A Java application for tissue section image analysis

Multi-scale tissue architecture analysis of favorable-risk prostate cancer: Correlation with biochemical recurrence

Purpose

Prostate cancer (PCa) with biopsy-based grade group (GG) 1 or 2 characteristics has a favorable outcome, yet some cases still progress after radical prostatectomy and present with biochemical recurrence (BCR). We hypothesized that the multi-scale tissue architecture (MSTA) analysis score would correlate with the aggressive PCa phenotype and could be used as a tool for risk assessment to improve the management of patients with favorable-risk PCa.

Materials and Methods

MSTA was evaluated in needle-biopsy samples from 115 patients with favorable-risk PCa, as defined by GG1 and GG2, a prostate-specific antigen (PSA) level of <10 ng/mL, a clinical stage of cT1c to cT2b, and general Gleason GG (GGG) and expert pathologist-assessed GG (EGG). Algorithms based on Voronoi diagrams were applied to all Feulgen-thionin-stained diagnostic areas. One hundred tissue architecture features were calculated and an MSTA score, a linear combination of the most discriminant features, was generated. Correlation of MSTA score with BCR and other clinical variables was investigated.

Results

In a univariate regression model, EGG, clinical stage, and MSTA were significant predictors of BCR (respective p-values: 0.0016, 0.016, and 0.028). Survival analysis showed that patients with a high MSTA score were more likely to experience BCR than were patients with a low MSTA score (odds ratio, 2.9). Combining MSTA with GG assessment resulted in a significant stratification of risk for BCR.

Conclusions

MSTA score could be used as an objective adjunct risk stratification tool to pathologist assessments and could improve the management of patients with favorable-risk PCa.

Identification of Malignancy-Associated Changes in Histologically Normal Tumor-Adjacent Epithelium of Patients with HPV-Positive Oropharyngeal Cancer

The incidence of HPV-positive oropharyngeal cancer (HPV+ OPC) is increasing, thus presenting new challenges for disease detection and management. Noninvasive methods involving brush biopsies of diseased tissues were recently reported as insufficient for tumor detection in HPV+ OPC patients, likely due to differences between the site of tumor initiation at the base of involuted crypts and the site of brush biopsy at the crypt surface. We hypothesized that histologically normal surface epithelial cells in the oropharynx contain changes in nuclear morphology that arise due to tumor proximity. We analyzed the nuclear phenotype of matched tumor, tumor-adjacent normal, and contralateral normal tissues from biopsies of nine HPV+ OPC patients. Measurements of 89 nuclear features were used to train a random forest-based classifier to discriminate between normal and tumor nuclei. We then extracted voting scores from the trained classifier, which classify nuclei on a continuous scale from zero (“normal-like”) to one (“tumor-like”). In each case, the average score of the adjacent normal nuclei was intermediate between the tumor and contralateral normal nuclei. These results provide evidence for the existence of phenotypic changes in histologically normal, tumor-adjacent surface epithelial cells, which could be used as brush biopsy-based biomarkers for HPV+ OPC detection.

Automatic labeling of molecular biomarkers of immunohistochemistry images using fully convolutional networks

This paper addresses the problem of quantifying biomarkers in multi-stained tissues based on the color and spatial information of microscopy images of the tissue. A deep learning-based method that can automatically localize and quantify the regions expressing biomarker(s) in any selected area on a whole slide image is proposed. The deep learning network, which we refer to as Whole Image (WI)-Net, is a fully convolutional network whose input is the true RGB color image of a tissue and output is a map showing the locations of each biomarker. The WI-Net relies on a different network, Nuclei (N)-Net, which is a convolutional neural network that classifies each nucleus separately according to the biomarker(s) it expresses. In this study, images of immunohistochemistry (IHC)-stained slides were collected and used. Images of nuclei (4679 RGB images) were manually labeled based on the expressing biomarkers in each nucleus (as p16 positive, Ki-67 positive, p16 and Ki-67 positive, p16 and Ki-67 negative). The labeled nuclei images were used to train the N-Net (obtaining an accuracy of 92% in a test set). The trained N-Net was then extended to WI-Net that generated a map of all biomarkers in any selected sub-image of the whole slide image acquired by the scanner (instead of classifying every nucleus image). The results of our method compare well with the manual labeling by humans (average F-score of 0.96). In addition, we carried a layer-based immunohistochemical analysis of cervical epithelium, and showed that our method can be used by pathologists to differentiate between different grades of cervical intraepithelial neoplasia by quantitatively assessing the percentage of proliferating cells in the different layers of HPV positive lesions.

