The microcomputer and image analysis in diagnostic pathology

Image-guided automatic triggering of a fractional CO2 laser in aesthetic procedures

BACKGROUND

Laser procedures in dermatology and aesthetic medicine are associated with the need for manual laser triggering. This leads to pulse overlapping and side effects.

METHODS

Automatic laser triggering based on image analysis can provide a secure fit to each successive doses of radiation. A fractional CO2 laser was used in the study. 500 images of the human skin of healthy subjects were acquired. Automatic triggering was initiated by an application together with a camera which tracks and analyses the skin in visible light. The tracking algorithm uses the methods of image analysis to overlap images. After locating the characteristic points in analysed adjacent areas, the correspondence of graphs is found. The point coordinates derived from the images are the vertices of graphs with respect to which isomorphism is sought. When the correspondence of graphs is found, it is possible to overlap the neighbouring parts of the image.

RESULTS

The proposed method of laser triggering owing to the automatic image fitting method allows for 100% repeatability. To meet this requirement, there must be at least 13 graph vertices obtained from the image. For this number of vertices, the time of analysis of a single image is less than 0.5s.

CONCLUSIONS

The proposed method, applied in practice, may help reduce the number of side effects during dermatological laser procedures resulting from laser pulse overlapping. In addition, it reduces treatment time and enables to propose new techniques of treatment through controlled, precise laser pulse overlapping.

Improving evaluation of the distribution and density of immunostained cells in breast cancer using computerized video image analysis

Quantitation of cell density in tissues has proven problematic over the years. The manual microscopic methodology, where an investigator visually samples multiple areas within slides of tissue sections, has long remained the basic ‘standard’ for many studies and for routine histopathologic reporting. Nevertheless, novel techniques that may provide a more standardized approach to quantitation of cells in tissue sections have been made possible by computerized video image analysis methods over recent years. The present study describes a novel, computer-assisted video image analysis method of quantitating immunostained cells within tissue sections, providing continuous graphical data. This technique enables the measurement of both distribution and density of cells within tissue sections. Specifically, the study considered immunoperoxidase-stained tumor infiltrating lymphocytes within breast tumor specimens, using the number of immunostained pixels within tissue sections to determine cellular density and number. Comparison was made between standard manual graded quantitation methods and video image analysis, using the same tissue sections. The study demonstrates that video image techniques and computer analysis can provide continuous data on cell density and number in immunostained tissue sections, which compares favorably with standard visual quantitation methods, and may offer an alternative.

Membranous glomerulonephritis: a morphometric study

Archival material from 45 renal biopsies with a diagnosis of idiopathic membranous glomerulonephritis (MGN) were studied by computer-aided image analysis in order to evaluate the prognostic significance of glomerular and interstitial morphometry in MGN. The control group consisted of thirty seven normal renal biopsy specimens. The surface area, the perimeter, the major axis length and the shape factor of renal glomeruli as well as the percentage of the interstitial fibrosis were measured. All the morphometric parameters related to the size of glomeruli had significantly higher values in the patient group (p = 0.000 for all the parameters). However, no significant difference of the glomerular size between different stages of MGN was observed. In contrast, the percentage of interstitial fibrosis increased as the MGN stage rose (median values: 10.3% in stage 1, 14.2% in stage II, 26.9% in stage III, 28.9% in stage IV and 34.2% in stage V, Kruskal-Wallis ANOVA H = 37.645, p = 0.000). In the multivariate analysis the percentage of interstitial fibrosis was the only independent prognostic factor (p = 0.013). Our findings suggest that, in membraneous glomerulonephritis, the interstitial fibrosis increases as the MGN stage progresses, while the size of renal glomeruli has increased at a very early stage of the disease. This fact may indicate that interstitial fibrosis, not glomerular lesions, is mainly responsible for the reduction of renal function.

