Eosin Images Research Articles

PD52-02 COMPUTER-EXTRACTED FEATURES OF GLAND MORPHOLOGY FROM DIGITAL TISSUE IMAGES IS COMPARABLE TO DECIPHER FOR PROGNOSIS OF BIOCHEMICAL RECURRENCE RISK POST-SURGERY

You have accessJournal of UrologyProstate Cancer: Markers II (PD52)1 Apr 2020PD52-02 COMPUTER-EXTRACTED FEATURES OF GLAND MORPHOLOGY FROM DIGITAL TISSUE IMAGES IS COMPARABLE TO DECIPHER FOR PROGNOSIS OF BIOCHEMICAL RECURRENCE RISK POST-SURGERY Patrick Leo*, Robin Elliott, Andrew Janowczyk, Nafiseh Janaki, Kaustav Bera, Rakesh Shiradkar, Ayah El-Fahmawi, Jessica Kim, Mohammed Shahait, Abhishek Shah, Hari Thulasidass, Ashutosh Tewari, Sanjay Gupta, Natalie Shih, Michael Feldman, Priti Lal, David Lee, and Anant Madabhushi Patrick Leo*Patrick Leo* More articles by this author , Robin ElliottRobin Elliott More articles by this author , Andrew JanowczykAndrew Janowczyk More articles by this author , Nafiseh JanakiNafiseh Janaki More articles by this author , Kaustav BeraKaustav Bera More articles by this author , Rakesh ShiradkarRakesh Shiradkar More articles by this author , Ayah El-FahmawiAyah El-Fahmawi More articles by this author , Jessica KimJessica Kim More articles by this author , Mohammed ShahaitMohammed Shahait More articles by this author , Abhishek ShahAbhishek Shah More articles by this author , Hari ThulasidassHari Thulasidass More articles by this author , Ashutosh TewariAshutosh Tewari More articles by this author , Sanjay GuptaSanjay Gupta More articles by this author , Natalie ShihNatalie Shih More articles by this author , Michael FeldmanMichael Feldman More articles by this author , Priti LalPriti Lal More articles by this author , David LeeDavid Lee More articles by this author , and Anant MadabhushiAnant Madabhushi More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000000954.02AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: It is currently clinically challenging to stratify patients’ risk of biochemical recurrence (BCR) of localized prostate cancer following radical prostatectomy (RP). Decipher, a 22-gene genomic classifier, is currently a part of NCCN guidelines to determine risk of metastasis and BCR following RP with improved accuracy beyond existing clinicopathologic factor based nomograms. Recently, computer extracted quantitative histomorphometric (QH) analysis of hematoxylin and eosin (H&E) images alone has shown increasing value in predicting risk of BCR following RP. In this work, we sought to compare the prognostic ability of QH against Decipher in BCR prognosis post-RP. As compared to Decipher, QH is non tissue-destructive, less time consuming, and cheaper. METHODS: A single diagnostic slide was collected from N=388 patients from three institutions and compose. Patients were split into training (N=215) and validation (N=173) sets. One institution was split across the training and testing set according to availability of Decipher results. Each slide was annotated for a single large, representative cancer region, from which 26 texture features were extracted. 41 training set patients were used to train a deep learning model for lumen segmentation, which was applied to all the cancerous regions. From the lumen segmentations, 216 features of lumen arrangement, shape, and disorder were extracted. Training set patients were used to fit an elastic-net penalized Cox regression model with these 242 features, which was applied to all patients. The Cox model output a risk score for each patient. This process was termed “Histotyping”. The concordance index (c-index) of Histotyping and Decipher in BCR-free survival were compared. RESULTS: The nine features selected by the Cox model included seven measures of average lumen shape and variation of lumen shape, one feature of lumen arrangement, and one feature of image texture. A greater variation across the tumor in lumen shape and density and a less smooth image texture were associated with increased BCR risk. Histotyping performed comparably to Decipher (c-index 0.66 vs. 0.70) in BCR prognosis. CONCLUSIONS: Computerized visual analysis of a routinely acquired H&E slides was found to perform comparably to Decipher in BCR prognosis. A limitation of this work is that Decipher is optimized for metastasis prognosis, though metastasis outcome information was not available for these patients. Source of Funding: National Cancer Institute, Department of Defense, National Science Foundation © 2020 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 203Issue Supplement 4April 2020Page: e1089-e1090 Advertisement Copyright & Permissions© 2020 by American Urological Association Education and Research, Inc.MetricsAuthor Information Patrick Leo* More articles by this author Robin Elliott More articles by this author Andrew Janowczyk More articles by this author Nafiseh Janaki More articles by this author Kaustav Bera More articles by this author Rakesh Shiradkar More articles by this author Ayah El-Fahmawi More articles by this author Jessica Kim More articles by this author Mohammed Shahait More articles by this author Abhishek Shah More articles by this author Hari Thulasidass More articles by this author Ashutosh Tewari More articles by this author Sanjay Gupta More articles by this author Natalie Shih More articles by this author Michael Feldman More articles by this author Priti Lal More articles by this author David Lee More articles by this author Anant Madabhushi More articles by this author Expand All Advertisement PDF downloadLoading ...