Large-scale DNA organization is a prognostic marker of breast cancer survival

Breast cancer is the leading cause of cancer-related deaths among women worldwide. We investigated whether changes in large-scale DNA organization (LDO) of tumor epithelial nuclei are an indicator of the aggressiveness of the tumor. We tested our algorithm on a set of 172 duplicates TMA cores samples coming from 95 breast cancer patients. Thirty-five patients died of breast cancer, and 60 were still alive 10 years after surgery. Duplicates cores were used to create training and test set. The TMA slides were stained with Feulgen-thionin and imaged using our in-house high-resolution Imaging system. Automated segmentation of cell nuclei followed by manual selection of intact, in-focus nuclei resulted in an average of 50 cell nuclei per sample available for analysis. Using forward stepwise linear discriminant analysis, a combination of six features that combined linearly gave the best discrimination between the two groups of cells: cells collected from 'deceased' patients TMA specimens and cells collected from "survivors" patients TMA specimens. Five of these features measure the spatial organization of DNA chromatin. The resulting canonical score is named cell LDO score. A patient LDO score, percentage of cell nuclei with a cell LDO score higher than a predefined cutoff value, was processed for the specimens in the test set, and a cutoff value was defined to classify patients with a low or a high LDO score. Using this binary test, 82.1% of patients were correctly classified are "deceased" or "survivors," with a specificity of 79% and a sensitivity of 88%. The relative risk of death of an individual with a high LDO score was nine times higher than for a patient with a low LDO score. When testing the combination of LDO score, node status, histological grade, and tumor grade to predict breast cancer survival, LDO was the most significant predictor. LDO classification was also highly associated with survival for only grade 1 and 2 patients as well as for only grade 3 patients. Our result confirms the potential of LDO to measure phenotypic changes associated with more aggressive disease and could be evaluated to identify patients more likely to benefit from adjuvant therapies.

Automated tumor analysis for molecular profiling in lung cancer

The discovery and clinical application of molecular biomarkers in solid tumors, increasingly relies on nucleic acid extraction from FFPE tissue sections and subsequent molecular profiling. This in turn requires the pathological review of haematoxylin & eosin (H&E) stained slides, to ensure sample quality, tumor DNA sufficiency by visually estimating the percentage tumor nuclei and tumor annotation for manual macrodissection. In this study on NSCLC, we demonstrate considerable variation in tumor nuclei percentage between pathologists, potentially undermining the precision of NSCLC molecular evaluation and emphasising the need for quantitative tumor evaluation. We subsequently describe the development and validation of a system called TissueMark for automated tumor annotation and percentage tumor nuclei measurement in NSCLC using computerized image analysis. Evaluation of 245 NSCLC slides showed precise automated tumor annotation of cases using Tissuemark, strong concordance with manually drawn boundaries and identical EGFR mutational status, following manual macrodissection from the image analysis generated tumor boundaries. Automated analysis of cell counts for % tumor measurements by Tissuemark showed reduced variability and significant correlation (p < 0.001) with benchmark tumor cell counts. This study demonstrates a robust image analysis technology that can facilitate the automated quantitative analysis of tissue samples for molecular profiling in discovery and diagnostics.

Quantification of confocal fluorescence microscopy for the detection of cervical intraepithelial neoplasia

BACKGROUND

Cervical cancer remains a major health problem, especially in developing countries. Colposcopic examination is used to detect high-grade lesions in patients with a history of abnormal pap smears. New technologies are needed to improve the sensitivity and specificity of this technique. We propose to test the potential of fluorescence confocal microscopy to identify high-grade lesions.

