A national treasure goes online: the Armed Forces Institute of Pathology.

With the advent of the 21st century, the Armed Forces Institute of Pathology is delivering online medical consultation to the global health care community. By incorporating widely available open system technology into the paradigm of telepathology, the AFIP is providing world class expertise in pathology and access to a wealth of information for physicians and the general public. The national treasure known as the Armed Forces Institute of Pathology has more than a glorious history; it has a pretty bright future, too.

Three-dimensional modeling of renal glomerular capillary networks.

OBJECTIVE: To quantitatively characterize three-dimensional (3D) vascular patterns based on the path length distribution in networks. STUDY DESIGN: Volume data were stacks of confocal fluorescence images of renal glomeruli, obtained using a confocal laser scanning microscope and spanning 60-130 microns in the z axis. After manual editing to remove nonglomerular components, transverse sections of glomerular capillary lumens were segmented automatically using two-dimensional morphologic filters. The center points and the overlap (between adjacent sections) of lumens segmented in the x vs. y plane were used to derive a graph (i.e., a multiply connected network) for each glomerulus. The average degree and the distance between connected nodes were used to derive the number of graph edges and the overall length of the capillary network. RESULTS: Renderings of the 3D reconstructions demonstrated well the lobular structure of the glomerular tufts. Mean capillary length ranged from 53 to 180 microns in 10 glomeruli. Total capillary length ranged from 3,500 to 9,500 microns (mean 5,833). CONCLUSION: Structural measurements based on confocal data require less effort than do measurements based on serial sections and make detailed study of diseased glomerular populations practical.

Uveal melanoma: the importance of large nucleoli in predicting patient outcome--an automated image analysis study.

BACKGROUND: Automated image capture and analysis (AICA) has not previously been used to measure the size of nucleoli in uveal melanoma. In this study, the measurements were tested for a possible association with patient outcome. METHODS: Sections from 63 uveal melanomas were stained using the silver stain for nucleolar organizing regions (AgNOR). AICA was used to measure the following five nucleolar features in ten microscopic fields of the tumor: area, circularity, maximum diameter, width, and length of the perimeter. For each tumor, the mean and standard deviation of each of the features were calculated based on all the nucleoli and on subsets of nucleoli with larger areas. For the five nucleolar features the mean of the largest value (MLV) in each of the ten fields was calculated. For comparison, a related visually measured nucleolar feature (MLN) was obtained from hematoxylin and eosin stained sections using a filar micrometer. RESULTS: Thirty-four patients died with metastatic disease and 29 patients survived at least 5 years without metastasis. A greater proportion of nucleoli larger than 3 mm2 in greatest dimension were from patients who died of their disease. The means of the nucleolar features were less significant outcome discriminators than the standard deviations. Means and standard deviations based on subgroups of nucleoli larger than 3 or 3.5 mm2 in greatest dimension were better discriminators. The MLVs were as effective discriminators as the corresponding standard deviations of the larger nucleoli and were better discriminators than MLN. CONCLUSIONS: AICA of AgNOR stained sections of uveal melanoma provides an excellent method for predicting the outcome of patients with uveal melanoma.

Differential diagnosis of Hürthle cell neoplasms on fine needle aspirates. Can we do any better with morphometry?

OBJECTIVE: To determine the value of computerized interactive morphometry in the preoperative prediction of malignancy in fine needle aspirates of Hürthle cell neoplasms. STUDY DESIGN: Alcohol-fixed, Papanicolaou-stained fine needle aspiration smears of histologically proven Hürthle cell adenomas (HCA) (n = 10) and Hürthle cell carcinomas (HCC) (n = 9) were studied by interactive computerized morphometry. The measured features included the areas, perimeters and shape factors of individual cells, nuclei and nucleoli; the nucleocytoplasmic and nucleolonuclear ratios; and the eccentricities of nuclei and nucleoli. RESULTS: Only nucleolar features showed statistically significant differences between HCA and HCC. These features were the nucleolar area and its standard deviation, the nucleolar form factor and circularity, and the nucleolonuclear ratio. The most effective, albeit imperfect, discrimination was achieved by the nucleolar form factor. CONCLUSION: Nucleolar features, such as size, variation in size and roundness, may be more effective than cellular or nuclear features in differentiating between HCA and HCC in fine needle aspiration smears.