Multipatch-GLCM for Texture Feature Extraction on Classification of the Colon Histopathology Images using Deep Neural Network with GPU Acceleration

Cancer is one of the leading causes of death in the world. It is the main reason why research in this field becomes challenging. Not only for the pathologist but also from the view of a computer scientist. Hematoxylin and Eosin (H&E) images are the most common modalities used by the pathologist for cancer detection. The status of cancer with histopathology images can be classified based on the shape, morphology, intensity, and texture of the image. The use of full high-resolution histopathology images will take a longer time for the extraction of all information due to the huge amount of data. This study proposed advance texture extraction by multi-patch images pixel method with sliding windows that minimize loss of information in each pixel patch. We use texture feature Gray Level Co-Occurrence Matrix (GLCM) with a mean-shift filter as the data pre-processing of the images. The mean-shift filter is a low-pass filter technique that considers the surrounding pixels of the images. The proposed GLCM method is then trained using Deep Neural Networks (DNN) and compared to other classification techniques for benchmarking. For training, we use two hardware: NVIDIA GPU GTX-980 and TESLA K40c. According to the study, Deep Neural Network outperforms other classifiers with the highest accuracy and deviation standard 96.72±0.48 for four cross-validations. The additional information is that training using Theano framework is faster than Tensorflow for both in GTX-980 and Tesla K40c.

Computerized Image Analysis of Tumor Cell Nuclear Morphology Can Improve Patient Selection for Clinical Trials in Localized Clear Cell Renal Cell Carcinoma

Background: Clinicopathological scores are used to predict the likelihood of recurrence-free survival for patients with clear cell renal cell carcinoma (ccRCC) after surgery. These are fallible, particularly in the middle range. This inevitably means that a significant proportion of ccRCC patients who will not develop recurrent disease enroll into clinical trials. As an exemplar of using digital pathology, we sought to improve the predictive power of “recurrence free” designation in localized ccRCC patients, by precise measurement of ccRCC nuclear morphological features using computational image analysis, thereby replacing manual nuclear grade assessment. Materials and Methods: TNM 8 UICC pathological stage pT1-pT3 ccRCC cases were recruited in Scotland and in Singapore. A Leibovich score (LS) was calculated. Definiens Tissue studio® (Definiens GmbH, Munich) image analysis platform was used to measure tumor nuclear morphological features in digitized hematoxylin and eosin (H&E) images. Results: Replacing human-defined nuclear grade with computer-defined mean perimeter generated a modified Leibovich algorithm, improved overall specificity 0.86 from 0.76 in the training cohort. The greatest increase in specificity was seen in LS 5 and 6, which went from 0 to 0.57 and 0.40, respectively. The modified Leibovich algorithm increased the specificity from 0.84 to 0.94 in the validation cohort. Conclusions: CcRCC nuclear mean perimeter, measured by computational image analysis, together with tumor stage and size, node status and necrosis improved the accuracy of predicting recurrence-free in the localized ccRCC patients. This finding was validated in an ethnically different Singaporean cohort, despite the different H and E staining protocol and scanner used. This may be a useful patient selection tool for recruitment to multicenter studies, preventing some patients from receiving unnecessary additional treatment while reducing the number of patients required to achieve adequate power within neoadjuvant and adjuvant clinical studies.