Image processing techniques for quantitative analysis of skin structures

Computer-based image processing and analysis techniques were developed for quantitative analysis of skin structures in color histological sections. Performance was compared with traditional non-image processing counting methods. Skin sections were stained with Masson's trichrome, hematoxylin and eosin, picrosirius red, or one of several elastin stains. The image processing software identified the top of the cellular epidermis and the dermal-epidermal junction and then calculated the volume of the cellular layer of the epidermis, epidermal thickness, and the ratio of the dermal-epidermal junction surface area to the in-plane surface area. It also identified cells and collagen and calculated cellular densities and collagen densities in the papillary and reticular layers of the dermis. Attempts to computationally process elastin-stained sections to determine elastin density were unsuccessful. The described techniques were used in a preliminary study to compare mechanically stressed skin with control skin. Results showed significant differences in cellular density in the papillary dermis and collagen density in the reticular dermis for skin subjected to combined shear/compression or tension compared with an unstressed control. Measurements made with the computer technique and traditional technique showed comparable results; the mean difference in measurements for epidermal features was 5.33% while for dermal features it was 2.76%. Significance testing between control and experimental groups showed similar results, though for three of the 28 comparisons the computer method identified a significant difference while the traditional method did not. The computer method took longer to conduct than the traditional method, though with recent advances in computer hardware this time difference would be eliminated.

A nonradioactive method of in situ hybridization that uses riboprobes and paraffin-embedded tissue and its combination with immunohistochemistry

Current research into cytokine production in tissue sections relies on the detection of cytokine proteins using a variety of immunohistochemical methods. The disadvantages of this technique are that precise localization to a particular cell is difficult and it is uncertain whether the cells detected by this method are the origin or target of the cytokine or rather have nonspecifically absorbed the secreted cytokine. This question can be clarified using in situ hybridization, but current techniques are insensitive, poorly localizing, or time consuming. Biotin-labeled riboprobes were generated from cDNA fragments sandwiched between two RNA polymerase promoters (SP6 and T7 RNA polymerases) using a commercial riboprobe generation kit containing biotin-labeled UTP. The in situ hybridization technique was used to demonstrate cytokine mRNA in a range of tissues containing an inflammatory infiltrate and with a range of cytokine probes. This technique of in situ hybridization was combined with immunohistochemistry using an immunoalkaline phosphatase technique to show the powerful combination of these two techniques. The biotin-labeled riboprobes were sensitive enough to detect a range of cytokine mRNAs in a variety of tissue sections. The technique can be completed over a 24-h period and produces a stable color product that can be stored for long periods and can be quantitated using image analysis techniques. This technique was performed on paraffin-embedded tissue as well as cryosections and allowed for the detection of mRNA in archival tissue. It was also successfully combined with immunohistochemical techniques to determine simultaneously the localization of a cytokine product in particular cell lineages. A nonradioactive method for in situ hybridization using biotin-labeled riboprobes is described; it is capable of detecting mRNA products from a range of genes in a variety of tissue samples. An amplification step in the method enhances the sensitivity to a level that approaches that of radioactive methods, while maintaining the speed, safety, and simplicity of an immunoperoxidase detection system. The ability to use paraffin-embedded tissue with this method allows for improved tissue architecture and examination of archival tissue. These features should ensure greater use of in situ hybridization techniques in future research studies.

Image analysis of tissue sections

The use of computers for the automated image analysis of tissue sections is becoming increasingly important. The paper presents an overview of current methodologies and summarizes developments in this field. A brief introduction followed by a survey is provided in each of these areas: image transformation, image segmentation and classification.

Simple PC-based system for morphometric analysis of synapses

The computer system designed for synaptic morphometry of aksosomatic and aksospine synapses of brain is described in the present paper. It is based on an AT-comparable personal computer equipped with low-cost frame grabber. This hardware configuration allows to input images from any TV source such as TV camera, videorecorder for further processing. The appropriate software was written in Microsoft Quick Basic to measure the main morphometric parameters of axon ending, dendritic spines, total contact, active zone and mitochondria. Number of synaptic vesicles (total and active) were also counted using a mouse as a pointing device. The derivative parameters (vesicle density, mitochondrial density) are then calculated. All measured data are stored in ASCII format, allowing ease in editing and export into other application programs. Statistical evaluation and calculation of histograms is performed by associated program also written in Quick Basic. The advantages and disadvantages of this approach are discussed.