Enumeration of interphase chromosomes. Comparison of visual in situ hybridization and confocal fluorescence in situ hybridization.

OBJECTIVE: To compare two visual enumeration methods to determine whether the confocal approach yielded better counts of chromosome-specific hybridization sites. STUDY DESIGN: Brightfield microscopy was used to count in situ hybridization (ISH) sites in 4-microns tissue sections. Confocal microscopy was used to collect three-dimensional (3D) data sets from fluorescence in situ hybridization (FISH) preparations made with sections of various thicknesses. Analysis of the confocal images relied on custom-built interactive visualization software. RESULTS: The confocal method yielded higher average counts of hybridization sites per nucleus due to fewer truncated nuclei in thicker sections and to visual exclusion of the truncated nuclei that remained. Optimal section thickness was 8-12 microns. Limited penetration by FISH reagents restricted the use of thicker sections. CONCLUSION: Analysis of intact nuclei visualized in three dimensions was more sensitive in demonstrating high centromere number than was brightfield ISH analysis of 4-microns sections. Improvements in semiautomated interactive software may make the confocal approach practical for accurate evaluation of chromosome number in precise histologic contexts.

Applications of neural networks in histopathology.

ANN methods are one of several nonparametric approaches to classification. They are based on an adaptive, connectionist model of information storage and processing and learn by example rather than explicitly programmed classification rules. Their greatest strength probably is the ease with which they capture nonlinear and interactive feature effects during training, though at a higher risk of overtraining then is found with traditional classifiers. ANN can be used to process image data, but so far not to classify complex natural images directly. Instead, image dimensionality is reduced to a set of extracted features for ANN input. ANN finds growing use in time series signal processing and radiology. In anatomic pathology, ANN pilot studies are published for tumor classification. Pap smear analysis, and chromosome identification. Each application requires proper sampling technique, minimising bias in training and testing to assure the ANN classifier will perform prospectively as expected. With careful arrangement of input features, ANN can be used for prognostic as well as diagnostic models.

Adult medulloblastoma: review of 13 cases with emphasis on MRI.

Medulloblastomas are generally associated with childhood, but 14-30% occur in adults, accounting for 1% of adult central nervous system tumors. While approximately one third of adult medulloblastomas present as vermian tumors similar to those seen in childhood, the majority differ substantially from the childhood variety. In this series of 13 patients, 5 had lateral, seemingly extra-axial masses in the cerebellopontine angle or at the tentorium, simulating meningiomas or acoustic neuromas, but angiographic hypovascularity in 2 of the latter suggested a diagnosis other than meningioma. Of 4 paramedian tumors, 3 diffusely infiltrated the cerebellar white matter, showed little or no gadolinium enhancement and were not associated with hydrocephalus. Hydrocephalus was present in less than half of our patients; in childhood the reported incidence is 85-100%. A possible association with pregnancy was noted.

Computerized nuclear morphometry and survival in renal cell carcinoma: comparison with other prognostic indicators.

Morphometric nuclear parameters were compared with patient survival for a series of 174 renal cell carcinomas (RCC) collected over a 30 yr period. Stepdown regression showed long diameter, average feret diameter, form factor and the ratio of average feret diameter to equivalent diameter to be significantly associated with survival. Nuclear area, nuclear perimeter, equivalent diameter, ratio of long diameter to average feret diameter and coefficients of variation of nuclear area and nuclear perimeter were not significantly correlated with survival. All parameters were correlated with a 3 division nuclear grading classification using analysis of variance. Multivariate analysis showed nuclear form factor, tumor stage, silver staining nucleolar organizer region numbers and proliferating cell nuclear antigen expression to be independently associated with survival. The results of this study indicate that form factor is the most discriminate morphometric parameter for RCC, providing survival data additional to that derived from tumor staging and from markers of tumor proliferation.

An alternative health-care reimbursement system--application of arthroscopy and financial warranty: results of a 2-year pilot study.