METHODS

We examined the quantification of ex vivo confocal fluorescence microscopy to differentiate among normal cervical tissue, low-grade Cervical Intraepithelial Neoplasia (CIN), and high-grade CIN. We sought to (1) quantify nuclear morphology and tissue architecture features by analyzing images of cervical biopsies; and (2) determine the accuracy of high-grade CIN detection via confocal microscopy relative to the accuracy of detection by colposcopic impression. Forty-six biopsies obtained from colposcopically normal and abnormal cervical sites were evaluated. Confocal images were acquired at different depths from the epithelial surface and histological images were analyzed using in-house software.

RESULTS

The features calculated from the confocal images compared well with those features obtained from the histological images and histopathological reviews of the specimens (obtained by a gynecologic pathologist). The correlations between two of these features (the nuclear-cytoplasmic ratio and the average of three nearest Delaunay-neighbors distance) and the grade of dysplasia were higher than that of colposcopic impression. The sensitivity of detecting high-grade dysplasia by analysing images collected at the surface of the epithelium, and at 15 and 30 μm below the epithelial surface were respectively 100, 100, and 92 %.

CONCLUSIONS

Quantitative analysis of confocal fluorescence images showed its capacity for discriminating high-grade CIN lesions vs. low-grade CIN lesions and normal tissues, at different depth of imaging. This approach could be used to help clinicians identify high-grade CIN in clinical settings.

Geometrical and topological analysis of in vivo confocal microscopy images reveals dynamic maturation of epidermal structures during the first years of life

Reflectance confocal microscopy is successfully used in infant skin research. Infant skin structure, function, and composition are undergoing a maturation process. We aimed to uncover how the epidermal architecture and cellular topology change with time. Images were collected from three age groups of healthy infants between one and four years of age and adults. Cell centers were manually identified on the images at the stratum granulosum (SG) and stratum spinosum (SS) levels. Voronoi diagrams were used to calculate geometrical and topological parameters. Infant cell density is higher than that of adults and decreases with age. Projected cell area, cell perimeter, and average distance to the nearest neighbors increase with age but do so distinctly between the two layers. Structural entropy is different between the two strata, but remains constant with time. For all ages and layers, the distribution of the number of nearest neighbors is typical of a cooperator network architecture. The topological analysis provides evidence of the maturation process in infant skin. The differences between infant and adult are more pronounced in the SG than SS, while cell cooperation is evident in all cases of healthy skin examined.

Proflavine Hemisulfate as a Fluorescent Contrast Agent for Point-of-Care Cytology

Proflavine hemisulfate, an acridine-derived fluorescent dye, can be used as a rapid stain for cytologic examination of biological specimens. Proflavine fluorescently stains cell nuclei and cytoplasmic structures, owing to its small amphipathic structure and ability to intercalate DNA. In this manuscript, we demonstrated the use of proflavine as a rapid cytologic dye on a number of specimens, including normal exfoliated oral squamous cells, cultured human oral squamous carcinoma cells, and leukocytes derived from whole blood specimens using a custom-built, portable, LED-illuminated fluorescence microscope. No incubation time was needed after suspending cells in 0.01% (w/v) proflavine diluted in saline. Images of proflavine stained oral cells had clearly visible nuclei as well as granular cytoplasm, while stained leukocytes exhibited bright nuclei, and highlighted the multilobar nature of nuclei in neutrophils. We also demonstrated the utility of quantitative analysis of digital images of proflavine stained cells, which can be used to detect significant morphological differences between different cell types. Proflavine stained oral cells have well-defined nuclei and cell membranes which allowed for quantitative analysis of nuclear to cytoplasmic ratios, as well as image texture analysis to extract quantitative image features.

