Deep learning for quantifying spatial patterning and formation process of early differentiated human-induced pluripotent stem cells with micropattern images.

Micropatterning is reliable method for quantifying pluripotency of human-induced pluripotent stem cells (hiPSCs) that differentiate to form a spatial pattern of sorted, ordered and nonoverlapped three germ layers on the micropattern. In this study, we propose a deep learning method to quantify spatial patterning of the germ layers in the early differentiation stage of hiPSCs using micropattern images. We propose decoding and encoding U-net structures learning labelled Hoechst (DNA-stained) hiPSC regions with corresponding Hoechst and bright-field micropattern images to segment hiPSCs on Hoechst or bright-field images. We also propose a U-net structure to extract extraembryonic regions on a micropattern, and an algorithm to compares intensities of the fluorescence images staining respective germ-layer cells and extract their regions. The proposed method thus can quantify the pluripotency of a hiPSC line with spatial patterning including cell numbers, areas and distributions of germ-layer and extraembryonic cells on a micropattern, and reveal the formation process of hiPSCs and germ layers in the early differentiation stage by segmenting live-cell bright-field images. In our assay, the cell-number accuracy achieved 86% and 85%, and the cell region accuracy 89% and 81% for segmenting Hoechst and bright-field micropattern images, respectively. Applications to micropattern images of multiple hiPSC lines, micropattern sizes, groups of markers, living and fixed cells show the proposed method can be expected to be a useful protocol and tool to quantify pluripotency of a new hiPSC line before providing it to the scientific community.

Quantitative Analyses for Early Tempo-spatial Patterning of Differentiated Human Induced Pluripotent Stem Cells on Micropatterns using Time-lapse Bright-field Microscopy Images.

Three germ layer formation on micropatterns are extremely useful for quantitative analysis of hiPSC (human induced pluripotent stem cells) pluripotency. Spatial patterns of stem cells differentiated on the micropatterns will be formed from about 24 hours after differentiation induction and usually quantitated near 48 hours. To delineate the germ layer formation process, temporal changes in spatial patterning of germ layers should be analyzed by noninvasive microscopy. This study proposed a series of image processing methods combined with a U-net automatic segmentation to segment differentiated hiPSCs captured by bright-field microscopy. High segmentation accuracy (83.3%) for the test bright-field images compared with their concurrent Hoechst images (85%) was achieved. Tempo-spatial patterning and formation process of germ layers on the micropatterns can be visualized and quantified by segmenting time-lapse bright-field microscopy images using our method.

Quantitative Analysis of Differentiation Activity for Mouse Embryonic Stem Cells by Deep Learning for Cell Center Detection using Three-Dimensional Confocal Fluorescence Microscopy Images.

Accurate single cell segmentation provides means to monitor the behavior of single cell within a population of cells. Time-lapse fluorescence images are used to reveal heterogeneous nature of single mouse embryonic stem cell (ESC) colony and monitor fluctuations of the cell states. Automatic quantification of speed and status shifts of the ESCs depends on accurate single cell segmentation that is used to calculate the 3D center of every cell and track this cell for the quantification. This study proposes a new 3D U-net to accurately detect center of each single cell in 3D confocal images. The dimension of input 3D images to the U-net is flexible so that multiple center detections from different image directions can be implemented simultaneously to improve the center detection accuracy. This study showed that our method can improve accuracy for cell center detection and thus the quantification for ESC speeds and status shifts.

Human induced pluripotent stem cell formation and morphology prediction during reprogramming with time-lapse bright-field microscopy images using deep learning methods.

BACKGROUND AND OBJECTIVE: Human induced pluripotent stem cells (hiPSCs) represent an ideal source for patient specific cell-based regenerative medicine; however, efficiency of hiPSC formation from reprogramming cells is low. We use several deep-learning results from time-lapse brightfield microscopy images during culture, to early detect the cells potentially reprogramming into hiPSCs and predict the colony morphology of these cells for improving efficiency of culturing a new hiPSC line. METHODS: Sets of time-lapse bright-field images are taken to track reprogramming process of CD34+ cells biologically identified as just beginning reprogramming. Prior the experiment, 9 classes of templates with distinct cell features clipped from microscopy images at various reprogramming stages are used to train a CNN model. The CNN is then used to classify a microscopy image as probability images of these classes. Probability images of some class are used to train a densely connected convolutional network for extracting regions of this class on a microscopy image. A U-net is trained to segment cells on the time-lapse images in early reprogramming stage during culture. The segmented cells are classified by the extracted regions to count various types of cells appearing in the early reprogramming stage for predicting the identified cells potentially forming hiPSCs. The probability images of hiPSC classes are also used to train a spatiotemporal RNN for predicting the future hiPSC colony morphology of the potential cells. RESULTS: Experimental results show the prediction (before 7 days after of beginning of the reprogramming) achieved 0.8 accuracy, and 66% of the identified cells under different culture conditions, predicted as forming, finally formed hiPSCs. The predicted hiPSC images and extracted colonies on the images show the prediction for future 1.5 days achieved high accuracy of hiPSC colony areas and image similarity. CONCLUSIONS: Our study proposes a method using several deep learning models to efficiently select the reprogramming cells possibly forming hiPSCs and to predict the shapes of growing hiPSC colonies. The proposed method is expected to improve the efficiency when establishing a new hiPSC line culture.

