Detecting colon polyps in endoscopic images using artificial intelligence constructed with automated collection of annotated images from an endoscopy reporting system.

BACKGROUND: Artificial intelligence (AI) has made considerable progress in image recognition, especially in the analysis of endoscopic images. The availability of large-scale annotated datasets has contributed to the recent progress in this field. Datasets of high-quality annotated endoscopic images are widely available, particularly in Japan. A system for collecting annotated data reported daily could aid in accumulating a significant number of high-quality annotated datasets. AIM: We assessed the validity of using daily annotated endoscopic images in a constructed reporting system for a prototype AI model for polyp detection. METHODS: We constructed an automated collection system for daily annotated datasets from an endoscopy reporting system. The key images were selected and annotated for each case only during daily practice, not to be performed retrospectively. We automatically extracted annotated endoscopic images of diminutive colon polyps that had been diagnosed (study period March-September 2018) using the keywords of diagnostic information, and additionally collect the normal colon images. The collected dataset was devised into training and validation to build and evaluate the AI system. The detection model was developed using a deep learning algorithm, RetinaNet. RESULTS: The automated system collected endoscopic images (47,391) from colonoscopies (745), and extracted key colon polyp images (1356) with localized annotations. The sensitivity, specificity, and accuracy of our AI model were 97.0%, 97.7%, and 97.3% (n = 300), respectively. CONCLUSION: The automated system enabled the development of a high-performance colon polyp detector using images in endoscopy reporting system without the efforts of retrospective annotation works.

Accurate and fast mitotic detection using an anchor-free method based on full-scale connection with recurrent deep layer aggregation in 4D microscopy images.

BACKGROUND: To effectively detect and investigate various cell-related diseases, it is essential to understand cell behaviour. The ability to detection mitotic cells is a fundamental step in diagnosing cell-related diseases. Convolutional neural networks (CNNs) have been successfully applied to object detection tasks, however, when applied to mitotic cell detection, most existing methods generate high false-positive rates due to the complex characteristics that differentiate normal cells from mitotic cells. Cell size and orientation variations in each stage make detecting mitotic cells difficult in 2D approaches. Therefore, effective extraction of the spatial and temporal features from mitotic data is an important and challenging task. The computational time required for detection is another major concern for mitotic detection in 4D microscopic images. RESULTS: In this paper, we propose a backbone feature extraction network named full scale connected recurrent deep layer aggregation (RDLA++) for anchor-free mitotic detection. We utilize a 2.5D method that includes 3D spatial information extracted from several 2D images from neighbouring slices that form a multi-stream input. CONCLUSIONS: Our proposed technique addresses the scale variation problem and can efficiently extract spatial and temporal features from 4D microscopic images, resulting in improved detection accuracy and reduced computation time compared with those of other state-of-the-art methods.

ViBrism DB: an interactive search and viewer platform for 2D/3D anatomical images of gene expression and co-expression networks.

Understanding anatomical structures and biological functions based on gene expression is critical in a systemic approach to address the complexity of the mammalian brain, where >25 000 genes are expressed in a precise manner. Co-expressed genes are thought to regulate cell type- or region-specific brain functions. Thus, well-designed data acquisition and visualization systems for profiling combinatorial gene expression in relation to anatomical structures are crucial. To this purpose, using our techniques of microtomy-based gene expression measurements and WebGL-based visualization programs, we mapped spatial expression densities of genome-wide transcripts to the 3D coordinates of mouse brains at four post-natal stages, and built a database, ViBrism DB (http://vibrism.neuroinf.jp/). With the DB platform, users can access a total of 172 022 expression maps of transcripts, including coding, non-coding and lncRNAs in the whole context of 3D magnetic resonance (MR) images. Co-expression of transcripts is represented in the image space and in topological network graphs. In situ hybridization images and anatomical area maps are browsable in the same space of 3D expression maps using a new browser-based 2D/3D viewer, BAH viewer. Created images are shareable using URLs, including scene-setting parameters. The DB has multiple links and is expandable by community activity.

A covalently linked probe to monitor local membrane properties surrounding plasma membrane proteins.