Stain Removal Through Color Normalization of Haematoxylin and Eosin Images: A Review

Histopathology images are used for the diagnosis and examination of cancer cells. Preparing and scanning histopathology slides consist of numerous steps of which staining are an important process. But these staining procedures cause multitude problems as there may be variations in the slides due to multiple reasons. Color variations in histopathology images can occur due to inconsistent staining of biopsy tissue, color responses of different scanners or difference in raw materials and manufacturing techniques used by stain. These factors hamper the generalization of automatic image analysis method. Hence there is a need for standardizing the image before analysis by performing stain normalization which is achieved by removing the stains for visual enhancement. Color normalization techniques play a vital in the developing computerized decision support system. In this paper, a detailed study and performance evaluation of various color normalization techniques such as Histogram Specification, Macenko Method, Reinhard Method on histopathology images are presented. In these normalization procedures the mean color of the target image is transferred onto the source image. Quality performance of different stain normalization techniques is evaluated based on Entropy and Structure Similarity Index Measure (SSIM). In this work, we record the time complexity of the color normalization algorithms. The normalization techniques are tested on histopathological images from UCSB dataset and Mitosis Atypia 2014 dataset. This article reviews and summarizes the color normalization techniques on histopathological images for visual improvement.

Computer-Aided Diagnosis in Histopathological Images of the Endometrium Using a Convolutional Neural Network and Attention Mechanisms.

Uterine cancer (also known as endometrial cancer) can seriously affect the female reproductive system, and histopathological image analysis is the gold standard for diagnosing endometrial cancer. Due to the limited ability to model the complicated relationships between histopathological images and their interpretations, existing computer-aided diagnosis (CAD) approaches using traditional machine learning algorithms often failed to achieve satisfying results. In this study, we develop a CAD approach based on a convolutional neural network (CNN) and attention mechanisms, called HIENet. In the ten-fold cross-validation on ∼3,300 hematoxylin and eosin (H&E) image patches from ∼500 endometrial specimens, HIENet achieved a 76.91 ± 1.17% (mean ± s. d.) accuracy for four classes of endometrial tissue, i.e., normal endometrium, endometrial polyp, endometrial hyperplasia, and endometrial adenocarcinoma. Also, HIENet obtained an area-under-the-curve (AUC) of 0.9579 ± 0.0103 with an 81.04 ± 3.87% sensitivity and 94.78 ± 0.87% specificity in a binary classification task that detected endometrioid adenocarcinoma. Besides, in the external validation on 200 H&E image patches from 50 randomly-selected female patients, HIENet achieved an 84.50% accuracy in the four-class classification task, as well as an AUC of 0.9829 with a 77.97% (95% confidence interval, CI, 65.27%∼87.71%) sensitivity and 100% (95% CI, 97.42%∼100.00%) specificity. The proposed CAD method outperformed three human experts and five CNN-based classifiers regarding overall classification performance. It was also able to provide pathologists better interpretability of diagnoses by highlighting the histopathological correlations of local pixel-level image features to morphological characteristics of endometrial tissue.