Prediction of Human Induced Pluripotent Stem Cell Formation Based on Deep Learning Analyses Using Time-lapse Brightfield Microscopy Images.

We use deep learning methods to predict human induced pluripotent stem cell (hiPSC) formation using time-lapse brightfield microscopy images taken from a cell identified as the beginning of entered into the reprogramming process. A U-net is used to segment cells and a CNN is used to classify the segmented cells into eight types of cells during the reprogramming and hiPSC formation based on cellular morphology on the microscopy images. The numbers of respective types of cells in cell clusters before the hiPSC formation stage are used to predict if hiPSC regions can be well formed lately. Experimental results show good prediction by the criteria using the numbers of different cells in the clusters. Time-series images with respective types of classified cells can be used to visualize and quantitatively analyze the growth and transition among dispersed cells not in cell clusters, various types of cells in the clusters before the hiPSC formation stage and hiPSC cells.

U-net structures for segmentation of single mouse embryonic stem cells using three-dimensional confocal microscopy images.

Cell segmentation at a single cell resolution is required to provide insights for basic biology and application study. However, there are issues of low signal-to-noise ratio, weak fluorescence response, and insufficient resolution along the image stacking direction in 3D confocal images (volume). It has been difficult to segment out single cells from close or contacted cells in a cell volume using image processing methods or together with geometric processing methods. Recently, 3D deep learning methods have been used to avoid tedious parameter settings in the image and geometric processing, but still not easy to segment out close or contacted single cells. This paper proposes a 2D U-net to segment cell regions in high accuracy and computing performance. Better 3D cell images and single cell segmentation for close or contacted cells are achieved by combining a 3D U-net to detect the centers of single cells in the volume.

High Resolution U-Net for Quantitatively Analyzing Early Spatial Patterning of Human Induced Pluripotent Stem Cells on Micropatterns.

Human induced pluripotent stem cells (hiPSCs) can differentiate into three germ layer cells, i.e. ectoderm, mesoderm and endoderm, on micropatterned chips in highly synchronous and reproducible manners. The cells are confined within the chip, expanding two-dimensionally as almost in the form of monolayer, thus to be ideal for serving quantitative analysis of their pluripotency. We present a new U-Net (MP-UNet) structure for cell segmentation of early spatial patterning of hiPSCs on micropattern chips using Hoechst fluorescence images. In this structure, the encoding/decoding layers can be dynamically adjusted to extract sufficient image features and be flexible to image sizes. Dice and weight loss functions are designed to identify slight difference in low signal-to-noise ratio, high boundary-to-area ratio and compacted cell images. Several sizes of Hoechst images were tested to show MP-UNet can achieve high accuracy in cell regions and number counting for various sizes of micropattern chips, thus to be excellent quantitative tool for early spatial patterning of hiPSCs.

Prediction for Morphology and States of Stem Cell Colonies using a LSTM Network with Progressive Training Microscopy Images.

We present a new LSTM (P-LSTM: Progressive LSTM) network, aiming to predict morphology and states of cell colonies from time-lapse microscopy images. Apparent short-term changes occur in some types of time-lapse cell images. Therefore, long-term-memory dependent LSTM networks may not predict accurately. The P-LSTM network incorporates the images newly generated from cell imaging progressively into LSTM training to emphasize the LSTM short-term memory and thus improve the prediction accuracy. The new images are input into a buffer to be selected for batch training. For real-time processing, parallel computation is introduced to implement concurrent training and prediction on partitioned images.Two types of stem cell images were used to show effectiveness of the P-LSTM network. One is for tracking of ES cell colonies. The actual and predicted ES cell images possess similar colony areas and the same transitions of colony states (moving, merging or morphology changing), although the predicted colony mergers may delay in several time-steps. The other is for prediction of iPS cell reprogramming from the CD34+ human cord blood cells. The actual and predicted iPS cell images possess high similarity evaluated by the PSNR and SSIM similarity evaluation metrics, indicating the reprogramming iPS cell colony features and morphology can be accurately predicted.

Human Induced Pluripotent Stem Cell Reprogramming Prediction in Microscopy Images using LSTM based RNN.

We present a LSTM (Long Short-Term Memory) based RNN (recurrent neural network) method for predicting human induced Pluripotent Stem (hiPS) cells in the reprogramming process. The method uses a trained LSTM network by time-lapse microscopy images to predict growth and transition of reprogramming processes of CD34+ human cord blood cells into hiPS cells. The prediction can be visualized by output time-series probability images. The growth and transition are thus analyzed quantitatively by region areas of distinct cells emerged during the iPS formation processes. The experimental results show that our LSTM network is a potentially powerful tool to predict the cells at the distinct phases of the reprogramming to hiPS cells. This method should be extremely useful not only for basic biology of iPS cells but also detection of the reprogramming cells that will become genuine hiPS cells even at early stages of hiPS formation. Such predictive power should greatly reduce cost, labor and time required for establishment of the genuine hiPS cells, thereby accelerating the practical use of hiPS cells in regenerative medicine.