Figure - available from: Scientific Data
This content is subject to copyright. Terms and conditions apply.
Steps involved in generating the ground truth segmentation labels for the inner core tissues. The original images pre-processed by employing rolling ball algorithm and bandpass filtering. Next, the adaptive thresholding has been employed to obtain binary images. Furthermore, morphological operations have been performed to refine the cell boundaries and remove the starch granules. By changing fcl, fch and R at pre-processing steps, possible binary images have been generated. Then, the best image has been selected for manual correction.
Source publication
We present a new large-scale three-fold annotated microscopy image dataset, aiming to advance the plant cell biology research by exploring different cell microstructures including cell size and shape, cell wall thickness, intercellular space, etc. in deep learning (DL) framework. This dataset includes 9,811 unstained and 6,127 stained (safranin-o,...
Similar publications
Background
Congenital syphilis (CS) is the infection of an infant or fetus with Treponema pallidum . The aim of this study was to investigate the clinical features and outcomes of serology reversion in infants diagnosed with confirmed or suspected congenital syphilis (CS).
Methods
Infants admitted to the neonatal department of Children's Hospital...
A loss-of-function mutation in the Lrp5 gene in mice leads to a low bone mass disorder due to the inhibition of the canonical Wnt signaling pathway; however, the role of bone marrow microenvironment in mice with this mutation remains unclear. In this study, we evaluated proliferation and osteogenic potential of mouse osteoblasts using the MTT assay...
Background: The prevalence of osteoarthritis (OA) is increasing, yet clinically effective and economical treatments are unavailable. We have previously proposed a cell-free fat extract (CEFFE) containing multiple cytokines, which possessed anti-apoptotic, anti-oxidative, and proliferation promotion functions, as a “cell-free” strategy. In this stud...
Background
The prevalence of osteoarthritis (OA) is increasing, yet clinically effective and economical treatments are unavailable. We have previously proposed a cell-free fat extract (CEFFE) containing multiple cytokines, which possessed antiapoptotic, anti-oxidative, and proliferation promotion functions, as a “cell-free” strategy. In this study,...
Background
The diagnosis of neurosyphilis is challenging due to the requirement of a lumbar puncture and cerebrospinal fluid (CSF) laboratory tests. Therefore, a convenient diagnostic nomogram for neurosyphilis is warranted. This study aimed to construct diagnostic models for diagnosing neurosyphilis.
Methods
This cross-sectional study included da...
Citations
... In biological microscopy, deep learning has demonstrated promising performance in a range of segmentation applications including semantic segmentation of human oocyte (Targosz et al., 2021), semantic and instance segmentation for cell nuclei (Caicedo et al., 2019) and semantic segmentation potato tuber (Biswas and Barma, 2020). Examples of plant phenotyping applications include semantic and instance segmentation for plant leaf detection and counting (Aich and Stavness, 2017;Giuffrida et al., 2018;Itzhaky et al., 2018;Jiang et al., 2019;Fan et al., 2022), semantic and instance segmentation for crop phenotyping (Jiang and Li, 2020), grapevine leaf semantic segmentation (Tamvakis et al., 2022), barley seed detection from instance segmentation (Toda et al., 2020) and many other applications (Kolhar and Jagtap, 2023). ...
In commercial forestry and large-scale plant propagation, the utilization of artificial intelligence techniques for automated somatic embryo analysis has emerged as a highly valuable tool. Notably, image segmentation plays a key role in the automated assessment of mature somatic embryos. However, to date, the application of Convolutional Neural Networks (CNNs) for segmentation of mature somatic embryos remains unexplored. In this study, we present a novel application of CNNs for delineating mature somatic conifer embryos from background and residual proliferating embryogenic tissue and differentiating various morphological regions within the embryos. A semantic segmentation CNN was trained to assign pixels to cotyledon, hypocotyl, and background regions, while an instance segmentation network was trained to detect individual cotyledons for automated counting. The main dataset comprised 275 high-resolution microscopic images of mature Pinus radiata somatic embryos, with 42 images reserved for testing and validation sets. The evaluation of different segmentation methods revealed that semantic segmentation achieved the highest performance averaged across classes, achieving F1 scores of 0.929 and 0.932, with IoU scores of 0.867 and 0.872 for the cotyledon and hypocotyl regions respectively. The instance segmentation approach demonstrated proficiency in accurate detection and counting of the number of cotyledons, as indicated by a mean squared error (MSE) of 0.79 and mean absolute error (MAE) of 0.60. The findings highlight the efficacy of neural network-based methods in accurately segmenting somatic embryos and delineating individual morphological parts, providing additional information compared to previous segmentation techniques. This opens avenues for further analysis, including quantification of morphological characteristics in each region, enabling the identification of features of desirable embryos in large-scale production systems. These advancements contribute to the improvement of automated somatic embryogenesis systems, facilitating efficient and reliable plant propagation for commercial forestry applications.
... Finally, we argue that the microscopic images are one potential to recognize plant disease at the cell level via combining compute vision and plant cell [61]. Besides, one microscopic image dataset is proposed to analyze potato tubes [62]. ...