Automated Identification and Quantification of Signals in Multichannel Immunofluorescence Images: The SignalFinder-IF Platform

Multimarker fluorescence analysis of tissue specimens offers the opportunity to probe the expression levels and locations of multiple markers in a single sample. Software is needed to fully capitalize on the advantages of this technology for sensitive, quantitative, and multiplexed data collection. A major challenge has been the automated identification and quantification of signals. We report on the software SignalFinder-IF, which meets that need. SignalFinder-IF uses a newly developed algorithm called Segment-Fit Thresholding, which showed robust performance for automated signal identification in side-by-side comparisons with several current methods. Two utilities provided with SignalFinder-IF enable downstream analyses. The first allows the quantification and mapping of relationships between an unlimited number of markers through user-defined sequences of AND, OR, and NOT operators. The second produces composite pictures of the signals or colocalization analysis on brightfield hematoxylin and eosin images, which is useful for understanding the morphologies and locations of cells meeting specific marker criteria. SignalFinder-IF enables high-throughput, rigorous analyses of whole-slide, multimarker data, and it promises to open new possibilities in many research and clinical applications.

Image quilting and wavelet fusion for creation of synthetic microscopy nuclei images

Background and ObjectiveIn this study a texture simulation methodology is proposed for composing synthetic tissue microscopy images that could serve as a quantitative gold standard for the evaluation of the reliability, accuracy and performance of segmentation algorithms in computer-aided diagnosis. MethodsA library of background and nuclei regions was generated using pre-segmented Haematoxylin and Eosin images of brain tumours. Background image samples were used as input to an image quilting algorithm that produced the synthetic background image. Randomly selected pre-segmented nuclei were randomly fused on the synthetic background using a wavelet-based fusion approach. To investigate whether the produced synthetic images are meaningful and similar to real world images, two different tests were performed, one qualitative by an experienced histopathologist and one quantitative using the normalized mutual information and the Kullback-Leibler tests. To illustrate the challenges that synthetic images may pose to object recognition algorithms, two segmentation methodologies were utilized for nuclei detection, one based on the Otsu thresholding and another based on the seeded region growing approach. ResultsResults showed a satisfactory to good resemblance of the synthetic with the real world images according to both qualitative and quantitative tests. The segmentation accuracy was slightly higher for the seeded region growing algorithm (87.2%) than the Otsu's algorithm (86.3%). ConclusionsSince we know the exact coordinates of the regions of interest within the synthesised images, these images could then serve as a ‘gold standard’ for evaluation of segmentation algorithms in computer-aided diagnosis in tissue microscopy.

Confocal scanning microscopy provides rapid, detailed intraoperative histological assessment of brain neoplasms: Experience with 106 cases

ObjectivesFrozen section histological analysis is currently the mainstay for intraprocedural tissue diagnosis during the resection of intracranial neoplasms and for evaluating tumor margins. However, frozen sections are time-consuming and often do not reveal the histological features needed for final diagnosis when compared with permanent sections. Confocal scanning microscopy (CSM) with certain stains may be a valuable technology that can add rapid and detailed histological assessment advantage for the neurosurgical operating room. This study describes potential advantages of CSM imaging of fresh human brain tumor tissues labeled with acriflavine (AF), acridine orange (AO), cresyl violet (CV), methylene blue (MB), and indocyanine green (ICG) within the neurosurgical operating room facility. Patients and methodsAcute slices from orthotopic human intracranial neoplasms were incubated with AF/AO and CV solutions for 10 s and 1 min respectively. Staining was also attempted with MB and ICG. Samples were imaged using a bench-top CSM system. Histopathologic features of corresponding CSM and permanent hematoxylin and eosin images were reviewed for each case. ResultsOf 106 cases, 30 were meningiomas, 19 gliomas, 13 pituitary adenomas, 9 metastases, 6 schwannomas, 4 ependymomas, and 25 other pathologies. CSM using rapid fluorophores (AF, AO, CV) revealed striking microvascular, cellular and subcellular structures that correlated with conventional histology. By rapidly staining and optically sectioning freshly resected tissue, images were generated for intraoperative consultations in less than one minute. With this technique, an entire resected tissue sample was imaged and digitally stored for tele-pathology and archiving. ConclusionCSM of fresh human brain tumor tissue provides clinically meaningful and rapid histopathological assessment much faster than frozen section. With appropriate stains, including specific cellular structure or antibody staining, CSM could improve the timeliness of intraoperative decision-making, and the neurosurgical-pathology workflow during resection of human brain tumors, ultimately improving patient care.

