Detection of underlying characteristics of nuclear chromatin patterns of thyroid tumor cells using texture and factor analyses

Pathologic Image Classification of Flat Urothelial Lesions Using Pathologic Criteria-Based Deep Learning

OBJECTIVES

Pathologic diagnosis of flat urothelial lesions is subject to high interobserver variability. We expected that deep learning could improve the accuracy and consistency of such pathologic diagnosis, although the learning process is a black box. We therefore propose a new approach for pathologic image classification incorporating the diagnostic process of the pathologist into a deep learning method.

METHODS

A total of 267 H&E-stained slides of normal urothelium and urothelial lesions from 127 cases were examined. Six independent convolutional neural networks were trained to classify pathologic images according to six pathologic criteria. We then used these networks in the main training for the final diagnosis.

RESULTS

Compared with conventional manual analysis, our method significantly improved the classification accuracy of images of flat urothelial lesions. The automated classification showed almost perfect agreement (weighted κ = 0.98) with the consensus reading. In addition, our approach provides the advantages of reliable diagnosis corresponding to histologic interpretation.

CONCLUSIONS

We used deep learning to establish an automated subtype classifier for flat urothelial lesions that successfully combines traditional morphologic approaches and complex deep learning to achieve a learning mechanism that seems plausible to the pathologist.

Computer vision reveals hidden variables underlying NF-κB activation in single cells

Individual cells are heterogeneous when responding to environmental cues. Under an external signal, certain cells activate gene regulatory pathways, while others completely ignore that signal. Mechanisms underlying cellular heterogeneity are often inaccessible because experiments needed to study molecular states destroy the very states that we need to examine. Here, we developed an image-based support vector machine learning model to uncover variables controlling activation of the immune pathway nuclear factor κB (NF-κB). Computer vision analysis predicts the identity of cells that will respond to cytokine stimulation and shows that activation is predetermined by minute amounts of “leaky” NF-κB (p65:p50) localization to the nucleus. Mechanistic modeling revealed that the ratio of NF-κB to inhibitor of NF-κB predetermines leakiness and activation probability of cells. While cells transition between molecular states, they maintain their overall probabilities for NF-κB activation. Our results demonstrate how computer vision can find mechanisms behind heterogeneous single-cell activation under proinflammatory stimuli.

Follicular Lesions with Papillary Nuclear Characteristics: Differences in Chromatin Detected by Computerized Image Analysis

Objective

Follicular lesions of the thyroid with papillary carcinoma nuclear characteristics are classified as infiltrative follicular variant of papillary thyroid carcinoma-FVPTC (IFVPTC), encapsulated/well demarcated FVPTC with tumour capsular invasion (IEFVPTC), and the newly described category "non-invasive follicular thyroid neoplasm with papillary-like nuclear features" (NIFTP) formerly known as non-invasive encapsulated FVPTC. This study evaluated whether computerized image analysis can detect nuclear differences between these three tumour subtypes.

Methods

Slides with histological material from 15 cases of NIFTP and 33 cases of FVPTC subtypes (22 IEFVPTC, and 11 IFVPTC) were analyzed using the Image J image processing program. Tumour cells were compared for both nuclear morphometry and chromatin textural characteristics.

Results

Nuclei from NIFTP and IFVPTC tumours differed in terms of chromatin textural features (grey intensity): mean (92.37 ± 21.01 vs 72.99 ± 14.73, p = 0.02), median (84.93 ± 21.17 vs 65.18 ± 17.08, p = 0.02), standard deviation (47.77 ± 9.55 vs 39.39 ± 7.18; p = 0.02), and coefficient of variation of standard deviation (19.96 ± 4.01 vs 24.75 ± 3.31; p = 0.003). No differences were found in relation to IEFVPTC.

Conclusion

Computerized image analysis revealed differences in nuclear texture between NIFTP and IFVPTC, but not for IEFVPTC.

