Search

The process of Nuclei detection in high-grade breast cancer images is quite challenging in the case of image processing techniques due to certain heterogeneous characteristics of cancer nuclei such as enlarged and irregularly shaped nuclei, highly coarse chromatin marginalized to the nuclei periphery and visible nucleoli. Recent reviews state that existing techniques show appreciable segmentation accuracy on breast histopathology images whose nuclei are dispersed and regular in texture and shape; however, typical cancer nuclei are often clustered and have irregular texture and shape properties. This paper proposes a novel segmentation algorithm for detecting individual nuclei from Hematoxylin and Eosin (H&E) stained breast histopathology images. This detection framework estimates a nuclei saliency map using tensor voting followed by boundary extraction of the nuclei on the saliency map using a Loopy Back Propagation (LBP) algorithm on a Markov Random Field (MRF). The method was tested on both whole-slide images and frames of breast cancer histopathology images. Experimental results demonstrate high segmentation performance with efficient precision, recall and dice-coefficient rates, upon testing high-grade breast cancer images containing several thousand nuclei. In addition to the optimal performance on the highly complex images presented in this paper, this method also gave appreciable results in comparison with two recently published methods—Wienert et al. (2012) and Veta et al. (2013), which were tested using their own datasets. [ABSTRACT FROM AUTHOR]

Background: Tumor proliferation speed, most commonly assessed by counting of mitotic figures in histological slide preparations, is an important biomarker for breast cancer. Although mitosis counting is routinely performed by pathologists, it is a tedious and subjective task with poor reproducibility, particularly among non-experts. Inter- and intraobserver reproducibility of mitosis counting can be improved when a strict protocol is defined and followed. Previous studies have examined only the agreement in terms of the mitotic count or the mitotic activity score. Studies of the observer agreement at the level of individual objects, which can provide more insight into the procedure, have not been performed thus far. Methods: The development of automatic mitosis detection methods has received large interest in recent years. Automatic image analysis is viewed as a solution for the problem of subjectivity of mitosis counting by pathologists. In this paper we describe the results from an interobserver agreement study between three human observers and an automatic method, and make two unique contributions. For the first time, we present an analysis of the object-level interobserver agreement on mitosis counting. Furthermore, we train an automatic mitosis detection method that is robust with respect to staining appearance variability and compare it with the performance of expert observers on an “external” dataset, i.e. on histopathology images that originate from pathology labs other than the pathology lab that provided the training data for the automatic method. Results: The object-level interobserver study revealed that pathologists often do not agree on individual objects, even if this is not reflected in the mitotic count. The disagreement is larger for objects from smaller size, which suggests that adding a size constraint in the mitosis counting protocol can improve reproducibility. The automatic mitosis detection method can perform mitosis counting in an unbiased way, with substantial agreement with human experts. [ABSTRACT FROM AUTHOR]

Precise detection of invasive cancer on whole-slide images (WSI) is a critical first step in digital pathology tasks of diagnosis and grading. Convolutional neural network (CNN) is the most popular representation learning method for computer vision tasks, which have been successfully applied in digital pathology, including tumor and mitosis detection. However, CNNs are typically only tenable with relatively small image sizes (200 × 200 pixels). Only recently, Fully convolutional networks (FCN) are able to deal with larger image sizes (500 × 500 pixels) for semantic segmentation. Hence, the direct application of CNNs to WSI is not computationally feasible because for a WSI, a CNN would require billions or trillions of parameters. To alleviate this issue, this paper presents a novel method, High-throughput Adaptive Sampling for whole-slide Histopathology Image analysis (HASHI), which involves: i) a new efficient adaptive sampling method based on probability gradient and quasi-Monte Carlo sampling, and, ii) a powerful representation learning classifier based on CNNs. We applied HASHI to automated detection of invasive breast cancer on WSI. HASHI was trained and validated using three different data cohorts involving near 500 cases and then independently tested on 195 studies from The Cancer Genome Atlas. The results show that (1) the adaptive sampling method is an effective strategy to deal with WSI without compromising prediction accuracy by obtaining comparative results of a dense sampling (∼6 million of samples in 24 hours) with far fewer samples (∼2,000 samples in 1 minute), and (2) on an independent test dataset, HASHI is effective and robust to data from multiple sites, scanners, and platforms, achieving an average Dice coefficient of 76%. [ABSTRACT FROM AUTHOR]