Dose-dependent Nerve Inflammatory Response to rhBMP-2 in a Rodent Spinal Nerve Model.

We developed a spinal nerve root wrapping rodent model to evaluate the relationship between recombinant human bone morphogenetic protein 2 (rhBMP-2) dosage and the degree of inflammation. To investigate the direct effects of recombinant human bone morphogenetic protein 2 (rhBMP-2) dosage and the degree of inflammation in rodent spinal nerve roots. rhBMP-2 is commonly used in clinical practice to augment spinal fusion. However, complications such as postoperative leg pain, and a higher rate of postoperative neurologic deficits have been reported. These may be attributable to the exposure of adjacent nerve roots to high doses of rhBMP-2. Eighteen rats were randomized into three groups as follows: Group 1: absorbable collagen sponge (ACS) + 10 μg rhBMP-2, Group 2: ACS + 1 μg rhBMP-2, and Group 3 ACS with 20 μL saline. The ACS containing rhBMP-2 or saline were then wrapped around the L5 nerve root and secured loosely with nonabsorbable sutures. At 1-week postoperation, the rats were sacrificed, and the L5 nerve root and dorsal root ganglion harvested for reverse transcription polymerase chain reaction (RT-PCR), histology and immunohistochemical staining. In our study, 10 μg rhBMP-2 induced a 10-fold increase in seroma compared with 1 μg group. Using RT-PCR, macrophage markers MIP3-α, and CD-68 were upregulated by 8- and 2-fold respectively in comparison with the saline group. Haematoxylin and eosin (H&E) images demonstrated disruption of nerve structures in the high dose 10 μg rhBMP-2, but not at 1 μg rhBMP-2 and with saline. High doses of rhBMP-2 induced neuroinflammation in a dose dependent manner, resulting in higher seroma volume, macrophage marker gene expressions, and higher proportions of immunohistochemically stained TNF-alpha and more macrophage infiltration. 2.

Computer-aided prognosis on breast cancer with hematoxylin and eosin histopathology images: A review.

With the advance of digital pathology, image analysis has begun to show its advantages in information analysis of hematoxylin and eosin histopathology images. Generally, histological features in hematoxylin and eosin images are measured to evaluate tumor grade and prognosis for breast cancer. This review summarized recent works in image analysis of hematoxylin and eosin histopathology images for breast cancer prognosis. First, prognostic factors for breast cancer based on hematoxylin and eosin histopathology images were summarized. Then, usual procedures of image analysis for breast cancer prognosis were systematically reviewed, including image acquisition, image preprocessing, image detection and segmentation, and feature extraction. Finally, the prognostic value of image features and image feature-based prognostic models was evaluated. Moreover, we discussed the issues of current analysis, and some directions for future research.

Correlation of breast tissue histology and optical signatures to improve margin assessment techniques.

Optical spectroscopy is sensitive to morphological composition and has potential applications in intraoperative margin assessment. Here, we evaluate ex vivo breast tissue and corresponding quantified hematoxylin & eosin images to correlate optical scattering signatures to tissue composition stratified by patient characteristics. Adipose sites (213) were characterized by their cell area and density. All other benign and malignant sites (181) were quantified using a grid method to determine composition. The relationships between mean reduced scattering coefficient (〈μs′〉), and % adipose, % collagen, % glands, adipocyte cell area, and adipocyte density were investigated. These relationships were further stratified by age, menopausal status, body mass index (BMI), and breast density. We identified a positive correlation between 〈μs′〉 and % collagen and a negative correlation between 〈μs′〉 and age and BMI. Increased collagen corresponded to increased 〈μs′〉 variability. In postmenopausal women, 〈μs′〉 was similar regardless of fibroglandular content. Contributions from collagen and glands to 〈μs′〉 were independent and equivalent in benign sites; glands showed a stronger positive correlation than collagen to 〈μs′〉 in malignant sites. Our data suggest that scattering could differentiate highly scattering malignant from benign tissues in postmenopausal women. The relationship between scattering and tissue composition will support improved scattering models and technologies to enhance intraoperative optical margin assessment.

