Path. Res. Pract. 182, 396-400 (1987)

Quantitative Pathology Today - A Technical View::-

J. P. A. Baak Pathological Institute, Free University Hospital, Amsterdam, The Netherlands

SUMMARY

This paper gives an overview of the different techniques which are currently used in quantitative pathology.

Morphometry is used to denote the interactive quantitative analysis of single cells or tissue sections by means of an eyepiece graticule, a projection microscope or a graphic tablet. It is a simple, inexpensive, relatively fast technique, and allows the quantitation of both cell and architectural (tissue) properties. Standard cell and tissue preparations can be employed for such investigations. This requires (minimal) standardization of the cell and tissue handling process. The degree of acidity (pH) of the fixation fluid, such as "neutral" (or buffered) formalin is probably the most important and easy to measure factor, and should be kept between 6 and 8 in order to obtain reproducible nuclear area measure­ments. Another important factor for reproducible results is the magnification in relation to the size of the particles measured. Careful selection of relevant areas, cells and nuclei done by a skilled pathologist, is often essential, as well as quality control of the measuring process.

In static cytometry, a relatively popular application is the measurement of the DNA content of single cells in slides. Although having the advantage of optical control, the measuring technique of the transmission systems is, at the present state of development, laborious. This restricts the measurable number of cells to one hundred or only a few hundred at maximum. As a result, the reproducibility rate, as well as the capacity to detect small differences is only moderate. It is expected that in the near future, very fast static DNA cytometers will become available, which measure the DNA content with a low coefficient of variation (nearly as low as currently possible with flow cytometers).

With flow cytometry, in contrast to static cytometry, thousands of cells can be measured in suspension within a few minutes. This can not be used only with freshly prepared tissue, but also with paraffin-embedded retrieved archival material. However, flow cytometry does not allow for the analysis of architecture-related features. There is also no direct visual control during the measuring process and the instruments used are rather expensive. Digital Image Processing (DIP) computers, in general, are somewhat less expensive than flow cytometers. They have the capacity to measure large numbers of objects at a reasona­ble speed, as well as architectural features of tissue. It is expected that with the further development of dedicated software, DIP computers will be able to cover most of the quantitative analyses in a pathology laboratory, most likely occurring (in a highly automa­tized) interactive way.

Laserscan microscopy (LSM) is a promising new imaging technique, as it has the capa­city to detect low intensity light events. Furthermore, the better (x-, y-, z-) resolution over

* Based on a presentation held at the Xth European Congress of Pathology, Athens, September 1985.

0344·0338/8710182-0396$3.5010 © 1987 by Gustav Fischer Verlag, Stuttgart

Quantitative Pathology Today· 397

the classical light microscope and the reduced depth of field makes it very suitable for optical sectioning. Thus, if coupled with a digital image processing computer, LSM theore­tically offers the possibility of 3 dimensional imaging of nuclei, in order to detect small clones of cells with a special function.

Furthermore, in the near future systems to archive large image and knowledge data bases, together with intelligent statistical software, will become of help as expert (consul­tant) systems.

Introduction

The increased interest amongst pathologists in quantita­tive pathology is obvious from the many articles over the past three years, in which quantitative microscopical tech­niques have been used. Furthermore, a number of books on morphometry and quantitative image analysis have been published since 1979, some of these having a more technical, others a more diagnosis-oriented content. At the IXth Congress of the European Society of Pathology (ESP), held in September 1983, only one afternoon was sufficient to contain the presentations on quantitative pathology. Since that time, at least four congresses related to that subject have been organized in Europe. At the ESP Con­gress in Athens, in September 1985, there was a full-day symposium on morphometry and image analysis.

The reason for this explosion of interest is, on the one hand, rooted in the objective nature of such quantitative analyses and, on the otheihand, in the growing realization that pathology diagnoses are not always as reproducible as was once thought. This is especially true in the so called "continuous lesions" 5, often found in benign and malig­nant proliferative conditions. Furthermore, quantitative of microscopic features can result in the detection of present hidden differences and changes.