Potential of epigenetic events in human thyroid cancer

Thyroid cancer remains the highest prevailing endocrine malignancy, and its incidence rate has progressively increased in the previous years. Above 95% of thyroid tumor are follicular cells types of carcinoma in which are considered invasive type of tumor. The pathogenesis and molecular mechanism of thyroid tumors are yet remains elucidated, in spite of activating RET, RAS and BRAF carcinogenesis have been well introduced. Nemours molecular alterations have been defined and have revealed promise for their diagnostic, prognostic and therapeutic capacity but still need further confirmation. Among different types of mechanisms, the current article reviews the importance of epigenetic modifications in thyroid cancer. Increasing data from previous reports demonstrate that acquired epigenetic abnormalities together with genetic changes plays an important role in alteration of gene expression patterns. Aberrant DNA methylation has been well known in the CpG regions and profile of microRNAs (mi-RNAs) expression also involved in cancer development. In addition, the gene expression through epigenetic control contribution to thyroid cancer is analyzed and it is semi considered in the clinic. However the epigenetic of the thyroid cancer is yet remains in its early stages, and it carries encouraging potential thyroid cancer detections in its early stages, assessment of prognosis and targeted cancer treatment.

Follicular thyroid lesions: is there a discriminatory potential in the computerized nuclear analysis?

BACKGROUND

Computerized image analysis seems to represent a promising diagnostic possibility for thyroid tumors. Our aim was to evaluate the discriminatory diagnostic efficiency of computerized image analysis of cell nuclei from histological materials of follicular tumors.

METHODS

We studied paraffin-embedded materials from 42 follicular adenomas (FA), 47 follicular variants of papillary carcinomas (FVPC) and 20 follicular carcinomas (FC) by the software ImageJ. Based on the nuclear morphometry and chromatin texture, the samples were classified as FA, FC or FVPC using the Classification and Regression Trees method.

RESULTS

We observed high diagnostic sensitivity and specificity rates (FVPC: 89.4% and 100%; FC: 95.0% and 92.1%; FA: 90.5 and 95.5%, respectively). When the tumors were compared by pairs (FC vs FA, FVPC vs FA), 100% of the cases were classified correctly.

CONCLUSION

The computerized image analysis of nuclear features showed to be a useful diagnostic support tool for the histological differentiation between follicular adenomas, follicular variants of papillary carcinomas and follicular carcinomas.

The role of HDAC2 in chromatin remodelling and response to chemotherapy in ovarian cancer

Chromatin undergoes structural changes in response to extracellular and environmental signals. We observed changes in nuclear morphology in cancer tissue biopsied after chemotherapy and hypothesised that these DNA damage-induced changes are mediated by histone deacetylases (HDACs). Nuclear morphological changes in cell lines (PE01 and PE04 models) and a xenograft model (OV1002) were measured in response to platinum chemotherapy by image analysis of nuclear texture. HDAC2 expression increased in PEO1 cells treated with cisplatin at 24h, which was accompanied by increased expression of heterochromatin protein 1 (HP1). HDAC2 and HP1 expression were also increased after carboplatin treatment in the OV1002 carboplatin-sensitive xenograft model but not in the insensitive HOX424 model. Expression of DNA damage response pathways (pBRCA1, γH2AX, pATM, pATR) showed time-dependent changes after cisplatin treatment. HDAC2 knockdown by siRNA reduced HP1 expression, induced DNA double strand breaks (DSB) measured by γH2AX, and interfered with the activation of DNA damage response induced by cisplatin. Furthermore, HDAC2 depletion affected γH2AX foci formation, cell cycle distribution, and apoptosis triggered by cisplatin, and was additive to the inhibitory effect of cisplatin in cell lines. By inhibiting expression of HDAC2, reversible alterations in chromatin patterns during cisplatin treatment were observed. These results demonstrate quantifiable alterations in nuclear morphology after chemotherapy, and implicate HDAC2 in higher order chromatin changes and cellular DNA damage responses in ovarian cancer cells in vitro and in vivo.

