Low Contrast Cell Research Articles

Analysis of WSI Images by Hybrid Systems with Fusion Features for Early Diagnosis of Cervical Cancer.

Cervical cancer is one of the most common types of malignant tumors in women. In addition, it causes death in the latter stages. Squamous cell carcinoma is the most common and aggressive form of cervical cancer and must be diagnosed early before it progresses to a dangerous stage. Liquid-based cytology (LBC) swabs are best and most commonly used for cervical cancer screening and are converted from glass slides to whole-slide images (WSIs) for computer-assisted analysis. Manual diagnosis by microscopes is limited and prone to manual errors, and tracking all cells is difficult. Therefore, the development of computational techniques is important as diagnosing many samples can be done automatically, quickly, and efficiently, which is beneficial for medical laboratories and medical professionals. This study aims to develop automated WSI image analysis models for early diagnosis of a cervical squamous cell dataset. Several systems have been designed to analyze WSI images and accurately distinguish cervical cancer progression. For all proposed systems, the WSI images were optimized to show the contrast of edges of the low-contrast cells. Then, the cells to be analyzed were segmented and isolated from the rest of the image using the Active Contour Algorithm (ACA). WSI images were diagnosed by a hybrid method between deep learning (ResNet50, VGG19 and GoogLeNet), Random Forest (RF), and Support Vector Machine (SVM) algorithms based on the ACA algorithm. Another hybrid method for diagnosing WSI images by RF and SVM algorithms is based on fused features of deep-learning (DL) models (ResNet50-VGG19, VGG19-GoogLeNet, and ResNet50-GoogLeNet). It is concluded from the systems' performance that the DL models' combined features help significantly improve the performance of the RF and SVM networks. The novelty of this research is the hybrid method that combines the features extracted from deep-learning models (ResNet50-VGG19, VGG19-GoogLeNet, and ResNet50-GoogLeNet) with RF and SVM algorithms for diagnosing WSI images. The results demonstrate that the combined features from deep-learning models significantly improve the performance of RF and SVM. The RF network with fused features of ResNet50-VGG19 achieved an AUC of 98.75%, a sensitivity of 97.4%, an accuracy of 99%, a precision of 99.6%, and a specificity of 99.2%.

Residual Network with Attention to Neural Cells Segmentation

Many neuroscience applications, including understanding the evolution of the brain, rely on neural cell instance segmentation, which seeks to integrate the identification and segmentation of neuronal cells in microscopic imagery. However, the task is complicated by cell adhesion, deformation, vague cell outlines, low-contrast cell protrusion structures, and background imperfections. On the other hand, existing segmentation approaches frequently produce inaccurate findings. As a result, an effective strategy for using the residual network with attention to segment cells is suggested in this paper. The segmentation mask of neural cells may be accurately predicted. This method is built on U-net, with EfficientNet serving as the encoder's backbone. The attention approach is employed in the detection and segmentation modules to guide the model's attention to the most valuable features. A massive collection of neural cell microscopic images tests the proposed method. According to the findings of the experiments, this technology can accurately detect and segment neuronal cell occurrences with an intersection over the union IoU of 95.47 and a Dice-Coeff of 98.34, which is superior to current state-of-the-art approaches.

MultiHeadGAN: A deep learning method for low contrast retinal pigment epithelium cell segmentation with fluorescent flatmount microscopy images

BackgroundRetinal pigment epithelium (RPE) aging is an important cause of vision loss. As RPE aging is accompanied by changes in cell morphological features, an accurate segmentation of RPE cells is a prerequisite to such morphology analyses. Due the overwhelmingly large cell number, manual annotations of RPE cell borders are time-consuming. Computer based methods do not work well on cells with weak or missing borders in the impaired RPE sheet regions. MethodTo address such a challenge, we develop a semi-supervised deep learning approach, namely MultiHeadGAN, to segment low contrast cells from impaired regions in RPE flatmount images. The developed deep learning model has a multi-head structure that allows model training with only a small scale of human annotated data. To strengthen model learning, we further train our model with RPE cells without ground truth cell borders by generative adversarial networks. Additionally, we develop a new shape loss to guide the network to produce closed cell borders in the segmentation results. ResultsIn this study, 155 annotated and 1,640 unlabeled image patches are included for model training. The testing dataset consists of 200 image patches presenting large impaired RPE regions. The average RPE segmentation performance of the developed model MultiHeadGAN is 85.4 (correct rate), 88.8 (weighted correct rate), 87.3 (precision), and 80.1 (recall), respectively. Compared with other state-of-the-art deep learning approaches, our method demonstrates its superior qualitative and quantitative performance. ConclusionsSuggested by our extensive experimental results, our developed deep learning method can accurately segment cells in RPE flatmount microscopy images and is promising to support large scale cell morphological analyses for RPE aging investigations.

Cell Localization and Counting Using Direction Field Map.