The background of deep learning and plant disease recognition are discussed in this thesis. Different methods and sensors are compared to perform plant disease recognition. Based on proximal RGB images, we conduct our study using deep learning.
To address the limited training dataset challenge, we propose a style-consistent image translation algorithm and a transfer learning strategy through this viewpoint. Inspired by one observation that the images of healthy leaf are much easier collected with high intra-class variation than some diseased one, the healthy leaves are translated into disease, yet image styles are retained using a style loss function, by which the high intra-class variation can be taken from healthy leaves to disease one. On the other hand, we noticed the difference between the dataset in general computer vision and plant disease. A common reasonable assumption is that similar source and target datasets are beneficial with the transfer learning paradigm. Therefore, a plant-related dataset, PlantCLEF2022, with 2,885,052 images and 80,000 classes, is utilized to pre-train a model. To reduce the time to train in PlantCLEF2022, we propose a two-step transfer learning where one public model pre-trained in ImageNet is pre-trained once more in PlantCLEF2022 before fine-tuning in a target dataset.
... It indicated that the cortex, vascular ring and medulla at the outside stem end part (No. 1-4, 6-9) generally had the highest DM levels, while the pith in the inside sections (No. 5, 14, 23) had the lowest DMC. Previous studies suggested that the potato tuber weight may be correlated with the cell volume of different tissue zones [46]. Microscopic images of different sections of number 1 potato were acquired during the experiment. ...
Potatoes are generally consumed directly as a staple food or used for processing, depending on the quality of raw materials. Dry matter content (DMC) is the most critical characteristic of potatoes, as it determines the processing and the final product quality. This study aimed to investigate the potential of different optical sensing systems in predicting the DMC of intact potato tubers, and the efficacy of classifying potatoes based on dry matter levels. The whole tubers were scanned using three optical sensing modes (transmittance spectra, interactance spectra and hyperspectral imaging). PLSR and different classifiers (PLSDA, SVM and ANN) were utilized to build the prediction and classification models, respectively. To extract the most influential wavelengths related to the prediction of DMC, the CARS and CSMW techniques were applied. The results indicated that the DMC of two sections on the equator of the potato tuber belly can well represent the DMC of the intact potato, and together with the spectral detection at the equatorial position, it provided good performance. The CARS-PLSR prediction model in transmittance mode showed stronger correlations than other systems, with Rp and RMSEP values of 0.968 and 0.413%, respectively. The CARS-SVM-Linear classification model exhibited the best performance with classification rates of 100% and 97.62% in the training and testing sets, respectively. Moreover, the spectra preferred by CARS and CSMW variable selection methods in transmittance mode overlapped near the absorption peak at 980 nm, indicating the importance of this band for predicting DMC. This study presented the feasible application of using spectroscopy to evaluate the DMC of intact potatoes and classify potatoes based on thresholds that are crucial to consumers and food processors.
Graphical abstract
... There is a need for automated, scalable software tools that could automatically analyze large datasets efficiently. An accurate but automated way to measure anatomical traits could also allow us to take advantage of increasingly available large image datasets [14]. In short, a faster means of measuring key characteristics of xylem anatomy would greatly broaden the scope of questions that can be asked about plant structure and function. ...
... Previous studies have applied machine learning algorithms to topics such as plant disease classification [19] and plant leaf classification [20]; however few studies have been conducted on machine learning at the cellular level. Those studies at this microscopic level have mainly been focused on the segmentation of plant tissues, then possibly classifying those segmented images [14,[21][22][23][24]. To our best knowledge, we are the first group to design a cascade-like machine learning model to classify tracheary elements (vessels), fibers, and parenchyma cells in plants, with discussions on the effects of inputting a contextualized image alongside the image to be classified. ...
Anatomical studies of plant hydraulic traits have traditionally been conducted by manual measurements of light micrographs. An automated process could expedite analysis and broaden the scope of questions that can be asked, but such an approach would require the ability to accurately classify plant cells according to their type. Our research evaluates a deep learning-based model which accepts a cropped cell image input alongside a broader cropped image which incorporates contextual information of that cell type’s original cropped image, and learns to segregate these plant cells based off of the features of both inputs. Whilst a single cropped image classification yielded adequate results with outputs matching the ground-truth labels, we discovered that a second image input significantly bolstered the model’s learning and accuracy (98.1%), indicating that local context provides important information needed to accurately classify cells. Finally our results imply a future application of our classifier to automatic cell-type detection in xylem tissue image cross sections.
... Various data acquisition systems are being used in Micro/Macroscopic DL, such as visible light sensors [27], infrared (IR) and near-infrared (NIR) [28], ultraviolet [29], hyperspectral [30], and even X-rays [31]. Furthermore, other supportive techniques are being used to increase the depth of the acquired imagery such as cell staining in microscopy [32] and fluorescence imaging systems [33]. There are also some attempts to employ plant's electrical signals measurements in DL research [34]. ...
