An Effective Deep Learning Framework for Cell Segmentation in Microscopy Images.

The segmentation of cells is necessary for biologists in the morphological statistics for quantitative and qualitative analysis in Phase-contrast Microscopy (PCM) images. In this paper, we address the cell segmentation problem in PCM images. Deep Neural Networks (DNNs) commonly is initialized with weights from a network pre-trained on a large annotated data set like ImageNet have superior performance than those trained from scratch on a small dataset. Here, we demonstrate how encoder-decoder type architectures such as U-Net and Feature Pyramid Network (FPN) can be improved by an alternative encoder which pre-trained on the ImageNet dataset. In particular, our experimental results confirm that the image descriptors from ResNet-18 are highly effective in accurate prediction of the cell boundary and have higher Intersection over Union (IoU) in comparison to the classical U-Net and require fewer training epochs.

We propose a novel cell segmentation approach by extracting Multi-exposure Maximally Stable Extremal Regions (MMSER) in phase contrast microscopy images on the same cell dish. Using our method, cell regions can be well identified by considering the maximally stable regions with response to different camera exposure times. Meanwhile, halo artifacts with regard to cells at different stages are leveraged to identify cells' stages. The experimental results validate that high quality cell segmentation and cell stage classification can be achieved by our approach.

A computer vision approach is presented for border detection of malaria infected cells in microscopic blood images for accurate diagnosis. First, the microscopic 24-bits RGB color blood image converted in to 8-bits gray scale image for a single channel procesing. The poroposed two-stage thresholdingmethod used for segmentation of malaria infected cells. Regarding border irregularities, the chosen descriptor is the perimeter factor and 4-connected neighbourhood. The experimental results on benchmark dataset that comprises around 300 images show that the proposed method successfully detects borders of malaria infected cells with no prior knowledge of the contents of the image without parameter tuning. The proposed one compared with other existing methods and results are discussed.

In this paper, a novel algorithm for microscopic color cell image edge detection based on Pulse Coupled Neural Networks (PCNN) is presented. Microscopic color image edge detection has proven to be a difficult task due to uneven brightness, cross-color existing between cell boundary and background caused by dyeing and noise. To solve these problems, this paper originally employs an improved PCNN model called multi-dimensional PCNN to implement edge detection of microscopic color cell image. PCNN is an advanced approach which aims at processing color images parellelly rather than separately dispose signal to every channel because it is characterized by synchronous neuronal burst and multi-dimensional convolution. To test the feasibility and effectiveness of multi-dimensional PCNN on edge detection of microscopic color cell images, experiments of several microscopic cell images are carried out. Empirical results show that multi-dimensional PCNN outperforms classical methods in terms of anti-noise and the accuracy of weak edge detection.

Computational analysis of fluorescent timelapse microscopy images at the single-cell level is a powerful approach to study cellular changes that dictate important cell fate decisions. Core to this approach is the need to generate reliable cell segmentations and classifications necessary for accurate quantitative analysis. Deep learning-based convolutional neural networks (CNNs) have emerged as a promising solution to these challenges. However, current CNNs are prone to produce noisy cell segmentations and classifications, which is a significant barrier to constructing accurate single-cell lineages. To address this, we developed a novel algorithm called Single Cell Track (SC-Track), which employs a hierarchical probabilistic cache cascade model based on biological observations of cell division and movement dynamics. Our results show that SC-Track performs better than a panel of publicly available cell trackers on a diverse set of cell segmentation types. This cell-tracking performance was achieved without any parameter adjustments, making SC-Track an excellent generalized algorithm that can maintain robust cell-tracking performance in varying cell segmentation qualities, cell morphological appearances and imaging conditions. Furthermore, SC-Track is equipped with a cell class correction function to improve the accuracy of cell classifications in multiclass cell segmentation time series. These features together make SC-Track a robust cell-tracking algorithm that works well with noisy cell instance segmentation and classification predictions from CNNs to generate accurate single-cell lineages and classifications.