Objectives: The extent of peritumoral tumor cell infiltrations in glioblastoma contributes to poor prognosis. We aimed to assess additive prognostic value of Minkowski functionals in analyzing heterogeneity of peritumoral hyperintensity on T2WI in glioblastoma patients. Methods: Clinical data (age, sex, extent of surgical resection), O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status and pre-operative T2WI of 113 pathologically confirmed glioblastoma patients (from our institution, n = 61; from the Cancer Imaging Archive, n = 52) were retrospectively reviewed. The patients were randomly grouped into a training set (n = 80) and a test set (n = 33). Peritumoral T2 hyperintensity was manually segmented and Minkowski functionals—a texture analysis method capturing heterogeneity of MR images—were computed as a function of 11 grayscale thresholds. The Cox proportional hazards models were fitted with clinical variables, Minkowski functionals features as well as both combined. The risk prediction performances of the Minkowski functionals and combined models were validated on a separate test dataset. The sex-specific survival difference of the entire cohort was analyzed according to MGMT methylation status via Kaplan-Meier survival curves. Results: Thirty-three Minkowski features (11 area, 11 perimeter and 11 genus) for each patient were acquired giving a total of 3729 features. Cox regression models fitted with clinical data, Minkowski features, and both combined had incremental concordance indices of 0.577 (P = 0.02), 0.706 (P = 0.02) and 0.714 (P = 0.01) respectively. The prediction error rate of the combined model—having clinical and Minkowski features—was lower than that of Minkowski functionals model (0.135 and 0.161, respectively) when validated on a test dataset. No sex-specific survival difference was found according to MGMT methylation status (male, P = 0.2; female, P = 0.22). Conclusions: Minkowski functionals features computed from peritumoral hyperintensity can capture heterogeneity of glioblastoma on T2WI and have additive prognostic value in predicting survival, demonstrating their potential in complementing currently available prognostic parameters. [ABSTRACT FROM AUTHOR]

The Ki-67 labeling index (LI) is an important prognostic factor in breast carcinoma. The Ki-67 LI is traditionally calculated via unaided microscopic estimation; however, inter-observer and intra-observer variability and low reproducibility are problems with this visual assessment (VA) method. For more accurate assessment and better reproducibility with Ki-67 LI, digital image analysis was introduced recently. We used both VA and automated digital image analysis (ADIA) (Ventana Virtuoso image management software) to estimate Ki-67 LI for 997 cases of breast carcinoma, and compared VA and ADIA results. VA and ADIA were highly correlated (intraclass correlation coefficient 0.982, and Spearman’s correlation coefficient 0.966, p<0.05). We retrospectively analyzed cases with a greater than 5% difference between VA and ADIA results. The cause of these differences was: (1) tumor heterogeneity (98 cases, 56.0%), (2) VA interpretation error (32 cases, 18.3%), (3) misidentification of tumor cells (26 cases, 14.9%), (4) poor immunostaining or slide quality (16 cases, 9.1%), and (5) Estimation of non-tumor cells (3 cases, 1.7%). There were more discrepancies between VA and ADIA results in the group with a VA value of 10–20% compared to groups with <10% and ≥20%. Although ADIA is more accurate than VA, there are some limitations. Therefore, ADIA findings require confirmation by a pathologist. [ABSTRACT FROM AUTHOR]

The spectral imaging technique has been shown to provide more discriminative information than the RGB images and has been proposed for a range of problems. There are many studies demonstrating its potential for the analysis of histopathology images for abnormality detection but there have been discrepancies among previous studies as well. Many multispectral based methods have been proposed for histopathology images but the significance of the use of whole multispectral cube versus a subset of bands or a single band is still arguable. We performed comprehensive analysis using individual bands and different subsets of bands to determine the effectiveness of spectral information for determining the anomaly in colorectal images. Our multispectral colorectal dataset consists of four classes, each represented by infra-red spectrum bands in addition to the visual spectrum bands. We performed our analysis of spectral imaging by stratifying the abnormalities using both spatial and spectral information. For our experiments, we used a combination of texture descriptors with an ensemble classification approach that performed best on our dataset. We applied our method to another dataset and got comparable results with those obtained using the state-of-the-art method and convolutional neural network based method. Moreover, we explored the relationship of the number of bands with the problem complexity and found that higher number of bands is required for a complex task to achieve improved performance. Our results demonstrate a synergy between infra-red and visual spectrum by improving the classification accuracy (by 6%) on incorporating the infra-red representation. We also highlight the importance of how the dataset should be divided into training and testing set for evaluating the histopathology image-based approaches, which has not been considered in previous studies on multispectral histopathology images. [ABSTRACT FROM AUTHOR]