Evaluation of HPV infection and smoking status impacts on cell proliferation in epithelial layers of cervical neoplasia

Accurate cervical intra-epithelial neoplasia (CIN) lesion grading is needed for effective patient management. We applied computer-assisted scanning and analytic approaches to immuno-stained CIN lesion sections to more accurately delineate disease states and decipher cell proliferation impacts from HPV and smoking within individual epithelial layers. A patient cohort undergoing cervical screening was identified (n = 196) and biopsies of varying disease grades and with intact basement membranes and epithelial layers were obtained (n = 261). Specimens were sectioned, stained (Mib1), and scanned using a high-resolution imaging system. We achieved semi-automated delineation of proliferation status and epithelial cell layers using Otsu segmentation, manual image review, Voronoi tessellation, and immuno-staining. Data were interrogated against known status for HPV infection, smoking, and disease grade. We observed increased cell proliferation and decreased epithelial thickness with increased disease grade (when analyzing the epithelium at full thickness). Analysis within individual cell layers showed a ≥50% increase in cell proliferation for CIN2 vs. CIN1 lesions in higher epithelial layers (with minimal differences seen in basal/parabasal layers). Higher rates of proliferation for HPV-positive vs. -negative cases were seen in epithelial layers beyond the basal/parabasal layers in normal and CIN1 tissues. Comparing smokers vs. non-smokers, we observed increased cell proliferation in parabasal (low and high grade lesions) and basal layers (high grade only). In sum, we report CIN grade-specific differences in cell proliferation within individual epithelial layers. We also show HPV and smoking impacts on cell layer-specific proliferation. Our findings yield insight into CIN progression biology and demonstrate that rigorous, semi-automated imaging of histopathological specimens may be applied to improve disease grading accuracy.

Ex vivo confocal imaging with contrast agents for the detection of oral potentially malignant lesions

OBJECTIVES

We investigated the potential use of real-time confocal microscopy in the non-invasive detection of occult oral potentially malignant lesions. Our objectives were to select the best fluorescence contrast agent for cellular morphology enhancement, to build an atlas of confocal microscopic images of normal human oral mucosa, and to determine the accuracy of confocal microscopy to recognize oral high-grade dysplasia lesions on live human tissue.

MATERIALS AND METHODS

Five clinically used fluorescent contrast agents were tested in vitro on cultured human cells and validated ex vivo on human oral mucosa. Images acquired ex vivo from normal and diseased human oral biopsies with bench-top fluorescent confocal microscope were compared to conventional histology. Image analyzer software was used as an adjunct tool to objectively compare high-grade dysplasia versus low-grade dysplasia and normal epithelium.

RESULTS

Acriflavine Hydrochloride provided the best cellular contrast by preferentially staining the nuclei of the epithelium. Using topical application of Acriflavine Hydrochloride followed by confocal microscopy, we could define morphological characteristics of each cellular layer of the normal human oral mucosa, building an atlas of histology-like images. Applying this technique to diseased oral tissue specimen, we were also able to accurately diagnose the presence of high-grade dysplasia through the increased cellularity and changes in nuclear morphological features. Objective measurement of cellular density by quantitative image analysis was a strong discriminant to differentiate between high-grade dysplasia and low-grade dysplasia lesions.

CONCLUSIONS

Pending clinical investigation, real-time confocal microscopy may become a useful adjunct to detect precancerous lesions that are at high risk of cancer progression, direct biopsy and delineate excision margins.

Development of image-processing software for automatic segmentation of brain tumors in MR images

Most of the commercially available software for brain tumor segmentation have limited functionality and frequently lack the careful validation that is required for clinical studies. We have developed an image-analysis software package called ‘Prometheus,’ which performs neural system–based segmentation operations on MR images using pre-trained information. The software also has the capability to improve its segmentation performance by using the training module of the neural system. The aim of this article is to present the design and modules of this software. The segmentation module of Prometheus can be used primarily for image analysis in MR images. Prometheus was validated against manual segmentation by a radiologist and its mean sensitivity and specificity was found to be 85.71±4.89% and 93.2±2.87%, respectively. Similarly, the mean segmentation accuracy and mean correspondence ratio was found to be 92.35±3.37% and 0.78±0.046, respectively.

