Excessive noise injection training of neural networks for markerless tracking in obscured and segmented environments.

In this letter, we demonstrate that the generalization properties of a neural network (NN) can be extended to encompass objects that obscure or segment the original image in its foreground or background. We achieve this by piloting an extension of the noise injection training technique, which we term excessive noise injection (ENI), on a simple feedforward multilayer perceptron (MLP) network with vanilla backward error propagation to achieve this aim. Six tests are reported that show the ability of an NN to distinguish six similar states of motion of a simplified human figure that has become obscured by moving vertical and horizontal bars and random blocks for different levels of obscuration. Four more extensive tests are then reported to determine the bounds of the technique. The results from the ENI network were compared to results from the same NN trained on clean states only. The results pilot strong evidence that it is possible to track a human subject behind objects using this technique, and thus this technique lends itself to a real-time markerless tracking system from a single video stream.

Monocyte aggregation around agarose beads in collagen gels: a 3-dimensional model of early granuloma formation?

Granulomas consist of organised collections of macrophages showing evidence of activation in response to an agent localised within a tissue. To date, in vitro models of granuloma formation have failed to reproduce the histological appearance of a granuloma, which forms a characteristic three-dimensional structure. The use of 3-D collagen gels allows agarose beads and human mononuclear cells to be suspended together in a milieu which permit the two to interact. The interaction between agarose beads and mononuclear cells produces monocyte-macrophage aggregates within 24 h which resemble an early granuloma. The monocytes show evidence of activation by direct microscopy, electron microscopy and immunohistochemistry. This model should permit the laboratory testing of hypotheses describing granuloma formation.

Effects of treatment on the histopathology of leprosy.

AIMS: To identify the histological changes in leprosy skin lesions over the first few weeks after the start of leprosy treatment and to examine their relationship to reversal reaction. METHODS: Sequential skin biopsy during treatment with multiple drug therapy. In this study, a series of 28 patients was studied, from whom two or more biopsies were taken at two week intervals. Fourteen patients had paucibacillary leprosy (PBL) and 14 had multibacillary leprosy (MBL). RESULTS: In most cases, granuloma fraction and bacterial index fell during treatment, the bacterial index being less sensitive than the granuloma fraction. Since the biopsies were fixed in buffered formalin and processed through to paraffin wax, little immunohistochemistry was feasible. However, there was strong evidence of immune activation, with increased expression of HLA-DR in the granulomas of MBL and PBL cases: the epidermis also expressed HLA-DR in several patients. Such changes may reflect gamma IFN production from granuloma lymphocytes. Patients with reversal reaction often showed HLA-DR expression on admission which decreased with corticosteroid treatment. CONCLUSIONS: The results suggest that activation of cell mediated immunity in leprosy lesions occurs during treatment with multiple drug therapy and may not be restricted to those with clinical evidence of reversal reaction.

Irregular distortion of the erythrocytes (acanthocytes, spur cells) in senile dementia.

An excess of irregularly distorted red cells with spiked forms (acanthocytes, spur cells) has been found in a substantial minority of patients with senile dementia of Alzheimer type (7 of 50 patients, 3 of 21 men and 4 of 29 women). Of 100 control patients, 42 men and 58 women), 5 (3 men and 2 women) showed comparable distortion, but, of these, one man may well have incipient dementia and the others had serious organic diseases which may be associated with comparable erythrocytic changes. The cause of the distortion is not yet clear, but the presence of occasional giant erythrocytes in the absence of general macrocytosis suggests a possible abnormality of cell-membrane synthesis. This distortion may be a useful marker in patients with loss of memory. Whether it is a manifestation of a haemopoietic clone or a constitutional anomaly associated with Alzheimer's disease remains to be seen.

Detection of tumour necrosis factor alpha in sarcoidosis and tuberculosis granulomas using in situ hybridisation.

AIMS: To determine the site of tumour necrosis factor alpha (TNF alpha) product and mRNA in granulomas. METHOD: In situ hybridisation with digoxigenin labelled or biotinylated oligonucleotide probes was used to demonstrate the presence of total mRNA, and then the presence of TNF alpha mRNA in the biopsy specimens of 37 granulomas (31 sarcoidosis, six tuberculosis). RESULTS: TNF alpha mRNA was detected in epithelioid cells, giant cells, and lymphocytes in the granulomas. Some sarcoidosis specimens did not contain detectable mRNA for TNF, but did contain TNF peptide in the epithelioid or giant cells on immunostaining. This may have been due to stored TNF present in cells in which mRNA for TNF is no longer being produced. CONCLUSION: The results suggest that giant cells should not be regarded as effete cells, as they contain large amounts of mRNA and seem to be actively producing TNF alpha.