Differences between standard and high-sensitivity immunohistology in tissue sections--comparison of immunoperoxidase staining methods using computerized video image analysis techniques

High-sensitivity immunoperoxidase labelling can be achieved using heavy metal enhancement of the di-amino (DAB) reaction product. For example nickel chloride combined with DAB improves the sensitivity of the method approximately 7-10 fold. This allows detection of approximately 100-200 molecules on cell surfaces. This has an obvious advantage over standard non-enhanced DAB methods which detect 1000-2000 molecules under similar conditions. This study compares standard and nickel enhanced DAB immunoperoxidase staining of cells in tissue sections using video-image analysis (VIA) measurement techniques. VIA is an objective method of evaluation of immunoperoxidase staining of separated cells and for immunostained cells in tissue sections. Nickel enhancement of the DAB reaction product reveals more positively stained cells, with higher contrast over background, giving superior qualities for VIA analysis providing continuous, reproducible, and objective data for statistical analysis.

Glycosaminoglycans of guinea pig prostate fibromuscular stroma: influence of estrogen and androgen on levels and location of chondroitin sulfate

The effects of aging and hormone manipulation on the glycosaminoglycan (GAG) content of prostatic stroma in guinea pigs were investigated. Total GAG and individual GAG classes (chondroitin, dermatan, and heparan sulfates, and hyaluronic acid) were measured biochemically in stromal extracts. Chondroitin sulfate was also measured and localized by video image analysis of immunocytochemically-stained tissue sections. The weight and total GAG (uronic acid) content of prostatic stroma increased between the ages of 2 weeks and 2 years by 7-8-fold and 4-5-fold respectively. GAG concentration per unit weight of stroma declined 4-fold during puberty and remained essentially unchanged thereafter. Similar results were obtained for each of the GAG classes. The decreases in GAG concentration were associated with a 3-fold increase in the size of the smooth muscle cells of the prostatic stroma during puberty. Hormonal control of GAG deposition in the prostatic stroma was investigated by steroid replacement in prepubertally-castrated animals. Administration of dihydrotestosterone (DHT) to castrate animals for 6 weeks resulted in significantly reduced concentrations of stromal uronic acid, compared with untreated castrate animals (P < 0.05). The GAG levels post-DHT treatment were similar to those observed after pubertal development in sham-operated control animals. Estradiol treatment had the opposite effect to that of DHT, resulting in a significantly increased concentration of uronic acid compared with castrate animals (P < 0.05). These steroid-induced changes in stromal GAG deposition were mostly contributed to by chondroitin and dermatan sulfates. Combined treatment with DHT and estradiol resulted in stromal uronic acid concentrations similar to those of animals receiving DHT alone, indicating that the effect of DHT on stromal GAG deposition is dominant over the effects of estradiol. Morphometric measurement, using computer-assisted video image analysis of a chondroitin sulfate epitope in prostatic sections stained with a monoclonal antibody (6C3), supported the biochemical data. Stereometric profiles across several sectioned glands demonstrated that chondroitin sulfate was confined to the periacinar basement membranes of the prostatic stroma in all groups except the estradiol-treated castrate animals, where the immunostaining extended from the periacinar basement membrane throughout the fibromuscular stroma. Treatment of castrate animals with estradiol alone also induced a physicochemical change in the chondroitin sulfate molecule, resulting in reduced electrophoretic mobility. In summary, this study identifies changes in the quantity, structure, and localization of chondroitin sulfate in the prostatic stroma of estradiol-treated guinea pigs. Furthermore, estradiol and DHT have opposing effects on the level of chondroitin and dermatan sulfate expression in the prostatic stroma.

Computer assisted data collection for stereology: rationale and description of point counting stereology (PCS) software

The paper describes microcomputer software for point counting stereology. Stereology includes a collection of statistical methods that quantify the images of light and transmission electron microscopy. The methods use test grids placed over images to collect raw data, which includes counts of points, intersections, transections, and profiles. In turn, the counts are included in stereological equations that give estimates of compartmental volumes, surfaces, lengths, or numbers. These parameters describe the composition of a structure in three-dimensional space. The PCS (point counting stereology) System Software III serves as a data collection, storage, and management tool. Users set up point counting protocols without programming, enter data by pressing predefined function (MS-DOS) or alphabetic keys (UNIX), store data in files, select files for analysis, and calculate results as stereological densities. The latest version of the PCS software includes a new user interface and is designed as a research "front end" that can feed data either into the calculation tools of a stereology tutorial (Bolender, 1992, this issue) or into the analysis routines of quantitative morphology databases (Bolender and Bluhm, 1992).