatically obtained high-magnification images. Once all information is stored in the database, it can be reviewed from multiple client computers using a custom software application that provides the ability to search, sort, and filter data.
|
RESULTS |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
shows carcinoma cells identified in a blood cell preparation from a breast cancer patient. Staining is evident in the cytoplasm of the tumor cells but is absent in surrounding normal mononuclear blood cells that demonstrate only the blue nuclear counterstain.
To examine the feasibility of automated detection of occult tumor cells, the blood sample was analyzed using the Rare Event Imaging System. The REIS detects likely cancer cells based on green fluorescent staining and user defined morphological characteristics (see "Material and Methods"). An image gallery of identified objects is built and presented for review by a laboratory professional. The classification of the cells into tumor and nontumor cells can be performed by solely reviewing the images or by relocating the cell on the microscope stage and viewing it at higher magnification. The user can delete nontumor cells from the image gallery and keep only true positive cells in the final scan file. In the case illustrated, six collected images were classified as containing one tumor cell each (Fig. 1)
. The REIS counted a total of 2,576,950 mononuclear cells on six specimen slides.
Sensitivity, Specificity, and Reproducibility of REIS.
Optimal ranges of the finding parameters (area, mean density, roundness) had to be determined for each tumor type and marker studied. This was accomplished by depositing separate mixtures of several breast (BT-474, SK-BR-3, MCF-7, and MDA-MB 435) and small cell lung cancer cell lines (SW-2, HTB-119, and H82) on adhesive slides. The cell lines were selected to represent the variety of different expression levels of the specific marker. The slides were subjected to the same staining protocols as described in "Materials and Methods" and were digitally measured using the REIS (results not shown). The minimal and maximal values of the finding parameters were recorded and used in additional studies.
To evaluate the sensitivity and specificity of the REIS for tumor cell detection in the red channel, a set of 30 blood samples from healthy donors was analyzed. Twenty of the slides contained spiked cancer cells (SK-BR-3) and 10 contained only normal WBCs. Ten of the spiked slides had high tumor cell loads (10–28 cells), and 10 slides had low tumor cell loads (1–5 cells). After two or three laboratory professionals had independently reviewed the slides manually, slides were scanned by the REIS. The two sets of data, manual and automated, were compared, and the result of this is presented in Table 1
. Manual scanning produced considerable interobserver variability, particularly in samples with a high tumor cell load. In 18 of 20 samples containing cancer cells, automated analysis was in agreement with the highest count on manual scans. In slide A1, one reviewer found two more positive cells than the REIS-assisted analysis, whereas the other two reviewers had identified the same number as the REIS. This discrepancy might be explained by repeat counts of the same cell(s) by the first reviewer. In slide A9, one additional positive cell was identified by the REIS but missed by all three reviewers. This might have been a weakly positive cell that was missed by the manual screen but was detected by the REIS. No positive samples were missed by the manual or automated procedure. Furthermore, no false-positive results were obtained with either method in the 10 negative cases. A similar experiment conducted to evaluate the sensitivity and specificity of the REIS for tumor cell detection in the green channel gave similar results (data not shown).
Slides obtained from two lung cancer and two breast cancer patients were subsequently analyzed in an effort to assess the reproducibility of REIS-assisted analysis results. Table 2
illustrates the result of this analysis in which five separate analyses on 5 days were performed for each specimen. The total number of red and green objects detected by the system (raw count), including the number of tumor cells identified by the laboratory professional (after review count) were registered (Table 2)
. Although the total number of automatically collected objects varied slightly from run to run, in every run the reviewer identified the same number of tumor cells.
Increasing Specificity and Shortening Hands-on Time with the "Nuclear Verification Feature."
Even when using carefully optimized staining protocols, the presence of nonspecifically fluorescing particles is unavoidable in fluorescence microscopy preparations. To increase specificity of the automated scan, we introduced a "nuclear verification" feature into the imaging process. This feature automatically obtains a DAPI channel image if a positive event is identified. Overlaying both images verifies that the positive signal originated from a cell (which contains a DAPI-labeled nuclei) and not from debris or dirt particles. Table 3
summarizes the results from six slides each from breast and lung cancer patients stained with monoclonal anticytokeratin/Alexa-Fluor488 (green) or polyclonal anticytokeratin/Rhodamine-Red-X (red), respectively. The nuclear verification decreased the number of raw images that had to be reviewed by a mean of 61% (range: 46–77%) without losing any true positive cells.
