Comparative Study on Automated Cell Nuclei Segmentation Methods for Cytology Pleural Effusion Images

Convolutional Blur Attention Network for Cell Nuclei Segmentation

Accurately segmented nuclei are important, not only for cancer classification, but also for predicting treatment effectiveness and other biomedical applications. However, the diversity of cell types, various external factors, and illumination conditions make nucleus segmentation a challenging task. In this work, we present a new deep learning-based method for cell nucleus segmentation. The proposed convolutional blur attention (CBA) network consists of downsampling and upsampling procedures. A blur attention module and a blur pooling operation are used to retain the feature salience and avoid noise generation in the downsampling procedure. A pyramid blur pooling (PBP) module is proposed to capture the multi-scale information in the upsampling procedure. The superiority of the proposed method has been compared with a few prior segmentation models, namely U-Net, ENet, SegNet, LinkNet, and Mask RCNN on the 2018 Data Science Bowl (DSB) challenge dataset and the multi-organ nucleus segmentation (MoNuSeg) at MICCAI 2018. The Dice similarity coefficient and some evaluation matrices, such as F1 score, recall, precision, and average Jaccard index (AJI) were used to evaluate the segmentation efficiency of these models. Overall, the proposal method in this paper has the best performance, the AJI indicator on the DSB dataset and MoNuSeg is 0.8429, 0.7985, respectively.

Label-free identification of microplastics in human cells: dark-field microscopy and deep learning study

The development of an automatic method of identifying microplastic particles within live cells and organisms is crucial for high-throughput analysis of their biodistribution in toxicity studies. State-of-the-art technique in the data analysis tasks is the application of deep learning algorithms. Here, we propose the approach of polystyrene microparticle classification differing only in pigmentation using enhanced dark-field microscopy and a residual neural network (ResNet). The dataset consisting of 11,528 particle images has been collected to train and evaluate the neural network model. Human skin fibroblasts treated with microplastics were used as a model to study the ability of ResNet for classifying particles in a realistic biological experiment. As a result, the accuracy of the obtained classification algorithm achieved up to 93% in cell samples, indicating that the technique proposed will be a potent alternative to time-consuming spectral-based methods in microplastic toxicity research.

Comparative single-cell analysis of biopsies clarifies pathogenic mechanisms in Klinefelter syndrome

Klinefelter syndrome (KS), also known as 47, XXY, is characterized by a distinct set of physiological abnormalities, commonly including infertility. The molecular basis for Klinefelter-related infertility is still unclear, largely because of the cellular complexity of the testis and the intricate endocrine and paracrine signaling that regulates spermatogenesis. Here, we demonstrate an analysis framework for dissecting human testis pathology that uses comparative analysis of single-cell RNA-sequencing data from the biopsies of 12 human donors. By comparing donors from a range of ages and forms of infertility, we generate gene expression signatures that characterize normal testicular function and distinguish clinically distinct forms of male infertility. Unexpectedly, we identified a subpopulation of Sertoli cells within multiple individuals with KS that lack transcription from the XIST locus, and the consequence of this is increased X-linked gene expression compared to all other KS cell populations. By systematic assessment of known cell signaling pathways, we identify 72 pathways potentially active in testis, dozens of which appear upregulated in KS. Altogether our data support a model of pathogenic changes in interstitial cells cascading from loss of X inactivation in pubertal Sertoli cells and nominate dosage-sensitive factors secreted by Sertoli cells that may contribute to the process. Our findings demonstrate the value of comparative patient analysis in mapping genetic mechanisms of disease and identify an epigenetic phenomenon in KS Sertoli cells that may prove important for understanding causes of infertility and sex chromosome evolution.

Automatic wavelet-based 3D nuclei segmentation and analysis for multicellular embryo quantification

Identification of individual cells in tissues, organs, and in various developing systems is a well-studied problem because it is an essential part of objectively analyzing quantitative images in numerous biological contexts. We developed a size-dependent wavelet-based segmentation method that provides robust segmentation without any preprocessing, filtering or fine-tuning steps, and is robust to the signal-to-noise ratio. The wavelet-based method achieves robust segmentation results with respect to True Positive rate, Precision, and segmentation accuracy compared with other commonly used methods. We applied the segmentation program to zebrafish embryonic development IN TOTO for nuclei segmentation, image registration, and nuclei shape analysis. These new approaches to segmentation provide a means to carry out quantitative patterning analysis with single-cell precision throughout three dimensional tissues and embryos and they have a high tolerance for non-uniform and noisy image data sets.

Automated classification of mitotic catastrophe by use of the centromere fragmentation morphology

Mitotic catastrophe is a common mode of tumor cell death. Cancer cells with a defective cell-cycle checkpoint often enter mitosis with damaged or under replicated chromosomes following genotoxic treatment. Premature condensation of the under-replicated (or damaged) chromosomes results in double-stranded DNA breaks at the centromere (centromere fragmentation). Centromere fragmentation is a morphological marker of mitotic catastrophe and is distinguished by the clustering of centromeres away from the chromosomes. We present an automated 2-step system for segmentation of cells exhibiting centromere fragmentation. The first step segments individual cells from clumps. We added two new terms, weighted local repelling term (WLRt) and weighted gradient term (WGt), in the energy functional of the traditional Chan-Vese based level set method. WLRt was used to generate a repelling force when contours of adjacent cells merged and then penalized the overlap. WGt enhances gradients between overlapping cells. The second step consists of a new algorithm, SBaN (shape-based analysis of each nucleus), which extracts features like circularity, major-axis length, minor-axis length, area, and eccentricity from each chromosome to identify cells with centromere fragmentation. The performance of SBaN algorithm for centromere fragmentation detection was statistically evaluated and the results were robust.