This issue of Pathology - Research and Practice con­tains a collection of papers related to the quantitative analysis of microscopical images. This paper intends to give an overview of the image analysis techniques used in quantitative pathology today. These techniques can be subdivided into: 1. Morphometry 2. Static cytometry 3. Flow cytometry 4. Digital image processing 5. Expert systems

The different aspects of measuring properties of quan­titative image analysis have been previously discussed8

•

Morphometry

Strictly spoken, morphometry is the quantitative description of a shape. In general, however, the term is more freely used with regard to the quantitative descrip­tion of a structurell . Thus, morphometry could imply any type of quantitative analysis. In practice, however, the term is used to denote a particular area of quantitative pathology. It is restricted to the use of simple equipment

for interactive quantitative image analysis of geometric features. The investigator thus uses its full pattern recogni­tion and diagnostic capacities in the analysis and measur­ing procedure.

Typical examples of quantitative analysis are mitotic counts and differential counts of leukocytes in a blood smear. Being one of the oldest types of quantitation, the daily worldwide use of these analyses clearly indicates the usefulness of such interactive morphometry.

If the particles of interest are randomly dispersed and widely spread, a standard microscope and a mechanical counter will fulfill the needs of the investigator. However, as soon as the situation is slightly more complicated, some additional equipment may be required.

An ocular grid can be a useful aid to help the observer remain on the correct path when meandering through tis­sue by visual inspection. Furthermore, such inexpensive graticules can be used for stereological analysis in order to assess features as volume percentages, surface, length and numerical densities.

However, in routine use, a projection microscope often i~ more practical for such analyses. Observer-designed line and point grids, drawn on a transparent sheet, can be attached to the projection head, thus adding to the flexibil­ity of the measuring-systems. Furthermore, a microcom­puter with dedicated software will facilitate the counting process, analysis and reporting. Several such programs have already been published4

•

Graphic tablets are widely available commercially and can be used to assess such features as area, perimeter, longest and shortest axes, shape factors and other particles of interest. Measurements can be performed through the microscope if equipped with a drawing mirror, or directly on the graphic tablet using image projections or micro­photographs. In order to guarantee good reproducibility (coefficient of variation below 1.5%), care should be taken to adjust the magnification to the size of the particle mea­sured. Because manual tremor can blur the measurement result, magnification should be such that the particle pro­jected on the graphic tablet has a minimum diameter of 15 mm (Fleege, Smeulders and Baak, 1987, in prepara­tion).

The time required for adequate analyses varies with the accuracy necessary and the features being investigated; the time required is usually between 10-60 minutes. In gen­eral, 'special time necessary for specimen handling is mini­mal. Unpublished studies performed in our laboratory have shown that the degree of acidity of the fixative is the major factor which can influence the nuclear area, and should preferably be kept constant (between 6 and 8).

398 . J. P. A. Baak

Static Cytometry

Static cytometry (SCM) is the measurement of cell and tissue characteristics in slides. Although nuclear mor­phometry falls within the boundaries set by this definition, it is usually not classed under static cytometry, as the latter technique uses (cyto)photometry rather than interactively measuring geometric characteristics.

At the present time, SCM is a much more expensive and time consuming technique than morphometry. It is tech­nically difficult to perform SCM in tissue sections because of overlap of nuclei and cutting of segments (a phenome­non called capping).

50

40

30

Number 20

10

DNA measumerement Static Cytometry

DNA Value Fig. 1. DNA (static) cytometry histogram which shows a narrow and uniform peak. The percentage of cells above 2.Sc is small.

50

40

30

Number 20

10

DNA measumerement Static Cytometry

DNA Value Fig. 2. DNA histogram of another, much more pleomorphic tumour. The DNA histogram is much broader than that of Fig. 1. As a result there are much more cells with a DNA content above 2.Sc.