The results of a 2-year pilot study on an alternative health-care reimbursement system are presented. This innovative system includes a 2-year warranty by the providers to protect the insured and insurer from additional expenses. It is based on the advantages of arthroscopy. This system provided access, choice, and affordability for 111 patients from a managed care environment who were predetermined to be orthopedic surgical candidates. The unique features included unlimited free consultations and office radiographs. Payment was made only if a patient had surgery. This single fee included all related physician and hospital charges for surgery and any subsequent service for the next 2 years under the warranty. The profit margin for the providers was based on the application of arthroscopy, which anticipated reduction in the customary hospitalization. Accountability was provided by reporting of surgical indications and incidence. The surgical incidence of 42% was less than what was projected. The health maintenance organization insurer saved in excess of $125,000.00. Both the hospital and the surgeon earned more than under existing reimbursement systems.

Computer-assisted image classification: use of neural networks in anatomic pathology.

Artificial neural networks (ANN) implemented on digital computers have received much attention for interpretation of images in pathology and cytology. Most such images are too complex for current ANN to interpret directly; instead, ANN usually classify the images according to numeric features extracted from them. In experiments on three distinct image classification problems, ANN classifiers performed as well or better than multivariate linear discriminant analysis (a traditional parametric statistical classifier). ANN empirically define non-linear multivariate decision boundaries, and can combine non-contiguous feature areas in mapping a classification. However, many training cases are required in order to map complex area boundaries precisely and with a low risk of 'overtraining.' Careful problem selection and attention to data dimensionality and format are important for efficient ANN use.

Domestic partner abuse treatment programs and cultural competence: the results of a national survey.

Partner abuse literature reveals that treatment is less effective with minorities than with their white counterparts. Our survey of partner abuse programs (N = 142) indicates that, for the most part, little or no special effort is being made to understand or accommodate the needs of minority populations. This colorblind approach lacks the effectiveness of a culturally competent approach, which fosters an environment that helps minority groups succeed in treatment. Do partner abuse programs make efforts to be culturally competent? The major characteristics of an organizational transformation towards cultural competence occur when these programs do the following: (a) network with the minority community, (b) locate outside consultants with expertise in working with minority clients, (c) obtain information concerning service delivery and programming for minority clients, and (d) have at least one bilingual counselor. These four characteristics accounted for .58 of the variance in the minority-focused activities.

Flow cytometric and quantitative histological parameters to predict occult disease in clinical stage I nonseminomatous testicular germ cell tumors.

The goal of this study was to determine if deoxyribonucleic acid (DNA) flow cytometric and quantitative histological parameters could predict occult metastases in clinical stage I nonseminomatous testicular cancer. Archival paraffin primary tumor tissue was available from 36 clinical stage I nonseminomatous germ cell testicular cancer patients who all had retroperitoneal lymphadenectomy and followup defining 2 groups: pathological stage I (23) and occult pathological stage II (13). Archival blocks were microdissected and individual histological components were subjected to flow cytometry. In addition, the primary histology was reevaluated for vascular invasion and per cent composition of histological components of embryonal carcinoma and other histologies. For flow cytometry parameters, no tumor was uniformly diploid, and the DNA index and per cent S phase cells were not useful in differentiating stages. Although mean per cent S phase for the aneuploid cell population and proliferative index were significantly greater for stage II cases by univariate logistic regression analysis, they are approximately 70% accurate in predicting occult disease as single tests and were not significant by multivariate analysis. The calculation of per cent embryonal carcinoma was also significantly greater in stage II cancer by univariate logistic regression testing and remained significant by multivariate analysis. Vascular invasion was marginally predictive of occult disease but was also not significant by multivariate analysis. Calculating the percentage of embryonal carcinoma of a primary testicular tumor may be a useful method to assess clinical stage I cancer patients for risk of occult disease. A larger study is needed to confirm the importance of per cent embryonal carcinoma and to clarify further if flow cytometry in combination is useful.

Use of remote video microscopy (telepathology) as an adjunct to neurosurgical frozen section consultation.