A Fuzzy-C-Means-Clustering Approach: Quantifying Chromatin Pattern of Non-Neoplastic Cervical Squamous Cells

Despite the effectiveness of Pap-smear test in reducing the mortality rate due to cervical cancer, the criteria of the reporting standard of the Pap-smear test are mostly qualitative in nature. This study addresses the issue on how to define the criteria in a more quantitative and definite term. A negative Pap-smear test result, i.e. negative for intraepithelial lesion or malignancy (NILM), is qualitatively defined to have evenly distributed, finely granular chromatin in the nuclei of cervical squamous cells. To quantify this chromatin pattern, this study employed Fuzzy C-Means clustering as the segmentation technique, enabling different degrees of chromatin segmentation to be performed on sample images of non-neoplastic squamous cells. From the simulation results, a model representing the chromatin distribution of non-neoplastic cervical squamous cell is constructed with the following quantitative characteristics: at the best representative sensitivity level 4 based on statistical analysis and human experts' feedbacks, a nucleus of non-neoplastic squamous cell has an average of 67 chromatins with a total area of 10.827 μm2; the average distance between the nearest chromatin pair is 0.508 μm and the average eccentricity of the chromatin is 0.47.

Accurate diagnosis of thyroid follicular lesions from nuclear morphology using supervised learning

Follicular lesions of the thyroid remain significant diagnostic challenges in surgical pathology and cytology. The diagnosis often requires considerable resources and ancillary tests including immunohistochemistry, molecular studies, and expert consultation. Visual analyses of nuclear morphological features, generally speaking, have not been helpful in distinguishing this group of lesions. Here we describe a method for distinguishing between follicular lesions of the thyroid based on nuclear morphology. The method utilizes an optimal transport-based linear embedding for segmented nuclei, together with an adaptation of existing classification methods. We show the method outputs assignments (classification results) which are near perfectly correlated with the clinical diagnosis of several lesion types' lesions utilizing a database of 94 patients in total. Experimental comparisons also show the new method can significantly outperform standard numerical feature-type methods in terms of agreement with the clinical diagnosis gold standard. In addition, the new method could potentially be used to derive insights into biologically meaningful nuclear morphology differences in these lesions. Our methods could be incorporated into a tool for pathologists to aid in distinguishing between follicular lesions of the thyroid. In addition, these results could potentially provide nuclear morphological correlates of biological behavior and reduce health care costs by decreasing histotechnician and pathologist time and obviating the need for ancillary testing.

Discriminant analysis between malignant mesothelioma and reactive mesothelium using nuclear three-dimensional analysis is useful for morphologically suspicious cases

OBJECTIVE

Morphological discrimination between malignant mesothelioma (MM) and reactive mesothelium (RM) is often difficult. Stereological analysis of nuclear luminance using centrifuged smear samples from coelomic fluid and discriminant analysis based on Mahalanobis distance may help to more accurately discriminate between MM and RM. In the present study, discriminant analysis was conducted on cytological specimens using the auto-smear method in a blinded manner with regard to histological results.

STUDY DESIGN

Coelomic fluid samples of 28 cases, cytologically diagnosed using the auto-smear method, were analyzed to determine pixel counts, the number of focus layers, 3-dimensional variation in the coefficient of variation of nuclear luminance between the focus layers as well as roundness in about 30-50 atypical cell nuclei per case. These measurements were employed to determine malignancy based on Mahalanobis distance.

RESULTS

Discrimination rates were as high as 91.7% for MM and 82.7% for RM. The discrimination rates of MM with histology were >80% in 8 of 10 suspicious cases with the initial cytology.

CONCLUSION

Our method allowed accurate discrimination between MM and RM and provides a useful alternative for the diagnosis of suspicious cases where morphological diagnosis of malignancy is difficult.