The article considers the labor intensity of the process of determining the white blood cell count of fish, which is simultaneously being highly significant and necessary in terms of monitoring the health of individuals. The authors present an approach to automating the compilation of the white blood cell count of fish (using sturgeon as an example) using a convolutional neural network model capable of recognizing and identifying cells in a microscopic blood image. The general scheme of hematopoiesis and standards for hematological parameters of sturgeon are considered. The procedure for preparing images for training a markup-based artificial neural network model is described. Software tools for interaction with images and artificial neural network models are described. As a result of the research, 14 microscopic images of fish blood cells based on markup were prepared, a convolutional neural network model was trained, the overall mean average precision (MAP) of which was 0.33. At the same time, the overall accuracy of cell recognition in individual images was 0.92, and the rate of red blood cell recognition was 0.94. The research results can serve as a basis for further study, development and application of convolutional neural networks for automating white blood cell count compilation based on high-precision recognition of fish blood cells in microscopic images.

In this study, we constructed and validated deep learning models capable of predicting the osteogenic differentiation stages of mesenchymal stem cells (MSCs) using only phase-contrast microscopy images. UE7T-13, an immortalized human MSC line, was cultured in osteoinductive medium. Phase-contrast microscopy images were acquired at D0, D1, D5, D10, and D14 of differentiation. Two deep learning models, ResNet-50 and DenseNet-121, were trained to perform multi-class classification of osteogenic differentiation stages. Model performance was evaluated using precision, sensitivity, F1 score, and overall accuracy. The overall accuracy of the ResNet-50 model was 0.700 and that of the DenseNet-121 model was 0.684. The highest F1 scores occurred at D5, which may reflect more distinctive morphological features during mid-stage differentiation. Our results suggest that deep learning has the potential to non-invasively identify osteogenic differentiation stages based on morphological features alone.

We investigate the relational classification of biological cells in 2D microscopy images. Rather than treating each cell image independently, we investigate whether and how the neighborhood information of a cell can be informative for its prediction. We propose a Relational Long Short-Term Memory (R-LSTM) algorithm, coupled with auto-encoders and convolutional neural networks, that can learn from both annotated and unlabeled microscopy images and that can utilize both the local and neighborhood information to perform an improved classification of biological cells. Experimental results on both synthetic and real datasets show that R-LSTM performs comparable to or better than six baselines.

Monitoring and tracking of cell motion is a key component for understanding disease mechanisms and evaluating the effects of treatments. Time-lapse optical microscopy has been commonly employed for studying cell cycle phases. However, usual manual cell tracking is very time consuming and has poor reproducibility. Automated cell tracking techniques are challenged by variability of cell region intensity distributions and resolution limitations. In this work, we introduce a comprehensive cell segmentation and tracking methodology. A key contribution of this work is that it employs multi-scale space-time interest point detection and characterization for automatic scale selection and cell segmentation. Another contribution is the use of a neural network with class prototype balancing for detection of cell regions. This work also offers a structured mathematical framework that uses graphs for track generation and cell event detection. We evaluated cell segmentation, detection, and tracking performance of our method on time-lapse sequences of the Cell Tracking Challenge (CTC). We also compared our technique to top performing techniques from CTC. Performance evaluation results indicate that the proposed methodology is competitive with these techniques, and that it generalizes very well to diverse cell types and sizes, and multiple imaging techniques.

An essential step for quantitative image analysis is cell segmentation, which is the process of defining the outline of individual cells in microscopy images. Segmentation of budding yeast is challenging due to their asymmetric cell division and mother-bud morphology. As a result, a dividing cell is frequently misidentified as two separate cells, causing errors in downstream analysis. Here, we overcame this challenge by developing YeastSAM, a deep learning-based segmentation framework derived from μSAM and optimized for budding yeast. YeastSAM achieves more than threefold higher accuracy in segmenting dividing cells compared to existing methods. When combined with single-molecule RNA imaging and organelle imaging, YeastSAM facilitates quantitative analysis of the spatial regulation of gene expression. This study offers an accessible, high-accuracy model for yeast cell segmentation, empowering researchers with minimal programming experience to perform quantitative image analysis.