Optical technologies and molecular imaging for cervical neoplasia: a program project update

There is an urgent global need for effective and affordable approaches to cervical cancer screening and diagnosis. In developing nations, cervical malignancies remain the leading cause of cancer-related deaths in women. This reality may be difficult to accept given that these deaths are largely preventable; where cervical screening programs have been implemented, cervical cancer-related deaths have decreased dramatically. In developed countries, the challenges of cervical disease stem from high costs and overtreatment. The National Cancer Institute-funded Program Project is evaluating the applicability of optical technologies in cervical cancer. The mandate of the project is to create tools for disease detection and diagnosis that are inexpensive, require minimal expertise, are more accurate than existing modalities, and can be feasibly implemented in a variety of clinical settings. This article presents the status and long-term goals of the project.

Potential use of quantitative tissue phenotype to predict malignant risk for oral premalignant lesions

The importance of early diagnosis in improving mortality and morbidity rates of oral squamous cell carcinoma (SCC) has long been recognized. However, a major challenge for early diagnosis is our limited ability to differentiate oral premalignant lesions (OPL) at high risk of progressing into invasive SCC from those at low risk. We investigated the potential of quantitative tissue phenotype (QTP), measured by high-resolution image analysis, to identify severe dysplasia/carcinoma in situ (CIS; known to have an increased risk of progression) and to predict progression to cancer within hyperplasia or mild/moderate dysplasia. We generated a nuclear phenotype score (NPS), a combination of five nuclear morphometric features that best discriminate 4,027 "normal" nuclei (selected from 29 normal oral biopsies) from 4,298 "abnormal" nuclei (selected from 30 SCC biopsies). This NPS was then determined for a set of 69 OPLs. Severe dysplasia/CIS showed a significant increase in NPS compared with hyperplasia or mild/moderate dysplasia. However, within the latter group, elevated NPS was strongly associated with the presence of high-risk loss of heterozygosity (LOH) patterns. There was a statistical difference between NPS of hyperplasia or mild/moderate dysplasia that progressed to cancer and those that did not. Individuals with a high NPS had a 10-fold increase in relative risk of progression. In the multivariate Cox model, LOH and NPS together were the strongest predictors for cancer development. These data suggest that QTP could be used to identify lesions that require molecular evaluation and should be integrated with such approaches to facilitate the identification of hyperplasia or mild/moderate dysplasia OPLs at high risk of progression.

Validation of novel optical imaging technologies: the pathologists' view

Noninvasive optical imaging technology has the potential to improve the accuracy of disease detection and predict treatment response. Pathology provides the critical link between the biological basis of an image or spectral signature and clinical outcomes obtained through optical imaging. The validation of optical images and spectra requires both morphologic diagnosis from histopathology and parametric analysis of tissue features above and beyond the declared pathologic "diagnosis." Enhancement of optical imaging modalities with exogenously applied biomarkers also requires validation of the biological basis for molecular contrast. For an optical diagnostic or prognostic technology to be useful, it must be clinically important, independently informative, and of demonstrated beneficial value to patient care. Its usage must be standardized with regard to methods, interpretation, reproducibility, and reporting, in which the pathologist plays a key role. By providing insight into disease pathobiology, interpretive or quantitative analysis of tissue material, and expertise in molecular diagnosis, the pathologist should be an integral part of any team that is validating novel optical imaging modalities. This review will consider (1) the selection of validation biomarkers; (2) standardization in tissue processing, diagnosis, reporting, and quantitative analysis; (3) the role of the pathologist in study design; and (4) reference standards, controls, and interobserver variability.

Visualization and other emerging technologies as change makers for oral cancer prevention

The genomic era has fueled a rapid emergence of new information at the molecular level with a great potential for developing innovative approaches to detection, risk assessment, and management of oral cancers and premalignant disease. As yet, however, little research has been done on complementary approaches that would use different technology in conjunction with molecular approaches to create a rapid and cost-effective strategy for patient assessment and management. In our ongoing 8-year longitudinal study, a set of innovative technologies is being validated alone and in combination to best correlate with patient outcome. The plan is to use these devices in a step-by-step sequence to guide key clinicopathological decisions on patient risk and treatment. The devices include a hand-held visualization device that makes use of tissue autofluorescence to detect and delineate abnormal lesions and fields requiring follow-up, to be used in conjunction with optical contrast agents such as toluidine blue. In addition, two semi-automated high-resolution computer microscopy systems will be used to quantitate the protein expression phenotype of cell nuclei in tissue sections and exfoliated cell brushings. Previously identified risk-associated molecular changes are being used to validate these systems as well as to establish their place in a population-based triage program that will filter out high-risk cases in the community and funnel them to dysplasia clinics where higher-cost molecular tools will guide intervention. A critical development for the translation of this technology into community settings is the establishment of an effective methodology for education and training of health practitioners on the front lines.