Digital imaging of H&E stained slides has enabled the application of image processing to support pathology workflows. Potential applications include computer-aided diagnostics, advanced quantification tools, and innovative visualization platforms. However, the intrinsic variability of biological tissue and the vast differences in tissue preparation protocols often lead to significant image variability that can hamper the effectiveness of these computational tools. We developed an alternative representation for H&E images that operates within a space that is more amenable to many of these image processing tools. The algorithm to derive this representation operates by exploiting the correlation between color and the spatial properties of the biological structures present in most H&E images. In this way, images are transformed into a structure-centric space in which images are segregated into tissue structure channels. We demonstrate that this framework can be extended to achieve color normalization, effectively reducing inter-slide variability. [ABSTRACT FROM AUTHOR]

Background: Antiangiogenic therapies for glioblastoma (GBM) such as bevacizumab (BVZ), have been unable to extend survival in large patient cohorts. However, a subset of patients having angiogenesis-dependent tumors might benefit from these therapies. Currently, there are no biomarkers allowing to discriminate responders from non-responders before the start of the therapy. Methods: 40 patients from the randomized GENOM009 study complied the inclusion criteria (quality of images, clinical data available). Of those, 23 patients received first line temozolomide (TMZ) for eight weeks and then concomitant radiotherapy and TMZ. 17 patients received BVZ+TMZ for seven weeks and then added radiotherapy to the treatment. Clinical variables were collected, tumors segmented and several geometrical measures computed including: Contrast enhancing (CE), necrotic, and total volumes; equivalent spherical CE width; several geometric measures of the CE ‘rim’ geometry and a set of image texture measures. The significance of the results was studied using Kaplan-Meier and Cox proportional hazards analysis. Correlations were assessed using Spearman correlation coefficients. Results: Kaplan-Meier and Cox proportional hazards analysis showed that total, CE and inner volume (p = 0.019, HR = 4.258) and geometric heterogeneity of the CE areas (p = 0.011, HR = 3.931) were significant parameters identifying response to BVZ. The group of patients with either regular CE areas (small geometric heterogeneity, median difference survival 15.88 months, p = 0.011) or those with small necrotic volume (median survival difference 14.50 months, p = 0.047) benefited substantially from BVZ. Conclusion: Imaging biomarkers related to the irregularity of contrast enhancing areas and the necrotic volume were able to discriminate GBM patients with a substantial survival benefit from BVZ. A prospective study is needed to validate our results. [ABSTRACT FROM AUTHOR]

Introduction: Confocal laser endomicroscopy (CLE) is becoming a popular method for optical biopsy of digestive mucosa for both diagnostic and therapeutic procedures. Computer aided diagnosis of CLE images, using image processing and fractal analysis can be used to quantify the histological structures in the CLE generated images. The aim of this study is to develop an automatic diagnosis algorithm of colorectal cancer (CRC), based on fractal analysis and neural network modeling of the CLE-generated colon mucosa images. Materials and Methods: We retrospectively analyzed a series of 1035 artifact-free endomicroscopy images, obtained during CLE examinations from normal mucosa (356 images) and tumor regions (679 images). The images were processed using a computer aided diagnosis (CAD) medical imaging system in order to obtain an automatic diagnosis. The CAD application includes image reading and processing functions, a module for fractal analysis, grey-level co-occurrence matrix (GLCM) computation module, and a feature identification module based on the Marching Squares and linear interpolation methods. A two-layer neural network was trained to automatically interpret the imaging data and diagnose the pathological samples based on the fractal dimension and the characteristic features of the biological tissues. Results: Normal colon mucosa is characterized by regular polyhedral crypt structures whereas malignant colon mucosa is characterized by irregular and interrupted crypts, which can be diagnosed by CAD. For this purpose, seven geometric parameters were defined for each image: fractal dimension, lacunarity, contrast correlation, energy, homogeneity, and feature number. Of the seven parameters only contrast, homogeneity and feature number were significantly different between normal and cancer samples. Next, a two-layer feed forward neural network was used to train and automatically diagnose the malignant samples, based on the seven parameters tested. The neural network operations were cross-entropy with the results: training: 0.53, validation: 1.17, testing: 1.17, and percent error, resulting: training: 16.14, validation: 17.42, testing: 15.48. The diagnosis accuracy error was 15.5%. Conclusions: Computed aided diagnosis via fractal analysis of glandular structures can complement the traditional histological and minimally invasive imaging methods. A larger dataset from colorectal and other pathologies should be used to further validate the diagnostic power of the method. [ABSTRACT FROM AUTHOR]