Measurement of the immunoperoxidase staining of macrophages within liver granulomata of mice infected with Mycobacterium tuberculosis.

A quantitative image analysis technique developed for the measurement of the extent of macrophage activation and epithelioid cell differentiation was performed on mice infected experimentally with Mycobacterium tuberculosis. The granulomatous inflammatory response within the liver reached a peak at day 23 and declined by day 33. Animals of strain B10.BR (H-2k) showed an increased granuloma fraction as compared to Balb/k (H-2k) mice, thus confirming the influence of non-H2 genes in the control of granuloma formation in mice. Using a monoclonal antibody against CD11b/CD18 (Mac1;CR3), we observed two subpopulations of macrophages within the granulomata. The small, darkly staining cells at the periphery of granulomata appear to be newly recruited macrophages. Larger, paler staining cells toward the center of granulomata represent activated and mature epithelioid macrophages. Using a semiautomated image analyzer (Quantimet 970), we measured the relative numbers of these macrophage subpopulations. There were more activated macrophages (epithelioid cells) associated with the increased granuloma fraction in the B10.BR mice than in the Balb/k. However, similar numbers of newly recruited peripheral macrophages were found in both Balb/k and B10.BR strains. This technique has shown qualitative as well as quantitative differences in the granulomatous inflammatory response in this murine model of tuberculosis in strains of mice with quite different antibody repertoires to mycobacterial antigens.

A novel tissue preservation medium for immunoperoxidase staining of skin biopsies.

The requirement for unfixed tissue is a major drawback in the use of immunohistochemistry for the diagnosis of inflammatory and neoplastic disease. We describe the use of a novel gel transport medium (U.K. Patent No. WO 90/07703, international patent applications pending) to preserve unfixed skin biopsies from allergic contact dermatitis reactions for 1 week prior to frozen section and immunohistochemistry for leucocyte antigens. The medium can be used with biopsies up to 6 mm3 and does not require any alteration in the usual frozen section or immunohistochemical staining procedures. The results show a subjective improvement in both morphology and staining quality. We believe this to be due to improved cutting of the sections and reduced background staining of collagen. This transport medium should be of considerable benefit to the provision of a clinical service, because it allows tissue for immunohistochemical examination to be taken at locations distant from the pathology laboratory.

Epidermal growth factor in the oesophagus.

Epidermal growth factor (EGF) has been implicated in mitogenesis and oncogenesis in the gastrointestinal tract. To determine the role of EGF in oesophageal disease, its quantity and distribution in the oesophageal mucosa of control subjects and patients with oesophageal disease were studied. Oesophageal biopsy specimens, taken 20-40 cm from the incisors in 72 patients, were graded histologically and adjacent specimens were taken for immunohistochemical analysis of the distribution of EGF. In patients with Barrett's columnar lined oesophagus, specimens were also taken from the gastric cardia for comparison. Twenty two biopsy specimens showed oesophagitis, 20 Barrett's mucosa, and 30 were histologically normal. EGF was found in the capillary endothelium of the normal oesophageal papillae and basal mucosa. Significantly more EGF positive papillae were found in the normal mucosa (81%) than in the inflamed mucosa (42%) (p < 0.001). The 20 patients with Barrett's mucosa showed abnormal expression of EGF in 25% of the isthmus and superficial epithelial cells. This study has shown that EGF is found only in the endothelial cells of the capillaries of the normal oesophageal mucosa and that the peptide is detectable significantly less frequently than normal in the inflamed oesophageal mucosa. EGF is also abnormally present, in large quantities, in the cytoplasm of the epithelial cells of Barrett's mucosa compared with gastric mucosa.

Oncogenes and onco-suppressor gene in adenocarcinoma of the oesophagus.

While the activation of the proto-oncogenes has been implicated in the development and progression of cancer of many tissues, the role of oncogenes in the development of oesophageal adenocarcinoma has not been defined. Fifteen patients who had undergone resection for oesophageal adenocarcinoma and 15 who had undergone oesophagectomy or biopsy for Barrett's oesophagus were studied. The latter patients also had adjacent normal gastric mucosa biopsied for comparison with the metaplastic oesophageal mucosa. The mucosal samples were snap frozen and subsequently stained with monoclonal antibodies to the following oncogene associated proteins; c-erbB2 (neu and CE-1) (external domain), c-erbB2 (NCL-CB11) (internal domain), c-src, c-ras, c-myc, c-fos, c-jun, and the onco-suppressor gene--p53. All tumours were well or moderately differentiated adenocarcinomas arising from the lower third of the oesophagus. Eleven specimens showed strong membraneous staining with both c-erbB2 (neu) and c-erbB2 (CBL-CB11). Seven specimens showed strong nuclear staining with p53 onco-suppressor gene. Three specimens were positive for c-ras and c-src, and two were positive for c-jun. In Barrett's epithelium, nine specimens were positive for c-erbB2 (neu and CB11), three were positive for c-src, two were positive for c-ras and c-jun, and one was positive for c-fos. Two of the gastric mucosal biopsy specimens expressed c-erbB2 weakly but no other oncogenes were found. The frequency of positive staining for c-erbB2 is very high, compared with the expression of these genes in other tumours. It is also concluded that errors in the onco-suppressor gene p53, and especially in the external and internal domains of c-erbB2, which is also often expressed in Barrett's mucosa, may be implicated in the development of adenocarcinoma of the oesophagus.