Significant Correlation between Chromosomal Aberration and Nuclear Morphology in Urothelial Carcinoma

We aimed to identify whether there is any correlation between chromosomal/genetic changes, nuclear morphology and the histological grade of urothelial carcinomas of the urinary bladder. Morphometry and multicolour fluorescence in situ hybridisation (FISH) techniques were applied to 250 cells in five low-grade cases and 350 cells in seven high-grade cases of urothelial carcinoma. Compared with low-grade carcinomas, most high-grade cases showed larger and more variable nuclear size, more frequent polysomy of centromere enumeration probes (CEPs) 3, 7 and 17, and the loss of the 9p21 locus. The number of CEP signals in cells was increased as the nuclear area of the cells became larger. Cells with gains in two or more types of CEP had significantly larger nuclei than cells with normal FISH signal patterns. In conclusion, the present study indicates that there was a correlation between nuclear morphology and chromosomal/genetic changes which were related to histological grading. Thus, we show that differences in the chromosomal/genetic aberrations present in low- and high-grade tumours can affect not only nuclear morphology but also the histopathological and clinical behaviour of urothelial carcinomas.

An optimal transportation approach for nuclear structure-based pathology

Nuclear morphology and structure as visualized from histopathology microscopy images can yield important diagnostic clues in some benign and malignant tissue lesions. Precise quantitative information about nuclear structure and morphology, however, is currently not available for many diagnostic challenges. This is due, in part, to the lack of methods to quantify these differences from image data. We describe a method to characterize and contrast the distribution of nuclear structure in different tissue classes (normal, benign, cancer, etc.). The approach is based on quantifying chromatin morphology in different groups of cells using the optimal transportation (Kantorovich-Wasserstein) metric in combination with the Fisher discriminant analysis and multidimensional scaling techniques. We show that the optimal transportation metric is able to measure relevant biological information as it enables automatic determination of the class (e.g., normal versus cancer) of a set of nuclei. We show that the classification accuracies obtained using this metric are, on average, as good or better than those obtained utilizing a set of previously described numerical features. We apply our methods to two diagnostic challenges for surgical pathology: one in the liver and one in the thyroid. Results automatically computed using this technique show potentially biologically relevant differences in nuclear structure in liver and thyroid cancers.

Detection and classification of thyroid follicular lesions based on nuclear structure from histopathology images

Follicular lesions of the thyroid are traditionally difficult and tedious challenges in diagnostic surgical pathology in part due to lack of obvious discriminatory cytological and microarchitectural features. We describe a computerized method to detect and classify follicular adenoma of the thyroid, follicular carcinoma of the thyroid, and normal thyroid based on the nuclear chromatin distribution from digital images of tissue obtained by routine histological methods. Our method is based on determining whether a set of nuclei, obtained from histological images using automated image segmentation, is most similar to sets of nuclei obtained from normal or diseased tissues. This comparison is performed utilizing numerical features, a support vector machine, and a simple voting strategy. We also describe novel methods to identify unique and defining chromatin patterns pertaining to each class. Unlike previous attempts in detecting and classifying these thyroid lesions using computational imaging, our results show that our method can automatically classify the data pertaining to 10 different human cases with 100% accuracy after blind cross validation using at most 43 nuclei randomly selected from each patient. We conclude that nuclear structure alone contains enough information to automatically classify the normal thyroid, follicular carcinoma, and follicular adenoma, as long as groups of nuclei (instead of individual ones) are used. We also conclude that the distribution of nuclear size and chromatin concentration (how tightly packed it is) seem to be discriminating features between nuclei of follicular adenoma, follicular carcinoma, and normal thyroid.

Conservation and alteration of chromosome territory arrangements in thyroid carcinoma cell nuclei

Chromosome territories (CTs) are intranuclear subregions occupied by individual chromosomes in an interphase cell. In this study, we investigated intranuclear CT positionings of chromosomes 10 (CS10), 18 (CS18), and 19 (CS19) in epithelial cells from four normal thyroid tissue (NT), four adenomatous goiters (AGs), six papillary carcinomas (PCs), and two undifferentiated carcinomas (UCs) using the multicolor fluorescence in situ hybridization method. In the NT and AGs, the radial positionings of CS10 and CS18 were detected at the periphery of nuclei in more than 60% and 80% of cells, respectively, whereas the radial positioning of CS19 was in the central region of the nuclei in more than 80% of cells. In the PCs, radial positioning pattern of CS10 and CS18 were similar to that in the NT. The nuclei with centrally located CS19 in PCs were less frequent than those in NT cells (p < 0.01). On the other hand, UCs with cells having DNA amplification demonstrated the locational abnormalities of the CS10, CS18, and CS19 radial positions. These findings indicate that alteration of CT positioning could be related to DNA amplification and, morphologically, may explain the nuclear atypia that accompanies the abnormal chromatin feature.