Functional membrane proteins in the plasma membrane are suggested to have specific membrane environments that play important roles to maintain and regulate their function. However, the local membrane environments of membrane proteins remain largely unexplored due to the lack of available techniques. We have developed a method to probe the local membrane environment surrounding membrane proteins in the plasma membrane by covalently tethering a solvatochromic, environment-sensitive dye, Nile Red, to a GPI-anchored protein and the insulin receptor through a flexible linker. The fluidity of the membrane environment of the GPI-anchored protein depended upon the saturation of the acyl chains of the lipid anchor. The local environment of the insulin receptor was distinct from the average plasma membrane fluidity and was quite dynamic and heterogeneous. Upon addition of insulin, the local membrane environment surrounding the receptor specifically increased in fluidity in an insulin receptor-kinase dependent manner and on the distance between the dye and the receptor.

Computer-aided demarcation of early gastric cancer: a pilot comparative study with endoscopists.

BACKGROUND: Precise area diagnosis of early gastric cancer (EGC) is critical for reliable endoscopic resection. Computer-aided diagnosis (CAD) shows strong potential for detecting EGC and reducing cancer-care disparities caused by differences in endoscopists' skills. To be used in clinical practice, CAD should enable both the detection and the demarcation of lesions. This study proposes a scheme for the detection and delineation of EGC under white-light endoscopy and validates its performance using 1-year consecutive cases. METHODS: Only 300 endoscopic images randomly selected from 68 consecutive cases were used for training a convolutional neural network. All cases were treated with endoscopic submucosal dissection, enabling the accumulation of a training dataset in which the extent of lesions was precisely determined. For validation, 462 cancer images and 396 normal images from 137 consecutive cases were used. From the validation results, 38 randomly selected images were compared with those delineated by six endoscopists. RESULTS: Successful detections of EGC in 387 cancer images (83.8%) and the absence of lesions in 307 normal images (77.5%) were achieved. Positive and negative predictive values were 81.3% and 80.4%, respectively. Successful detection was achieved in 130 cases (94.9%). We achieved precise demarcation of EGC with a mean intersection over union of 66.5%, showing the extent of lesions with a smooth boundary; the results were comparable to those achieved by specialists. CONCLUSIONS: Our scheme, validated using 1-year consecutive cases, shows potential for demarcating EGC. Its performance matched that of specialists; it might therefore be suitable for clinical use in the future.

A Cascade of 2.5D CNN and Bidirectional CLSTM Network for Mitotic Cell Detection in 4D Microscopy Image.

Mitosis detection is one of the challenging steps in biomedical imaging research, which can be used to observe the cell behavior. Most of the already existing methods that are applied in detecting mitosis usually contain many nonmitotic events (normal cell and background) in the result (false positives, FPs). In order to address such a problem, in this study, we propose to apply 2.5-dimensional (2.5D) networks called CasDetNet_CLSTM, which can accurately detect mitotic events in 4D microscopic images. This CasDetNet_CLSTM involves a 2.5D faster region-based convolutional neural network (Faster R-CNN) as the first network, and a convolutional long short-term memory (CLSTM) network as the second network. The first network is used to select candidate cells using the information from nearby slices, whereas the second network uses temporal information to eliminate FPs and refine the result of the first network. Our experiment shows that the precision and recall of our networks yield better results than those of other state-of-the-art methods.

A statistical image analysis framework for pore-free islands derived from heterogeneity distribution of nuclear pore complexes.

Nuclear pore complexes (NPCs) maintain cellular homeostasis by mediating nucleocytoplasmic transport. Although cyclin-dependent kinases (CDKs) regulate NPC assembly in interphase, the location of NPC assembly on the nuclear envelope is not clear. CDKs also regulate the disappearance of pore-free islands, which are nuclear envelope subdomains; this subdomain gradually disappears with increase in homogeneity of the NPC in response to CDK activity. However, a causal relationship between pore-free islands and NPC assembly remains unclear. Here, we elucidated mechanisms underlying NPC assembly from a new perspective by focusing on pore-free islands. We proposed a novel framework for image-based analysis to automatically determine the detailed 'landscape' of pore-free islands from a large quantity of images, leading to the identification of NPC intermediates that appear in pore-free islands with increased frequency in response to CDK activity. Comparison of the spatial distribution between simulated and the observed NPC intermediates within pore-free islands showed that their distribution was spatially biased. These results suggested that the disappearance of pore-free islands is highly related to de novo NPC assembly and indicated the existence of specific regulatory mechanisms for the spatial arrangement of NPC assembly on nuclear envelopes.