Growth regulatory peptides in gastric mucosa.

1. Epidermal growth factor and the related peptide transforming growth factor alpha have been implicated in the stimulation of gastric mucosal proliferation. We assessed the immunohistochemical distribution of these peptides and their receptor, epidermal growth factor receptor, in mucosa from the antrum and body of the stomach from 28 patients. Twenty-three of the 56 biopsies were histologically normal (12 antrum and 11 body), whereas the other 33 showed varying degrees of inflammation. 2. Epidermal growth factor, transforming growth factor alpha and epidermal growth factor receptor had maximal density of distribution on the apical surfaces of the superficial epithelial cells, but were also expressed to a lesser extent on neck and body cells of the glandular tissue (P less than 0.05). We also demonstrated that epidermal growth factor expression was greater in the epithelial cells of inflamed mucosa than in those of normal mucosa (P less than 0.05). 3. We assessed mucosal proliferation by the Ki-67 labelling index. Ki-67-positive cells were found predominantly in the neck area of the glands and were more frequent in glandular antral tissue than in body glandular tissue (P less than 0.05). Expression of epidermal growth factor receptor in the neck and isthmus cells had a significant correlation with the Ki-67 labelling index (P less than 0.05). 4. We conclude that epidermal growth factor and epidermal growth factor receptor may be important in the adaptation of gastric mucosa to inflammation.

Epidermal growth factor receptors in the oesophagus.

The quantity and distribution of epidermal growth factor receptors (EGF-R) in oesophageal mucosa was studied in the oesophagus in order to determine its role in oesophageal disease. Fifty five biopsies were taken from different levels of the oesophagus in 25 consecutive patients undergoing endoscopy. Another group of eight patients with histologically proven Barrett's oesophagitis had a biopsy taken from the area of columnar lined oesophagus. A peripheral, membranous pattern was seen predominantly confined to the basal and immediately suprabasal cells in all of the first group of patients. In the superficial cells a few granular cytoplasmic structures were positive. All patients with Barrett's oesophagitis showed EGF-R staining of the surface epithelium. A computerised planimeter was used to determine the proportion of stained areas of squamous cells which were expressed as a percentage of the total area of squamous cells. The difference in the area of cells stained for EGF-R between normal and inflamed oesophageal mucosa (29.5% and 43.1% respectively) was significant (p less than 0.001).

Immunohistochemical demonstration of B and T-lymphocyte cell surface markers: an assessment of tissue transport/storage media.

A number of solutions were tested on lymphoid tissue with a view to devising a tissue transport/long-term storage medium which would preserve the morphology and antigenicity of lymphocytes, for demonstration of B- and T-cell surface markers using monoclonal antibodies. Some of the solutions produced total loss of any recognisable tissue structure, but one newly-devised medium gave excellent preservation of morphology and nuclear detail, together with extremely good monoclonal antibody staining. This sterilised medium, Carmichael's medium*, contains a buffer and cryoprotectant: it forms a gel at 4 degrees C, but it is easily liquefied in warm water just prior to immersion of the biopsy. Sections from tissue preserved in Carmichael's Medium were found to be of equally good quality as, and in some cases better than, those which had been fresh frozen. *International patents are currently pending on the formulation of this medium, which will be available commercially through Raymond Lamb Ltd, London, England, UK.

Morphometric analysis of Mycobacterium tuberculosis infection in mice suggests a genetic influence on the generation of the granulomatous inflammatory response.

There is evidence in natural human disease and experimental infection in mice that host genetic factors influence susceptibility to infection with Mycobacterium tuberculosis and the progress of the disease. In mouse models, both H-2 and non-H-2 genes have been implicated. In this study, four inbred strains of mice (Balb/b, Balb/k, B10, B10.BR), selected for combinations of two different H-2 haplotypes on two different non-H-2 backgrounds, were inoculated with M. tuberculosis, strain H37Rv, by intraperitoneal injection. The histological features of the granulomatous inflammatory response in the liver and lungs were investigated during the first 18 weeks of the infection. Granuloma fraction, mean granuloma area, bacillary load, and the density of acid-fast bacilli within granulomata were measured. Animals of all four strains showed the same general pattern of infection with an early, and later self-limiting, infection of the liver and delayed onset, but progressive, infection of the lung. The non-H-2 related genetic background appears to influence the morphology of the granulomatous inflammatory response. In comparison, H-2 differences appeared to be small and inconsistent.