Alginate Affects Bioactivity of Chimeric Collagen-Binding LL37 Antimicrobial Peptides Adsorbed to Collagen-Alginate Wound Dressings

Chronic infected wounds cause more than 23,000 deaths annually. Antibiotics and antiseptics are conventionally used to treat infected wounds; however, they can be toxic to mammalian cells, and their use can contribute to antimicrobial resistance. Antimicrobial peptides (AMPs) have been utilized to address the limitations of antiseptics and antibiotics. In previous work, we modified the human AMP LL37 with collagen-binding domains from collagenase (cCBD) or fibronectin (fCBD) to facilitate peptide tethering and delivery from collagen-based wound dressings. We found that cCBD-LL37 and fCBD-LL37 were retained and active when bound to 100% collagen scaffolds. Collagen wound dressings are commonly made as composites with other materials, such as alginate. The goal of this study was to investigate how the presence of alginate affects the tethering, release, and antimicrobial activity of LL37 and CBD-LL37 peptides adsorbed to commercially available collagen-alginate wound dressings (FIBRACOL Plus-a 90% collagen and 10% alginate wound dressing). We found that over 85% of the LL37, cCBD-LL37, and fCBD-LL37 was retained on FIBRACOL Plus over a 14-day release study (90.3, 85.8, and 98.6%, respectively). Additionally, FIBRACOL Plus samples loaded with peptides were bactericidal toward Pseudomonas aeruginosa, even after 14 days in release buffer but demonstrated no antimicrobial activity against Escherichia coli, Staphylococcus aureus, and Staphylococcus epidermidis. The presence of alginate in solution induced conformational changes in the cCBD-LL37 and LL37 peptides, resulting in increased peptide helicity, and reduced antimicrobial activity against P. aeruginosa. Peptide-loaded FIBRACOL Plus scaffolds were not cytotoxic to human dermal fibroblasts. This study demonstrates that CBD-mediated LL37 tethering is a viable strategy to reduce LL37 toxicity, and how substrate composition plays a crucial role in modulating the antimicrobial activity of tethered AMPs.

Automated oral squamous cell carcinoma identification using shape, texture and color features of whole image strips

Despite profound knowledge of the incidence of oral cancers and a large body of research beyond it, it continues to beat diagnosis and treatment management. Post physical observation by clinicians, a biopsy is a gold standard for accurate detection of any abnormalities. Towards the application of artificial intelligence as an aid to diagnosis, automated cell nuclei segmentation is the most essential step for the recognition of the cancer cells. In this study, we have extracted the shape, texture and color features from the histopathological images collected indigenously from regional hospitals. A dataset of 42 whole slide slices was used to automatically segment and generate a cell level dataset of 720 nuclei. Next, different classifiers were applied for classification purposes. 99.4 % accuracy using Decision Tree Classifier, 100 % accuracy using both SVM and Logistic regression and 100 % accuracy using SVM, Logistic regression and Linear Discriminant were acquired for shape, textural and color features respectively. The in-depth analysis showed SVM and Linear Discriminant classifier gave the best result for texture and color features respectively. The achieved result can be effectively converted to software as an assistant diagnostic tool.

Computer Aided Diagnosis System for Detection of Cancer Cells on Cytological Pleural Effusion Images

Cytological screening plays a vital role in the diagnosis of cancer from the microscope slides of pleural effusion specimens. However, this manual screening method is subjective and time-intensive and it suffers from inter- and intra-observer variations. In this study, we propose a novel Computer Aided Diagnosis (CAD) system for the detection of cancer cells in cytological pleural effusion (CPE) images. Firstly, intensity adjustment and median filtering methods were applied to improve image quality. Cell nuclei were extracted through a hybrid segmentation method based on the fusion of Simple Linear Iterative Clustering (SLIC) superpixels and K-Means clustering. A series of morphological operations were utilized to correct segmented nuclei boundaries and eliminate any false findings. A combination of shape analysis and contour concavity analysis was carried out to detect and split any overlapped nuclei into individual ones. After the cell nuclei were accurately delineated, we extracted 14 morphometric features, 6 colorimetric features, and 181 texture features from each nucleus. The texture features were derived from a combination of color components based first order statistics, gray level cooccurrence matrix and gray level run-length matrix. A novel hybrid feature selection method based on simulated annealing combined with an artificial neural network (SA-ANN) was developed to select the most discriminant and biologically interpretable features. An ensemble classifier of bagged decision trees was utilized as the classification model for differentiating cells into either benign or malignant using the selected features. The experiment was carried out on 125 CPE images containing more than 10500 cells. The proposed method achieved sensitivity of 87.97%, specificity of 99.40%, accuracy of 98.70%, and F-score of 87.79%.