Stratification of HPV-induced cervical pathology using the virally encoded molecular marker E4 in combination with p16 or MCM

High-risk human papillomavirus (HPV) types cause cervical lesions of varying severity, ranging from transient productive infections to high-grade neoplasia. Disease stratification requires the examination of lesional pathology, and possibly also the detection of biomarkers. P16INK4a and MCM are established surrogates of high-risk HPV E6/E7 activity, and can be extensively expressed in high-grade lesions. Here we have combined these two cellular biomarkers with detection of the abundant HPV-encoded E4 protein in order to identify both productive and transforming lesions. This approach has allowed us to distinguish true papillomavirus infections from similar pathologies, and has allowed us to divide the heterogeneous CIN2 category into those that are CIN1-like and express E4, and those that more closely resemble nonproductive CIN3. To achieve this, 530 lesional areas were evaluated according to standard pathology criteria and by using a multiple staining approach that allows us to superimpose biomarker patterns either singly or in combination onto an annotated hematoxylin and eosin (H&E) image. Conventional grading of neoplasia was established by review panel, and compared directly with the composite molecular pathology visualized on the same tissue section. The detection of E4 coincided with the onset of vacuolation, becoming abundant in koilocytes as the MCM marker declined and cells lost their defined nuclear margins as visualized by standard H&E staining. Of the dual marker approaches, p16INK4a and E4 appeared most promising, with E4 generally identifying areas of low-grade disease even when p16INK4a was present. Extensive p16INK4a expression usually coincided with an absence of E4 expression or its focal retention in sporadic cells within the lesion. Our results suggest that a straightforward molecular evaluation of HPV life-cycle deregulation in cervical neoplasia may help improve disease stratification, and that this can be achieved using complementary molecular biomarker pairs such as MCM/E4 or, more promisingly, p16INK4a/E4 as an adjunct to conventional pathology.

Sparse Non-negative Matrix Factorization (SNMF) based color unmixing for breast histopathological image analysis

Color deconvolution has emerged as a popular method for color unmixing as a pre-processing step for image analysis of digital pathology images. One deficiency of this approach is that the stain matrix is pre-defined which requires specific knowledge of the data. This paper presents an unsupervised Sparse Non-negative Matrix Factorization (SNMF) based approach for color unmixing. We evaluate this approach for color unmixing of breast pathology images. Compared to Non-negative Matrix Factorization (NMF), the sparseness constraint imposed on coefficient matrix aims to use more meaningful representation of color components for separating stained colors. In this work SNMF is leveraged for decomposing pure stained color in both Immunohistochemistry (IHC) and Hematoxylin and Eosin (H&E) images. SNMF is compared with Principle Component Analysis (PCA), Independent Component Analysis (ICA), Color Deconvolution (CD), and Non-negative Matrix Factorization (NMF) based approaches. SNMF demonstrated improved performance in decomposing brown diaminobenzidine (DAB) component from 36 IHC images as well as accurately segmenting about 1400 nuclei and 500 lymphocytes from H & E images.

Poor survival with wild-type TP53 ovarian cancer?