To retrospectively determine the association between breast lesion morphology and malignancy and to determine the optimal value of lesion enhancement (HU, Hounsfield units) to improve the diagnostic accuracy of breast cancer in patients with incidental breast lesions (IBLs). A total of 97 patients with 102 IBLs detected from July 2009 to December 2012 were enrolled in this study. Two radiologists analyzed CT images for the presence of malignancy based on the morphology of the lesions alone and in combination with an enhancement value (HU) analysis. There were 36 malignant and 66 benign IBLs. When the morphology and enhancement values were combined, the sensitivity, specificity, and accuracy were 92%, 97%, and 95%, respectively, for reader 1 and 89%, 94%, and 92%, respectively, for reader 2. The addition of HU values led to correct changes in the diagnosis; specifically, the accuracy of the diagnosis of reader 1 and reader 2 improved by 6.9% and 11.8%, respectively. The addition of the enhancement value (HU) to the CT morphology improved the diagnostic accuracy in the differentiation of malignant from benign IBLs by using the region of interest (ROI) to measure the HU within the most suspicious part of the lesion. [ABSTRACT FROM AUTHOR]

Background: Ki67 labeling index (LI) is critical for treatment options and prognosis evaluation in breast cancer. Visual assessment (VA) is widely used to assess Ki67 LI, but has some limitations. In this study, we compared the consistency between VA and automated digital image analysis (DIA) of Ki67 LI in breast cancer, and to evaluate the application value of DIA in Ki67 LI assessment. Methods: Ki67 immunostained slides of 155 cases of primary invasive breast cancer were eyeballing assessed by five breast pathologists and automated digital image analyzed by one breast pathologist respectively. Two score methods, hot-spot score and average score, were used to choose score areas. The intra-class correlation coefficient (ICC) was used to analyze the consistency between VA and DIA, and Wilcoxon signed-rank test was used to compare the median of paired-difference between VA and DIA values. Results: (1) A perfect agreement was demonstrated between VA and DIA of Ki67 LI by ICC analysis (P<0.0001) in the whole cohort. A perfect agreement between VA and DIA of Ki67 LI was also showed in G2-G3, ER positive/HER2 negative cases. Average score and hot-spot score methods both demonstrated a perfect concordance between VA and DIA of Ki67 LI. (2) All cases were classified into three groups by VA values (≤10%, 11%-30% and >30% Ki67 LI). The concordance was relatively lower in intermediate Ki67 LI group (11%-30%) compared with high (>30%) Ki67 LI groups according to both methods. (3) All cases were classified into three groups by paired-difference (d) between VA values of hot-spot score and average score (d<5, 5≤d<10, d≥10) to evaluate the correlation between Ki67 staining distribution (heterogeneous or homogenous) and reproducibility of assessment. A perfect agreement was all demonstrated in three groups, and a slightly better Ki67 LI agreement between VA and DIA was indicated in homogenous staining slides than in heterogeneous staining ones. (4) VA values were relatively smaller than DIA values (average score: median of paired-difference -3.72; hot-spot score: median of paired-difference -9.12). Conclusions: An excellent agreement between VA and DIA of Ki67 LI in breast cancer was demonstrated in the whole mixed cohort, suggesting that VA and DIA both could be used to assess Ki67 LI in clinical practice. Average score and hot-spot score methods both demonstrated a perfect concordance between VA and DIA of Ki67 LI. The almost perfect agreement between VA and DIA was observed in high Ki67 LI cases, displaying a homogenous staining pattern. The consistency between VA and DIA was relatively low in intermediate Ki67 LI group. The heterogeneity of tumors may slightly affect the concordance between VA and DIA of Ki67 LI. Assessment of VA provides lower Ki67 values than DIA, the biological importance of these values are not known at the moment. [ABSTRACT FROM AUTHOR]

Purpose: To assess the measurement variability of subsolid nodules (SSNs) in follow-up situations and to compare the degree of variability between measurement metrics. Methods: Two same-day repeat-CT scans of 69 patients (24 men and 45 women) with 69 SSNs were randomly assigned as initial or follow-up scans and were read by the same (situation 1) or different readers (situation 2). SSN size and solid portion size were measured in both situations. Measurement variability was calculated and coefficients of variation were used for comparisons. Results: Measurement variability for the longest and average diameter of SSNs was ±1.3 mm (±13.0%) and ±1.3 mm (±14.4%) in situation 1, and ±2.2 mm (±21.0%) and ±2.1 mm (±21.3%) in situation 2, respectively. For solid portion, measurement variability on lung and mediastinal windows was ±1.2 mm (±27.1%) and ±0.8 mm (±24.0%) in situation 1, and ±3.7 mm (±61.0%) and ±1.5 mm (±47.3%) in situation 2, respectively. There were no significant differences in the degree of variability between the longest and average diameters and between the lung and mediastinal window settings (p>0.05). However, measurement variability significantly increased when the follow-up and initial CT readers were different (p<0.001). Conclusions: A cutoff of ±2.2 mm can be reliably used to determine true nodule growth on follow-up CT. Solid portion measurements were not reliable in evaluating SSNs’ change when readers of initial and follow-up CT were different. [ABSTRACT FROM AUTHOR]