The role that deep learning plays in modern life is undeniably essential. It is also certain that deep learning, with its various approaches, is contributing significantly to plant science. Whether by explaining the acquired data or converting and refining these data to a more profound level, deep learning techniques are pushing the frontiers of plant research further than ever before. This study is an attempt to shed light on recent advances and applications of deep learning in plant science. These applications were systematically reviewed at omics, micro/macroscopic, and population levels. Future aspects were also discussed to some extent.
... Research results show that FSL can overcome some of the shortcomings of ML, for example, by significantly reducing the amount of data needed to train ML models and thereby reducing the time required to collect and label large datasets. Areas of application include image classification [20], image retrieval [21], and gesture recognition [22]. Fink [23] proposed a framework that could learn object classifiers from a single example, meaning that the data collection work involved in image classification tasks could be greatly reduced. ...
Few-shot learning (FSL) is a core topic in the domain of machine learning (ML), in which the focus is on the use of small datasets to train the model. In recent years, there have been many important data-driven ML applications for intrusion detection. Despite these great achievements, however, gathering a large amount of reliable data remains expensive and time-consuming, or even impossible. In this regard, FSL has been shown to have advantages in terms of processing small, abnormal data samples in the huge application space of intrusion detection. FSL can improve ML for scarce data at three levels: the data, the model, and the algorithm levels. Previous knowledge plays an important role in all three approaches. Many promising methods such as data enrichment, the graph neural network model, and multitask learning have also been developed. In this paper, we present a comprehensive review of the latest research progress in the area of FSL. We first introduce the theoretical background to ML and FSL and then describe the general features, advantages, and main methods of FSL. FSL methods such as embedded learning, multitask learning, and generative models are applied to intrusion detection to improve the detection accuracy effectively. Then, the application of FSL to intrusion detection is reviewed in detail, including enriching the dataset by extracting intermediate features, using graph embedding and meta-learning methods to improve the model. Finally, the difficulties of this approach and its prospects for development in the field of intrusion detection are identified based on the previous discussion.
... A dataset of potato tuber was established using a microscope. Bismas et al. 35 created a dataset consisting of 9811 unstained and 6127 stained potato samples. Through deep learning assessment, by utilising a CNN, and using a train-test split of 80-20, an accuracy of 94.27% was achieved for the unstained dataset. ...
... For example, this approach allows studying the ultrastructural features of the cell wall (Yarbrough et al., 2009), which enables us to assess cellular biomechanics indirectly. The combination of large-scale annotated image datasets and deep learning approaches is a promising technique for annotating physical, morphological, and tissue grading cellular properties (Fricker, 2016;Biswas and Barma, 2020). ...
Single-cell technology is a relatively new and promising way to obtain high-resolution transcriptomic data mostly used for animals during the last decade. However, several scientific groups developed and applied the protocols for some plant tissues. Together with deeply-developed cell-resolution imaging techniques, this achievement opens up new horizons for studying the complex mechanisms of plant tissue architecture formation. While the opportunities for integrating data from transcriptomic to morphogenetic levels in a unified system still present several difficulties, plant tissues have some additional peculiarities. One of the plants’ features is that cell-to-cell communication topology through plasmodesmata forms during tissue growth and morphogenesis and results in mutual regulation of expression between neighboring cells affecting internal processes and cell domain development. Undoubtedly, we must take this fact into account when analyzing single-cell transcriptomic data. Cell-based computational modeling approaches successfully used in plant morphogenesis studies promise to be an efficient way to summarize such novel multiscale data. The inverse problem’s solutions for these models computed on the real tissue templates can shed light on the restoration of individual cells’ spatial localization in the initial plant organ—one of the most ambiguous and challenging stages in single-cell transcriptomic data analysis. This review summarizes new opportunities for advanced plant morphogenesis models, which become possible thanks to single-cell transcriptome data. Besides, we show the prospects of microscopy and cell-resolution imaging techniques to solve several spatial problems in single-cell transcriptomic data analysis and enhance the hybrid modeling framework opportunities.
Given the significant impact of biofilms on human health and material corrosion, research in this field urgently needs more accessible techniques to facilitate the testing of new control agents and general understanding of biofilm biology. Microtiter plates offer a convenient format for standardized evaluations, including high-throughput assays of alternative treatments and molecular modulators. This study introduces a novel Biofilm Analysis Software (BAS) for quantifying biofilms from microtiter plate images. We focused on early biofilm growth stages and compared BAS quantification to common techniques: direct turbidity measurement, intrinsic fluorescence detection linked to pyoverdine production, and standard crystal violet staining which enables image analysis and optical density measurement. We also assessed their sensitivity for detecting subtle growth effects caused by cyclic AMP and gentamicin. Our results show that BAS image analysis is at least as sensitive as the standard method of spectrophotometrically quantifying the crystal violet retained by biofilms. Furthermore, we demonstrated that bacteria adhered after short incubations (from 10 min to 4 h), isolated from planktonic populations by a simple rinse, can be monitored until their growth is detectable by intrinsic fluorescence, BAS analysis, or resolubilized crystal violet. These procedures are widely accessible for many laboratories, including those with limited resources, as they do not require a spectrophotometer or other specialized equipment.