ObjectiveThe objective of this study is to investigate whether wild-type TP53 status in high-grade serous ovarian carcinoma is associated with poorer survival. MethodsClinical and genomic data of 316 sequenced samples from The Cancer Genome Atlas (TCGA) ovarian high-grade serous carcinoma study were downloaded from TCGA data portal. Association between wild-type TP53 and survival was analyzed with Kaplan Meier method and Cox regression. The diagnosis of high-grade serous carcinomas was evaluated by reviewing pathological reports and high-resolution hematoxylin and eosin (H&E) images from frozen sections. The authenticity of wild-type TP53 in these tumor samples was assessed by analyzing SNP array data with ASCAT algorithm, reverse phase protein array (RPPA) data and RNAseq data. ResultsFifteen patients with high grade serous ovarian carcinomas were identified to have wild-type TP53, which had significantly shorter survival and higher chemoresistance than those with mutated TP53. The authenticity of wild-type TP53 status in these fifteen patients was supported by SNP array, RPPA, and RNAseq data. Except four cases with mixed histology, the classification as high grade serous carcinomas was supported by pathological reports and H&E images. Using RNAseq data, it was found that EDA2R gene, a direct target of wild-type TP53, was highly up-regulated in samples with wild-type TP53 in comparison to samples with either nonsense or missense TP53 mutations. ConclusionAlthough patients with wild-type TP53 ovarian cancer were rare in the TCGA high grade ovarian serous carcinomas cohort, these patients appeared to have a poorer survival and were more chemoresistant than those with mutated TP53. Differentially expressed genes in these TP53 wild-type tumors may provide insight in the molecular mechanism in chemotherapy resistance.

Histological and ultrastructural findings of corneal tissue after failed descemet membrane endothelial keratoplasty

To investigate in a retrospective review the histological and ultrastructural findings after failed primary and early Descemet membrane endothelial keratoplasty (DMEK), propose possible pathomechanisms of graft failure and give clinical implications. The explanted grafts underwent light- and electromicroscopical investigations in eight failed DMEK cases. Haematoxylin - Eosin, periodic acid Schiff and Alcian blue stainings were performed. Special note was given to any residual stromal remnants, absence of endothelial cells, lamellar structure and 'activation' of keratocytes. Of the eight cases, six were re-DMEKs and two penetrating keratoplasties. Partial graft separation was seen in six and no graft separation in two of the cases. The average time-interval to the re-DMEK or penetrating keratoplasty was 4.6 months. Light and electron microscopy of the two explanted stromal specimens showed varying degrees of keratocyte activation. Endothelial cell loss was observed in essentially all explants with varying degrees and positive correlation with intraoperative difficulty. Assumed upside-down situations showed large areas of intact endothelial cells. In addition, a new layer, situated between the endothelial cell layer and the posterior nonbanded layer, was observed with loose intercellular structure. A loss of the endothelial cell layer of varying degrees and positive correlation with intraoperative difficulty are the prominent feature of primary and early DMEK graft failure. Of note is the upside-down situation, in which in some cases, the endothelial cell layer not only remains intact but also demonstrates metabolical activity in forming a novel cellular layer.

Distinctive lipid profiles of human breast cancer and adjacent normal tissues by desorption electrospray ionization mass spectrometry imaging.

1132 Background: Routine intra-operative distinction between normal breast tissue and tumor is currently not possible in breast conserving surgery (BCS). This limitation affects the success of surgery, resulting in up to 40% requiring more than one operative procedure. Desorption electrospray ionization mass spectrometry (DESI MS) has been successfully used to discriminate between normal and cancerous human tissues from anatomical sites such as the liver and brain. The aim of this proof of concept study was to determine the feasibility of using DESI MS imaging for tissue identification and differentiation of breast cancer versus normal tissue. Methods: DESI MS imaging was carried out on 14 human invasive breast cancer samples. Breast cancer and adjacent normal paired human tissue sections (margin of tumor, 2cm and 5 cm from tumor) from 14 patients undergoing mastectomy were flash frozen in liquid nitrogen, sectioned, and thaw mounted to glass slides. All samples were imaged using DESI MS at 200 μm imaging resolution. DESI MS images were overlaid and compared with hematoxylin and eosin (H&E) images of the same sections. Results: Discrimination between cancer and adjacent normal tissue was achieved on the basis of the spatial distribution and varying intensities of particular fatty acids and lipid species. Several fatty acids such as oleic acid (m/z 281) and arachidonic acid (m/z 303) displayed much greater signal intensities in the cancer specimen compared to low or undetectable intensities in normal tissue. The cancer margins delineated by the DESI MS images of these molecules were consistent with H and E images of the tumor edge. Cancerous tissue was distinguished from normal tissue based on the qualitative assessment of molecular signatures and the distinction was in agreement with expert histopathology evaluation in 85% of samples. Conclusions: Our findings offer proof of concept that examination and classification of breast normal and cancer tissue by mass spectrometry imaging is highly accurate. The results are encouraging for development of a MS-based method that could be utilized intra-operatively for rapid detection of residual cancer tissue in the lumpectomy bed in BCS.