We investigated the use of remote video microscopy (telepathology) to assist in the diagnosis of 52 neurosurgical frozen section cases. The TelMed system (Discovery Medical Systems, Overland Park, KS), in which the referring pathologist selects appropriate fields for transmission to the consultant, was used for the study. There was a high degree of concordance between the diagnosis rendered on the basis of transmitted video images and that rendered on the basis of direct evaluation of frozen sections; however, in seven cases there was substantial disagreement. Remote evaluation was associated with a more rapid consultation from the standpoint of the consultant, who spent approximately 2 minutes less per case when using remote microscopy; this was achieved at the expense of considerably greater effort on the part of the referring pathologist, who spent approximately 16 minutes per case selecting an average of 4.5 images for transmission to the consultant. The use of remote video microscopy for pathology consultation is associated with a complex series of tradeoffs involving cost, information loss, and timeliness of consultation.

Standardization and quality control of quantitative microscopy in pathology.

Standardization and quality control of quantitative microscopy techniques are distinct but related concerns. The first deals with the great variety of quantitative methods, measured features, and even response variables used in investigation of biological or clinical processes. The latter deals with reproducibility of results from those investigations across time and test performance sites. Though distinct, efforts for standardization and quality control are inherently interactive. Consensus on standard methods, instrumentation, and data analysis is hard to achieve in fields developing as rapidly as quantitative microscopy. Consensus is possible, however, on the issues that affect test performance and interpretation. For example, issues of specimen type, fixation, processing, and staining affect image cytometry just as they do flow cytometry. Raw data acquisition issues include area sampling rules and fidelity of optical and sensor systems (light wavelength, glare/stray light, lens aberrations, numerical aperture, depth of focus, scan precision, pixel spacing and depth, sensor linearity, and stability). Intermediate data issues are primarily related to image foreground/background segmentation techniques--automated versus manual, object-specific versus field-based. Data reduction and interpretation procedures also provide many roads for divergence from uniformity. Each of these issues must be considered in terms of its effect on comparability and utility of quantitative microscopy results. Quality control for quantitative microscopy is as important as standardization for its use in research programs and with clinical specimens. The sine qua non of quality control is comparison of experimental results against a known "correct" value to estimate accuracy, and against other experimental results to estimate precision.(ABSTRACT TRUNCATED AT 250 WORDS)

Morphometric distinction of sclerosing adenosis from tubular carcinoma of the breast.

Objective features have been identified that assist in distinguishing sclerosing adenosis from tubular carcinoma of the breast. Hematoxylin and eosin stained paraffin sections of 18 sclerosing adenoses and 18 tubular carcinomas were studied using a TAS Plus image analysis system. Histological measurements from lumens and glands included stereologic features of architecture and morphometry of size and shape (the latter by Fourier coefficients). Cytological measurements included nuclear area, perimeter, diameter and formfactor. Initial analysis suggested utility for several individual features. However, after a modified Bonferroni procedure only two of the features were statistically significant, i.e. the coefficient of variation of luminal form factor and the surface density of glands. Multivariate discriminant analysis using these two variables correctly classified 86% of the cases, with three adenoses and two carcinomas misclassified. Validity of the discriminant rules was supported by classification using measurements from a separate, independent set of cases (ten sclerosing adenoses and nine tubular carcinomas). The classification function computed from the first set misclassified only one case from the second set, a tubular carcinoma, leaving 95% of the cases successfully categorized. Chi square test for 2 x 2 contingency tables gave a p-value < 0.001 for both sets of cases. The results suggest that morphometric features are helpful in distinguishing tubular carcinoma from sclerosing adenosis and point out the need for conservative analysis of high-dimensional data sets.

Computer-assisted image interpretation: use of a neural network to differentiate tubular carcinoma from sclerosing adenosis.

Measurement of nuclear and glandular size and shape features was carried out on 18 cases of sclerosing adenosis and 18 cases of tubular carcinoma. Modified Bonferroni analysis showed that glandular surface density and the coefficient of variation of luminal form factor were significant in discriminating between these two lesions. These two histologic features, together with the diagnosis, were used to train a neural network implementing a backpropagation algorithm. Following training, the network correctly classified 33 of the 36 cases in the training set (92%). Furthermore, the network correctly classified 19 of 19 cases in a test set consisting of cases that were not used to train the network. These results suggest that neural networks may be useful to assist in the differential diagnosis of histologically similar lesions.