Nonetheless, it has been shown that with endometrial and ovarian tumours, DNA SCM is an essential prognos­tic factor. Patients with a lower percentage of cells above 2.Sc have a better prognosis than those with higher per­centages above 2.Scl , 6.

Figures 1 and 2 show two examples of SCM. In Fig. 1, a DNA ploidy pattern is shown from a tumour with rather monomorphous nuclei. As a result, the histogram is nar­row and uniform. In contrast to Fig. 1, Fig. 2 shows a DNA ploidy pattern which is much broader. As a result the tumour has a higher percentage of cells above 2.5 c. Indeed the nuclei of this tumour were much more pleomorphic.

We see that the reproducibility of SCM is a crucial point. Unfortunately, we found that the correlation between two different observers was rather poor (unpub­lished results). Another factor which could possibly have contributed to this is the fact that in practice only 100 nuclei could be measured.

Flow Cytometry

The measurement of a higher number of nuclei will reduce the standard error of the mean DNA value. This is possible with flow cytometers which are now becoming available at a somewhat lower price than in the past (espe­cially the mercury lamp based systems).

Different ploidy patterns are discernible such as diploid, tetraploid, multiploid and aneuploid, each probably hav­ing a different prognostic significance. A disadvantage of this technique was the fact that fresh material was required. However, Friedlander and his colleagues in Syd­ney have developed a technique for the analysis of cellular DNA content of paraffin embedded material. Fig. 3 shows results of this procedure (original H + E section, the single cell specimen with intact nuclei) and Fig. 4 the DNA his­togram. The Feulgen stained single cell slide can be used for static cytometry, but serves mainly as a visual control. However, it can be used for SCM as well. For FCM, other stains (propidium iodine, ethidium bromide, DAPI) have proved to be useful. The results of these measurements on paraffin retrieved material are highly comparable with fresh tissue as far as the DNA-index is concerned. By using this technique in advanced ovarian cancers, diploid tumours had a better prognosis than the aneuploid tumours2

• Fig. S shows an aneuploid tumour and Fig. 6 the corresponding DNA histogram. It is clear that large prospective studies are required to determine the value of DNA cytometry in suspension, either on fresh or paraffin embedded material. The first of such studies are beginning to appear with encouraging results. However, it is not clear whether morpho metrical and DNA-FCM features have independent prognostic value, or that they are strongly correlated. Studies which will answer this ques­tion are currently being carried out.

It must be admitted that FCM has several disadvantages in spite of the charm of this technique. During measure­ment there is no visual control. Also, samples are lost after use unless cell sorting is applied. Furthermore, FCM equipment is rather expensive.

Fig. 3. Original H + E stained paraffin section of solid tumour (ovary), and its single cell suspension.

Ieee

1600

1490

1990

Fig. 4. DNA histogram of the diploid tumour shown in Fig. 3.

Quantitative Pathology Today . 399

Fig. 5. Same as Fig. 3, but for a more pleomorphic ovarian tumour.

3600

Fig. 6. Aneuploid DNA histogram of the tumour shown in Fig. 5.

400 . J. P. A. Baak

Digital Image Analysis

It would be a definite advantage to have an instrument which could measure cells in slides at high speed with the possibility of analyzing other parameters such as architec­tural features of tissue.

In principle, digital image processing (DIP) computers have these capacities. Developments in this area are increasing at a rapid rate. In the near future, certain his­tologic features will probably be detected and measured in a highly automatized and reproducible way, using a DIP computer. These features include mitoses, epithelium/ stroma characteristics, nuclei, nucleoli, cellular DNA con­tent and others. In a feasibility study of breast cancer, mitoses could be detected with fairly simple thresholding and binary image manipulation3

• Epithelium and stroma could also be accurately detected and measured7

• Static DIP DNA cytometers have now become available as well.

Laserscan Microscopy

Laserscan microscopes (LSM) are in between electron and light microscopy. They offer the possibility to detect particles which are (just) not visible with a light micro­scope, as they can detect very low intensity light events. LSM thus can be of help to detect certain oncogens, and perhaps oncoproteins.