A Hierarchical Clustering Algorithm Based Computer Aided Molecular Modeling with Haematoxylin & Eosin Images of Colon Cancer

Color Image segmentation plays a crucial role in many medical imaging applications. In this paper we present a novel algorithm for computer aided molecular modeling in MATLAB R2012, a real time simulation environment. This manuscript presents a novel algorithm for color image segmentation of Haematoxilin and Eosin images of colon cancer. With help of this algorithm we can classify the differences between benign and malignant tumor cells.

Statistical shape analysis of the rat hippocampus in epilepsy

In this study, we aimed to (1) propose landmarks for the hippocampus in a rat brain using an experimental study and (2) investigate hippocampus shape changes in a rat brain with epilepsy using the statistical shape analysis method. We have used the statistical shape analysis method to illustrate hippocampal shape deformation due to epilepsy. Statistical shape analysis is of increasing interest to the neuroimaging community because of its potential for locating morphological changes. Nineteen rat brains (ten healthy and nine epileptic) with hematoxylin and eosin images of the hippocampus were used. The results strongly indicated that the normalized hippocampal shape of the epileptic group was different from the nonepileptic group; deformation was seen most significantly in the medial regions of the cornu ammonis (CA1 and CA3) of the hippocampus. In conclusion, our landmark-based methodology detected regional differences in the hippocampus in epilepsy. This study may serve as an initial reference for future studies on shape alteration of the hippocampus associated with certain medical conditions.

A systems pathology model for predicting overall survival in patients with refractory, advanced non-small cell lung cancer (NSCLC) treated with gefitinib

18065 Background: The abundant expression of the epidermal growth factor receptor (EGFR) in a variety of solid tumors including non-small cell lung cancer (NSCLC), head and neck, breast, colon and brain has made it an attractive target for various selective molecular therapeutics, including the tyrosine kinase inhibitor gefitinib. The recent evidence of activating mutations in EGFR combined with clinical - demographic features has suggested that subgroups of patients with NSCLC are most likely to respond to selective therapies. We sought to determine whether the integration of clinical variables, tumor morphometry and quantitative protein profiles using support vector machines could identify a set of features which predicts overall survival in patients with NSCLC treated with gefitinib. Methods: We analyzed tumor samples from 109 patients with advanced refractory NSCLC treated with gefitinib. Formalin fixed, paraffin embedded tissue samples were evaluated with the following assays: Hematoxylin and Eosin image morphometry, EGFR DNA mutation analysis, EGFR protein immunohistochemistry and quantitative immunofluorescence with the following antibodies: CK18, Ki67, Caspase 3 activated, cd34, EGFR, phosphorylated-EGFR, phosphorylated-ERK, phosphorylated-AKT, PTEN, Cyclin D1, phosphorylated-m-TOR, PI3-K, VEGF, KDR (VEGFR2) and phosphorylated KDR. A predictive model was developed using support vector regression for censored data. Results: 4 of 87 patients had tyrosine kinase domain mutations in exons 19, 20 or 21. Utilizing 51 patients with complete data profiles (i.e. clinical, image morphometry and immunofluorescence), a model predicting overall survival was developed with a concordance index of 0.74. Poor performance status, poorly differentiated histology by morphometry and increased levels of activated caspase 3, phosphorylated KDR (VEGFR2) and cyclin D1 were associated with reduced survival. Conclusion: The integration of clinical, imaging and biomarker data identified a set of features which were associated with a more aggressive disease phenotype and resulted in overall poor survival. [Table: see text]