Toxicosis in horses after ingestion of hoary alyssum.

Fever, limb edema, and laminitis were observed in horses 18 to 36 hours after they consumed hoary alyssum (Berteroa incana) under field and experimental conditions. Clinical signs were not observed in all horses that had ingested the plant. Diagnosis in the field cases was limited to observation of clinical signs and evidence of plant ingestion in hay or on pasture. In most cases, clinical remission was observed 2 to 4 days after empirical treatment, removal of the plant source, or both.

Prediction of prognosis in untreated stage A2 prostatic carcinoma.

Carcinoma is found unexpectedly in approximately 10% or more of the 400,000 prostatectomies performed annually in the United States. Patients with Stage A2 carcinoma die of their disease in only 35% of the cases. To alter the course of disease in these patients, 65% of Stage A2 patients may be treated unnecessarily by radical prostatectomy, radiation therapy, or hormonal therapy. An accurate method to predict the outcome of patients with Stage A2 carcinoma is needed. Histologic sections from 18 patients with Stage A2 prostatic carcinoma followed without further treatment until progression, or followed without progression, were evaluated by several investigators who did not have knowledge of patient outcomes and who employed standard pathologic grading systems as well as morphometric, cytophotometric, flow cytometric, and immunohistochemical techniques. Outcome was predicted correctly by random sampled absolute (17 of 18 cases) and relative (16 of 18) nuclear roundness factor (NRF), tumor volume expressed as percent of specimen (13 of 16), primary (13 of 18), secondary (14 of 18), sum (15 of 18), and worse (14 of 18) Gleason grades and prostate-specific antigen immunohistochemical findings (13 of 18) that produced statistically significant separation of the two groups. Significant separation was not obtained with Mostofi's pattern, nuclear, sum, and worse grades, Johns Hopkins' grade, absolute tumor volume, nuclear DNA content measured by image cytophotometric study of Feulgen-stained histologic sections and flow cytometric study of propidium iodide-labeled suspensions of nuclei obtained from paraffin blocks, nonrandom sampled NRF of worse and most prevalent neoplastic areas, and prostatic acid phosphatase and peanut agglutinin immunohistochemical study. NRF measured by a random technique best predicted outcome in these patients with A2 prostatic carcinoma and should be evaluated prospectively as a means for selecting patients who require therapy.

Antibody responses to Haemophilus influenzae type B vaccines in men with human immunodeficiency virus infection.

BACKGROUND: Persons with human immunodeficiency virus (HIV) infection are at increased risk for serious infections caused by Haemophilus influenzae, yet there are few data on their antibody responses to the H. influenzae type b vaccines. METHODS: We evaluated antibody responses in 248 men who were randomly assigned to receive a single dose of either the H. influenzae type b polysaccharide (PRP) vaccine or the polysaccharide-mutant diphtheria toxoid conjugate vaccine (PRP-CRM). The subjects were stratified into four groups: seronegative men (67 subjects), men with asymptomatic HIV infection (79), men with symptomatic HIV infection (47), and men with the acquired immunodeficiency syndrome (AIDS) (55). RESULTS: Before immunization, the subjects with AIDS had the lowest PRP-antibody titers; 40 percent had titers below the putative protective level (less than 0.15 micrograms per milliliter). In the seronegative subjects, those with asymptomatic HIV infection, and those with symptomatic HIV infection, the PRP-CRM vaccine led to a threefold greater increase in geometric mean antibody titers than did the PRP vaccine (P less than 0.01). However, the subjects with AIDS had a greater antibody response to the PRP vaccine. The antibody response of HIV-seropositive men to the PRP-CRM vaccine correlated significantly with the number of CD4 lymphocytes (r = 0.47, P less than 0.0001, as compared with r = -0.01 for the PRP vaccine). In these HIV-infected men, both vaccines elicited the dominant anti-PRP idiotype described previously in populations not infected with HIV. CONCLUSIONS: Immunization with the PRP-CRM conjugate vaccine early in the course of HIV infection is likely to confer protection against disease caused by H. influenzae type b.