Subvisual chromatin changes in cervical epithelium measured by texture image analysis and correlated with HPV

OBJECTIVES

In this study, we are testing the hypothesis that human papillomavirus (HPV) positivity is correlated with chromatin texture in the cell. Interim analyses are important since this study involves 2000 patients and generates 6000 biopsy specimens that will be subjected to quantitative histopathological analysis and correlated to HPV positivity as measured by the Hybrid Capture II test (Digene; Gaithersberg, MD) and both HPV-DNA and mRNA by the polymerase chain reaction (PCR). The studies of optical technologies, from which we derive this sample, use the colposcopically directed and histopathologically classified cervical biopsy as the gold standard. In this report, we describe the results of an interim analysis of quantitative histopathology and chromatin texture as correlates of HPV infection using the cyto-savant system in cytologically and histopathologically negative specimens.

METHODS

A group of 1544 patients entered the optical technology trials, generating 3275 biopsies and 1544 Papanicolaou readings. Two hundred forty-eight patients were cytologically and histopathologically negative. Study pathologists reviewed histologic samples 3 times in a blinded fashion. Non-overlapping, quantitatively stained nuclei were selected from the samples by the pathologists. HPV testing was done using the PCR method and the Hybrid Capture II test. Statistical analysis involved the creation of a classification matrix using a linear discriminant analysis. The matrix was trained on HPV-positive cells by PCR. The analysis included the random creation of both a training set and a validation set that were classified based on the discrimination score obtained by correlating nuclear texture with HPV positivity.

RESULTS

The sensitivity of the classification was 52-54% and the specificity was 77-78%. Overall, a 68% predicted accuracy was achieved for both the training set and the test set. The agreement of a test and training set shows that the sets created randomly are indeed similar, and that the discrimination score worked equally well in both sets of cells. Once a cell-by-cell algorithm for HPV positivity was derived, HPV positivity was recalculated on the basis of cell-by-cell texture features. HPV positivity was then recalculated on both a per-biopsy basis and a per-patient basis. For HPV 16 and 18, the positivity rate was 70% on a per-biopsy basis and 73% on a per-patient basis.

CONCLUSIONS

Although these results are preliminary, they suggest that texture features reflecting chromatin condensation may correlate with HPV positivity. The current sample is histologic, the analysis suggests that in a cytologic sample, HPV positivity could be detected or confirmed by texture features computed as part of an HPV-associated score. Additional biologic markers could be used as needed. While this study was performed on histologic samples, a study of cytologic samples would be more useful. Future studies will examine chromatin texture compared to HPV integration and mRNA HPV expression.

Classification using the cumulative log-odds in the quantitative pathologic diagnosis of adenocarcinoma of the cervix

INTRODUCTION

This study develops a method that discriminates between normal and cancerous tissue sections (i.e., populations of cells) using a statistical model applied to high-dimensional quantitative measurements made on a sample of cells.

MATERIALS AND METHODS

We use a cumulative log-odds model to create a score for a tissue section using the information from the cells within that tissue section. Then, a threshold is determined using receiver operating characteristic (ROC) curve analysis. The method was tested using data from cervical adenocarcinomas, adenocarcinoma in situ, and normal columnar tissue.

RESULTS

Using 120 potential features, we analyzed the data for staining-independent features. Twenty-two features were statistically significant. We then calculated the log-odds and created a score, followed by ROC curve analysis. The operating point which maximizes the sum of the specificity and sensitivity achieved a sensitivity of 100% with a specificity of 85%.

CONCLUSION

The cumulative log-odds performs well in classifying tissue sections using high-dimensional data measured at the cellular level, like that of quantitative pathology. This methodology potentially has applications in pathology, radiology, and optical technologies.