Application of microscopic Forster resonance energy transfer to cytological diagnosis of the thyroid tumors

We propose a novel application of microscopic Forster resonance energy transfer (FRET) to clinical cytological diagnosis based on sensitive measurements of distance changes between fluorescently labeled deoxyribose nucleic acid (DNA) molecules. We have employed the microscopic FRET imaging for investigation of six papillary carcinomas and eight benign cases. In each case the FRET images of 20 cells stained by the AT-specific donor Hoechst 33258 and the GC-specific acceptor 7-aminoactinomycin D were acquired and analyzed by texture analysis. We have not found significant difference of the mean FRET efficiency between the benign and malignant groups. On the other hand, the texture analysis revealed a significant difference of the intranuclear spatial distribution of FRET efficiencies between the benign and malignant groups. The results indicate that despite the similar average distance between the AT- and the GC-rich DNA segments in the papillary carcinomas and the benign cases, the former has more heterogeneous distribution of the AT- and the GC-rich DNA segments in nuclei compared to the benign groups. We have demonstrated that the FRET imaging is a helpful tool for the medical cytological diagnosis of human tumors by giving information on the chromatin topology on the scale below the resolution of conventional optical microscopes. (c) 2005 Society of Photo-Optical Instrumentation Engineers.

Morphological abstraction of thyroid tumor cell nuclei using morphometry with factor analysis

Various morphonuclear studies by digital image analysis have successfully been applied to quantify the nuclear morphology, including chromatin distribution pattern, in cytology of various organs; however, the majority of past reports have not shown correlation between the quantitative data by digital image analysis and cytological findings in practical diagnosis. In this report, we present the usefulness of morphological abstraction to combine the objective data and subjective observation in cytological diagnosis. Randomly selected, 100 cells in each Papanicolaou-stained ABC smear samples of 39 benign and malignant thyroid tumor cases were studied. Gray-level image data provided seven parameters for nuclear size, four parameters for nuclear shape, and 16 parameters showing the nuclear chromatin patterns from high-dimensional texture analysis of using co-occurrence and run-length matrices. To statistically abstract nuclear morphology, factor analysis was used. Factor analysis classified morphological nuclear characters as abstraction parameter into five abstract parameters composed of nuclear size, shape, heterogeneity, and contrast and homogeneity of chromatin pattern. The nuclei of papillary carcinoma showed larger size, more irregular shape, and higher contrast of chromatin pattern than those of the benign group. The follicular carcinomas have larger nucleus in each cell and more monotonous chromatin pattern among cells in each case than those of the benign group. Morphological abstraction by morphometry with factor analysis may provide a practical approach to the detection of the underlying characteristics of nuclear morphology in aspiration biopsy cytology.

Spatial distribution analysis of AT- and GC-rich regions in nuclei using corrected fluorescence resonance energy transfer

We employed microscopic intensity-based fluorescence resonance energy transfer (FRET) images with correction by donor and acceptor concentrations to obtain unbiased maps of spatial distribution of the AT- and GC-rich DNA regions in nuclei. FRET images of 137 bovine aortic endothelial cells stained by the AT-specific donor Hoechst 33258 and the GC-specific acceptor 7-aminoactinomycin D were acquired and corrected for the donor and acceptor concentrations by the Gordon's method based on the three fluorescence filter sets. The corrected FRET images were quantitatively analyzed by texture analysis to correlate the spatial distribution of the AT- and GC-rich DNA regions with different phases of the cell cycle. Both visual observation and quantitative texture analysis revealed an increased number and size of the low FRET efficiency centers for cells in the G(2)/M-phases, compared to the G(1)-phase cells. We have detected cell cycle-dependent changes of the spatial organization and separation of the AT- and GC-rich DNA regions. Using the corrected FRET (cFRET) technique, we were able to detect early DNA separation stages in late interphase nuclei.