Preservation of keratin expression in oral mucosa using a novel transport medium.

Cellular expression of keratins (intermediate filaments) can be demonstrated by immunocytochemistry using unfixed tissue. However, for practical reasons, provision of fresh tissue to the laboratory is often difficult. Recently a fresh tissue transport medium (Patent applied for) has been developed which allows keratin immunocytochemistry to be performed up to 4 days after biopsy. In this study oral mucosal biopsies from 10 patients were hemisected, half placed in the new transport medium at 4 degrees C and the other half immediately frozen in liquid nitrogen. After 4 days frozen sections of both halves of the biopsies were stained by immunocytochemistry using various antikeratin antibodies. The morphology and staining characteristics of the two halves of the biopsies were then assessed. No significant difference could be found in either morphology or preservation of keratin expression in specimens stored in transport medium, as compared to those in liquid nitrogen. This new transport medium may offer considerable advantage for the provision of a histologic and immunocytochemical diagnostic service.

Unsupervised clustering and centroid estimation using dynamic competitive learning.

In this paper, an unsupervised learning algorithm is developed. Two versions of an artificial neural network, termed a differentiator, are described. It is shown that our algorithm is a dynamic variation of the competitive learning found in most unsupervised learning systems. These systems are frequently used for solving certain pattern recognition tasks such as pattern classification and k-means clustering. Using computer simulation, it is shown that dynamic competitive learning outperforms simple competitive learning methods in solving cluster detection and centroid estimation problems. The simulation results demonstrate that high quality clusters are detected by our method in a short training time. Either a distortion function or the minimum spanning tree method of clustering is used to verify the clustering results. By taking full advantage of all the information presented in the course of training in the differentiator, we demonstrate a powerful adaptive system capable of learning continuously changing patterns.

Measurement of the tissue distribution of immunoperoxidase staining with polyclonal anti-BCG serum in lung granulomata of mice infected with Mycobacterium tuberculosis.

Mice inoculated with Mycobacterium tuberculosis, strain H37Rv were used as a model of human tuberculosis. The microanatomical location of immunoperoxidase staining with a polyclonal anti-BCG serum was within macrophages and appeared granular rather than delineating whole bacilli. Immunoperoxidase staining appears to demonstrate degraded mycobacterial antigens from disrupted organisms and so reflects prior turnover of bacilli. On Ziehl-Neelsen staining, intact or almost intact bacilli are seen and so the extent of this form of staining reflects the current bacillary load. Both methods have limited sensitivity, but with larger mycobacterial loads the area of immunoperoxidase stain measured on a semi-automated image analyser correlated with the numbers of bacilli observed. The immunoperoxidase method will be useful in the evaluation of residual antigen in studying the pathogenesis of experimental murine tuberculosis. In human mycobacterial granulomata, this immunohistochemical technique should provide an alternative method of estimating the extent of bacillary load: this approach may also provide evidence of mycobacterial infection from residual antigen deposits in the tissue when whole bacilli have been successfully cleared.

Improved silver staining of nucleolar organiser regions in paraffin wax sections using an inverted incubation technique.

A new simple modification to the silver staining of nucleolar organiser regions (AgNORs) was devised which, by performing the incubation with the slide inverted, results in minimal undesirable background staining, a persistent problem. Inverted incubation is facilitated by the use of a commercially available plastic coverplate. This technique has several additional advantages over other published staining protocols. In particular, the method is straightforward, fast, and maintains a high degree of contrast between the background and the AgNORs.

Expression of CD15 antigen in urinary bladder transitional cell carcinoma.

The biopsy specimens of 91 patients between the ages of 38 and 94 with transitional cell carcinoma of the bladder were retrospectively reviewed to determine if the expression of CD15 antigen detected by a monoclonal antibody MC2 was correlated with prognosis. Expression was variable, ranging from strong expression of the antigen by only the superficial cells in well differentiated papillary lesions to weak expression by most cells in solid or invasive tumours. In the invasive component there was a correlation between MC2 expression and tumour type, suggesting that the cell surface carbohydrate detected by MC2 may have a role in cell adhesion. There was no correlation between staining and survival. It is concluded that tumour type, grade, and stage remain the best prognostic indicators of urothelial tumours.