Scanning of Clinical Samples using REIS.
The performance of REIS was further evaluated on 40 slides of patients with lung cancer and 40 slides of patients with breast cancer. All specimens were blindly evaluated manually and using REIS-assisted analysis. The results of these experiments are shown in Table 4
. In 14 of 37 slides containing cancer cells (37.8%), the reviewer detected one or more tumor cell(s) using REIS-assisted analyses that were not detected by manual microscopy. In two slides, manual microscopy detected the presence of tumor cells, whereas REIS concluded the absence of positive cells. At a later date, the specimens were manually reanalyzed, blinded to the original classification. Careful manual screening at higher magnification (using the 20x lens) revealed that there were no tumor cells in either specimen, findings consistent with the REIS result but contrasting with the original manual microscopic analysis. It can be speculated that in these two cases, the originally identified cells exhibited a very weak fluorescence signal that further weakened during storage. Despite successful efforts to reduce photobleaching by embedding cells in ProLong mounting medium and storing slides at 4°C, cells occasionally lose their fluorescence during prolonged storage.5
Characterization of Occult Tumor Cells Using REIS.
Fluorescent microscopy-based analysis has the advantage over brightfield microscopy to readily use multiple markers to characterize the protein expression profile of OTCs. The functionality of the multicount feature, implemented into our second generation REIS, was demonstrated using normal PB, added to a mixture of MCF-7 and SK-BR-3 breast cancer cell lines. Both cell lines showed a strong expression of cytokeratins. However, MCF-7 cells were negative for HER-2/neu in contrast to the SK-BR-3 cells, which were strongly stained. The REIS-finding parameters were set to only identify the highly HER-2/neu expressing cells as red positives. Subsequently, the slide was scanned with the REIS using both the red and the green channels. Twenty-five total positive cells were detected on the slide. Twelve cells were categorized as being only green (cytokeratin) positive and 13 cells were categorized as being double (cytokeratin and HER-2/neu) positive (Table 5)
. After the scan, the positive cells were relocated. Inspection of their morphometric and nuclear features suggested that the single-labeled cells were indeed MCF-7 cells and that the SK-BR-3 cells were correctly categorized as double-labeled cells.
|
DISCUSSION |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. Furthermore, the REIS showed a high day-to-day reproducibility and reliable characterization of OTCs based on their protein expression profile. Today, OTCs are not only detected by microscopy but also by flow cytometry and nucleic acid-based (molecular) methods. In recent clinical studies, real-time PCR was used to detect circulating cancer cells in peripheral blood of patients with breast, gastric, hepatocellular, and colorectal cancers (28, 29, 30, 31) . However, despite the high detection sensitivity of molecular methods, an exact enumeration of OTCs relative to the number of hematopoietic cells is not possible. Furthermore, there is no characterization on an individual cell basis (e.g., regarding the presence of multiple markers or gene amplifications). Flow cytometry analyzes cells on an individual basis and allows for the fast analysis of tens of thousands of cells in a short time. However, this technique is limited by a detection sensitivity of one tumor cell in the background of 104 hematopoietic cells.
Microscopic methods also analyze samples on an individual cell basis. One important advantage over flow cytometry is that the result is reported as a cellular image that can be interpreted by a laboratory professional. Furthermore, the cell is not "lost" after analysis but can be relocated and subjected to additional analyses (e.g., double-labeling or fluorescence in situ hybridization) for increased specificity regarding the nature of the identified cell. The main disadvantage of microscopic methods, particularly if screening is done manually, is that they are time-consuming and subject to considerable intra- and interobserver variation. Automated microscopic methods, such as REIS-assisted analysis, can reduce pathologist time requirements considerably and also guarantee a reproducible performance.