Accurate and automated detection and tracking of cells in microscopy images is a persistent challenge in biotechnology and biomedical research. Effective detection and tracking are crucial for understanding biological processes and extracting meaningful data for subsequent simulations. In this study, we present an integrated pipeline that leverages a fine-tuned YOLOv8x model for detecting cells and cell divisions across microscopy image series. While YOLOv8x exhibits strong detection capabilities, it occasionally misses certain cells, leading to gaps in data. To mitigate this, we incorporate the DeepSORT tracking algorithm, which enhances data association and reduces the cells’ identity (ID) switches by utilizing a pre-trained convolutional network for robust multi-object tracking. This combination ensures continuous detection and compensates for missed detections, thereby improving overall recall. Our approach achieves a recall of 93.21% with the enhanced DeepSORT algorithm, compared to the 53.47% recall obtained by the original YOLOv8x model. The proposed pipeline effectively extracts detailed information from structured image datasets, providing a reliable approximation of cellular processes in culture environments.

Time-lapse imaging in combination with fluorescence microscopy techniques enable the investigation of gene regulatory circuits and uncovered phenomena like culture heterogeneity. In this context, computational image processing for the analysis of single cell behaviour plays an increasing role in systems biology and mathematical modelling approaches. Consequently, we developed a software package with graphical user interface for the analysis of single bacterial cell behaviour. A new software called TLM-Tracker allows for the flexible and user-friendly interpretation for the segmentation, tracking and lineage analysis of microbial cells in time-lapse movies. The software package, including manual, tutorial video and examples, is available as Matlab code or executable binaries at http://www.tlmtracker.tu-bs.de.

Support Vector Machine (SVM) integrated with subsampling methods provides the latest approach to cell segmentation of microscopic images. This approach aims to increase computing efficiency and also maintain segmentation accuracy. To begin with, the processing load associated with large-scale microscopic images is reduced by employing subsampling techniques. On applying Support Vector Machine (SVM), classification of the picture patches into cell and non-cell portions based on the gathered attributes. The produced method is highly valuable for high-throughput analysis in medical and biological diagnostics as it provides competitive segmentation performance while drastically cutting down time for its process. Through a detailed analysis from rigorous testing and comparison with existing methods, efficiency and efficacy are demonstrated in accurately differentiating cells from microscopic images and extending solutions for vast biomedical applications.

Phase contrast microscopy, as a noninvasive imaging technique, has been widely used to monitor the behavior of transparent cells without staining or altering them. Due to the optical principle of the specifically-designed microscope, phase contrast microscopy images contain artifacts such as halo and shade-off which hinder the cell segmentation and detection tasks. Some previous works developed simplified computational imaging models for phase contrast microscopes by linear approximations and convolutions. The approximated models do not exactly reflect the imaging principle of the phase contrast microscope and accordingly the image restoration by solving the corresponding deconvolution process is not perfect. In this paper, we revisit the optical principle of the phase contrast microscope to precisely formulate its imaging model without any approximation. Based on this model, we propose an image restoration procedure by reversing this imaging model with a deep neural network, instead of mathematically deriving the inverse operator of the model which is technically impossible. Extensive experiments are conducted to demonstrate the superiority of the newly derived phase contrast microscopy imaging model and the power of the deep neural network on modeling the inverse imaging procedure. Moreover, the restored images enable that high quality cell segmentation task can be easily achieved by simply thresholding methods. Implementations of this work are publicly available at https://github.com/LiangHann/Phase-Contrast-Microscopy-Image-Restoration.

A probabilistic approach to joint cell tracking and segmentation in high-throughput microscopy videos.