Furthermore, confo,cal LSM offers better (x-, y-, z-) resolution over the classical light microscope and the reduced depth of field makes them very suitable for optical sectioning. Different forms of image processing are possi­ble as well, as certain laserscan microscopes can be cou­pled with digital image processing computers. In this way, 3-dimensional imaging of nuclei, and detection of small clones of particular cells with a special function may become possible. Therefore, LSM is a neW promising and potentially important imaging technique for quantitative pathological examination of (sub) cellular and tissue com­ponents.

Expert systems

Computer programs capable of performing a task which normally requires human intelligence or an expert are called expert systemslO

• Such inclusion of artificial intelli­gence as a diagnostic aid or consultant is a promising development in quantitative pathology. This assistance can consist of a certain (quantitative) decision rule, but also could be a computerized archive of criteria or images

Received October 14, 1986 . Accepted October 28, 1986

linked with the occurrence of certain histological changes and features.

As to the latter, it is thought that in the near future, optical discs will, at least partially, replace colour atlas books. Currently, optical discs can contain more than 50,000 colour slide images. For practical reasons, analog discs are presently in an easily accessible state, but in the near future digital storage of images will be a more likely solution. Certain commercially available digital image processing computers already have a restricted possibility for archiving (approximately 250) and fast displaying of images. This opens the possibility of additional quantita­tive cell and tissue analyses. Connection of optical disc recorders with personal computers which are already available will make such systems considerably more economical.

Acknowledgement. I would like to thank Mr. Vanderstaal and Mrs. F. Smith, Melbourne, Australia, for correction of the text and Ms. Jose Konneman and Ms. Hanneke Kruyswijk for typing the manuscript.

References

1 Erhardt K, Auer G, Forsslund G et al (1984) Prognostic sig­nificance of nuclear DNA content of serous ovarian tumours. Cancer Res 44: 2198-2202

2 Friedlander ML, Hedley DW, Taylor IW, Russel P, Coates AS, Tattersall MHN (1984) Influence of cellular DNA content on survival in advanced ovarian cancer. Cancer Res 44: 397-400

3 Kaman EJ, Smeulders AWM, Verbeek PW et al (1984) Image processing for mitoses in sections of breast cancer: A feasibility study. Cytometry 5: 244-249

4 Knol RGF, de Wilde PCM, Slootweg PJ (1986) A basic pro­gram for computer-assisted point-counting techniques. An Quant Cytol Histol 8: 75-80

5 Langley FA, Baak JPA Oort J (1983) Diagnosis making: error sources. In: A manual of morphometry in diagnostic pathology, Baak JPA and Oort J (eds), Springer, Berlin, p 6-14

6 Moberger B, Auer G, Forsslund G, Moberger G (1984) The prognostic significance of DNA measurements in endometrial carcinoma. Cytometry 54: 430-436

7 Schipper N, Smeulders AWM, Baak JPA, Young T (1985) Staining evaluation and scene segmentation of ovarian tumors. Proc An Cytol11: 438

8 Smeulders A WM, Dorst L (1985) Measurement in Mor­phometry. An Quant Cytol Histol 7: 242-249

9 Voort van der HTM, Brakenhoff G J, Valkenburg J A C, Nan­ninga N (1985) Design and use of a computer controlled confocal microscope for biological applications. Scanning 7: 66-78

10 de Vries PH, de Vries Robbee PF (1985) An overview of medical expert systems. Meth Inf Med 24: 57-64

11 Weibel ER (1979) Stereological methods practical methods for biological morphometry. Vollo London: Academic Press

Key words: Quantitative pathology - Static cytometry - Flow cytometry - Digital image processing - Morphometry techniques

J. P. A. Baak, Pathological Institute, Free University Hospital, De Boelelaan 1117, 1007 MB Amsterdam, The Netherlands