Automatic cell counting in pathology images is challenging due to blurred boundaries, low-contrast, and overlapping between cells. In this paper, we train a convolutional neural network (CNN) to predict a two-dimensional direction field map and then use it to localize cell individuals for counting. Specifically, we define a direction field on each pixel in the cell regions (obtained by dilating the original annotation in terms of cell centers) as a two-dimensional unit vector pointing from the pixel to its corresponding cell center. Direction field for adjacent pixels in different cells have opposite directions departing from each other, while those in the same cell region have directions pointing to the same center. Such unique property is used to partition overlapped cells for localization and counting. To deal with those blurred boundaries or low contrast cells, we set the direction field of the background pixels to be zeros in the ground-truth generation. Thus, adjacent pixels belonging to cells and background will have an obvious difference in the predicted direction field. To further deal with cells of varying density and overlapping issues, we adopt geometry adaptive (varying) radius for cells of different densities in the generation of ground-truth direction field map, which guides the CNN model to separate cells of different densities and overlapping cells. Extensive experimental results on three widely used datasets (i.e., VGG Cell, CRCHistoPhenotype2016, and MBM datasets) demonstrate the effectiveness of the proposed approach.

Attentive neural cell instance segmentation.

Neural cell instance segmentation, which aims at joint detection and segmentation of every neural cell in a microscopic image, is essential to many neuroscience applications. The challenge of this task involves cell adhesion, cell distortion, unclear cell contours, low-contrast cell protrusion structures, and background impurities. Consequently, current instance segmentation methods generally fall short of precision. In this paper, we propose an attentive instance segmentation method that accurately predicts the bounding box of each cell as well as its segmentation mask simultaneously. In particular, our method builds on a joint network that combines a single shot multi-box detector (SSD) and a U-net. Furthermore, we employ the attention mechanism in both detection and segmentation modules to focus the model on the useful features. The proposed method is validated on a dataset of neural cell microscopic images. Experimental results demonstrate that our approach can accurately detect and segment neural cell instances at a fast speed, comparing favorably with the state-of-the-art methods. Our code is released on GitHub. The link is https://github.com/yijingru/ANCIS-Pytorch.

Automated tracking of label-free cells with enhanced recognition of whole tracks

Migration and interactions of immune cells are routinely studied by time-lapse microscopy of in vitro migration and confrontation assays. To objectively quantify the dynamic behavior of cells, software tools for automated cell tracking can be applied. However, many existing tracking algorithms recognize only rather short fragments of a whole cell track and rely on cell staining to enhance cell segmentation. While our previously developed segmentation approach enables tracking of label-free cells, it still suffers from frequently recognizing only short track fragments. In this study, we identify sources of track fragmentation and provide solutions to obtain longer cell tracks. This is achieved by improving the detection of low-contrast cells and by optimizing the value of the gap size parameter, which defines the number of missing cell positions between track fragments that is accepted for still connecting them into one track. We find that the enhanced track recognition increases the average length of cell tracks up to 2.2-fold. Recognizing cell tracks as a whole will enable studying and quantifying more complex patterns of cell behavior, e.g. switches in migration mode or dependence of the phagocytosis efficiency on the number and type of preceding interactions. Such quantitative analyses will improve our understanding of how immune cells interact and function in health and disease.

Curvelet initialized level set cell segmentation for touching cells in low contrast images

Cell segmentation is an important element of automatic cell analysis. This paper proposes a method to extract the cell nuclei and the cell boundaries of touching cells in low contrast images. First, the contrast of the low contrast cell images is improved by a combination of multiscale top hat filter and h-maxima. Then, a curvelet initialized level set method has been proposed to detect the cell nuclei and the boundaries. The image enhancement results have been verified using PSNR (Peak Signal to noise ratio) and the segmentation results have been verified using accuracy, sensitivity and precision metrics. The results show improved values of the performance metrics with the proposed method.

Localized, Macromolecular Transport for Thin, Adherent, Single Cells Via an Automated, Single Cell Electroporation Biomanipulator

Single cell electroporation (SCE), via microcapillary, is an effective method for molecular, transmembrane transport used to gain insight on cell processes with minimal preparation. Although possessing great potential, SCE is difficult to execute and the technology spans broad fields within cell biology and engineering. The technical complexities, the focus and expertise demanded during manual operation, and the lack of an automated SCE platform limit the widespread use of this technique, thus the potential of SCE has not been realized. In this study, an automated biomanipulator for SCE is presented. Our system is capable of delivering molecules into the cytoplasm of extremely thin cellular features of adherent cells. The intent of the system is to abstract the technical challenges and exploit the accuracy and repeatability of automated instrumentation, leaving only the focus of the experimental design to the operator. Each sequence of SCE including cell and SCE site localization, tip-membrane contact detection, and SCE has been automated. Positions of low-contrast cells are localized and "SCE sites" for microcapillary tip placement are determined using machine vision. In addition, new milestones within automated cell manipulation have been achieved. The system described herein has the capability of automated SCE of "thin" cell features less than 10 μm in thickness. Finally, SCE events are anticipated using visual feedback, while monitoring fluorescing dye entering the cytoplasm of a cell. The execution is demonstrated by inserting a combination of a fluorescing dye and a reporter gene into NIH/3T3 fibroblast cells.

Translational velocity measurement for single floating cell based on optical Fourier transform theory

This letter reports on translational velocity measurement, which is needed for tracking a low contrast cell. We propose a new optical spatial filtering method that is based on the optical Fourier transform theory. In this method, a pinhole is installed as a spatial filter on the optical Fourier transform plane. By means of this spatial filter, the arbitrary component of the spatial frequency is derived from the random refractive index distribution as the periodic light intensity distribution. By observing the changes of this light intensity, we can obtain the translational velocity of a low-contrast cell by means of a high-response photodiode.