The number of automatically detected objects in each slide (raw object count) varied between tens and hundreds of objects (Tables 2
and 3
). The raw object count depends on the prepared biological material (e.g., the kind of material analyzed, such as blood from healthy donors versus blood from patients before, during, and after treatment or the type of antibody used for labeling, such as monoclonal versus polyclonal)6
and on the computer’s finding parameters. There is usually a higher raw count, reflecting a higher background, in patient material as compared with model samples.6
For the computer’s finding parameters, the REIS was set to obtain a high level of sensitivity to detect all tumor cells, including heterogeneously stained tumor cells. This high level of sensitivity leads to the concomitant pick up of fluorescent signals from debris, not originating from cells. To reduce the number of raw objects that have to be reviewed by the laboratory professional, we implemented a "nuclear verification feature" into the automated screening algorithm. This feature excludes fluorescing dirt and cell debris that do not contain a cell nucleus (no DAPI signal) and reduces the number of raw objects that have to be reviewed by >50% (Table 3)
.
The data from day-to-day comparison, which included runs with identical finding parameters (Table 2)
, also showed a considerable heterogeneity in the raw object count. These variations are most likely the result of slightly different scan areas and planes of focus. Thus, objects located in the periphery of the scanned area, or objects with morphometric features near the set threshold limits for detection, might be included in one scan but excluded in others. However, the final counts after manual verification were 100% reproducible in the day-to-day comparison.
Sometimes, even careful morphological analysis cannot support or exclude the true nature of a positive cell. One way to address this is to use multiple markers in the same specimen. Multiple marker analysis can be easily achieved in a fluorescence-based system, such as the REIS. We developed double-labeling protocols for patient specimens and implemented an automated multicount-scanning algorithm into the REIS-assisted analysis to detect cells labeled with more than one marker. Although the number of available antibodies for the characterization of tumor cells is limited at this time, the multicount feature will become more valuable as new markers are identified through high-throughput genetic analyses. Another option for the additional characterization of OTCs is the relocation feature of the REIS. This allows for the additional analysis of positive cells on the slide by fluorescence in situ hybridization (fluorescence in situ hybridization; Ref. 23 , 25 ). The advantage is that only the slides with a positive cell need to be prepared for fluorescence in situ hybridization, saving expensive reagents and labor time. Furthermore, one could microdissect cells of interest by laser capture techniques and then subject them to molecular methods that have been designed for the analysis of single cells or small cell populations (33, 34, 35) . This additional information will increase the specificity of tumor cell detection and give further insight into the biology of rare cancer cell dissemination.
In summary, we have developed a fluorescence-based automated microscope system that offers a sensitive and reliable assessment of OTCs in peripheral blood. This approach presents a superior alternative to manual microscopic detection. It now appears feasible to use REIS to conduct larger prospective clinical trials to explore the significance of the occurrence of rare cancer cells in cancer patient specimens. Furthermore, the technology described may be applicable for other types of malignancies or diseases were the enumeration of rare cells in blood samples is of interest.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Lan Bo Chen, Dana-Farber Cancer Institute, 44 Binney Street, Boston MA, 02115. Phone: (617) 632-3386; Fax: (617) 632-4470; E-mail: drchen{at}shore.net
5 A. Ladanyi et al., manuscript in preparation. 
6 S-K. Kraeft and A. Ladanyi, unpublished observations.
Received 10/ 7/03; revised 12/16/03; accepted 1/ 8/04.
|
REFERENCES |
|---|
|
TopABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
1,4-N-acetylglucosaminyltransferase for the detection of gastric cancer. Lab Investig, 83: 187-97, 2003.[Medline]
This article has been cited by other articles:
|
|
 |
|
  W. Janni, B. Rack, K. Lindemann, and N. Harbeck Detection of Micrometastatic Disease in Bone Marrow: Is It Ready for Prime Time? Oncologist, August 1, 2005; 10(7): 480 - 492. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. T. Krivacic, A. Ladanyi, D. N. Curry, H. B. Hsieh, P. Kuhn, D. E. Bergsrud, J. F. Kepros, T. Barbera, M. Y. Ho, L. B. Chen, et al. A rare-cell detector for cancer PNAS, July 20, 2004; 101(29): 10501 - 10504. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Cell Growth & Differentiation |