Analysis of the equine ovarian structure during the first twelve months of life by three-dimensional internal structure microscopy.

A three-dimensional internal structure microscopy (3D-ISM) can clarify the anatomical arrangement of internal structures of equine ovaries. In this study, morphological changes of the equine ovary over the first 12 months of life were investigated by 3D-ISM in 59 fillies and by histological analysis in 2 fillies. The weight and volume of the paired ovaries initially decreased from 0 to 1 months to 2 to 3 months of age and then significantly increased at 8 to 12 months of age. The ovulation fossa was first observed around the 3rd month and became evident after the 6th month. The number of follicles with a diameter of ≥10 mm and the diameter of the largest follicle increased gradually after 6 months of age. On a volume basis, the medulla accounted for nearly 90% of the whole ovary at 0 to 1 months of age, but significantly decreased from 2 to 3 months of age. The volume of the cortex increased progressively after birth and reached approximately 60% of the total volume at 8 to 12 months of age. This significant development of the cortex coincided with the increased number and size of large follicles observed from 6 months of age. These results suggest that the development of the cortex plays a role in the maturation of the follicles and the equine ovary undergoes substantial morphological changes postnatally until puberty.

Broad integration of expression maps and co-expression networks compassing novel gene functions in the brain.

Using a recently invented technique for gene expression mapping in the whole-anatomy context, termed transcriptome tomography, we have generated a dataset of 36,000 maps of overall gene expression in the adult-mouse brain. Here, using an informatics approach, we identified a broad co-expression network that follows an inverse power law and is rich in functional interaction and gene-ontology terms. Our framework for the integrated analysis of expression maps and graphs of co-expression networks revealed that groups of combinatorially expressed genes, which regulate cell differentiation during development, were present in the adult brain and each of these groups was associated with a discrete cell types. These groups included non-coding genes of unknown function. We found that these genes specifically linked developmentally conserved groups in the network. A previously unrecognized robust expression pattern covering the whole brain was related to the molecular anatomy of key biological processes occurring in particular areas.

Transcriptome tomography for brain analysis in the web-accessible anatomical space.

Increased information on the encoded mammalian genome is expected to facilitate an integrated understanding of complex anatomical structure and function based on the knowledge of gene products. Determination of gene expression-anatomy associations is crucial for this understanding. To elicit the association in the three-dimensional (3D) space, we introduce a novel technique for comprehensive mapping of endogenous gene expression into a web-accessible standard space: Transcriptome Tomography. The technique is based on conjugation of sequential tissue-block sectioning, all fractions of which are used for molecular measurements of gene expression densities, and the block- face imaging, which are used for 3D reconstruction of the fractions. To generate a 3D map, tissues are serially sectioned in each of three orthogonal planes and the expression density data are mapped using a tomographic technique. This rapid and unbiased mapping technique using a relatively small number of original data points allows researchers to create their own expression maps in the broad anatomical context of the space. In the first instance we generated a dataset of 36,000 maps, reconstructed from data of 61 fractions measured with microarray, covering the whole mouse brain (ViBrism: http://vibrism.riken.jp/3dviewer/ex/index.html) in one month. After computational estimation of the mapping accuracy we validated the dataset against existing data with respect to the expression location and density. To demonstrate the relevance of the framework, we showed disease related expression of Huntington's disease gene and Bdnf. Our tomographic approach is applicable to analysis of any biological molecules derived from frozen tissues, organs and whole embryos, and the maps are spatially isotropic and well suited to the analysis in the standard space (e.g. Waxholm Space for brain-atlas databases). This will facilitate research creating and using open-standards for a molecular-based understanding of complex structures; and will contribute to new insights into a broad range of biological and medical questions.

Local nucleosome dynamics facilitate chromatin accessibility in living mammalian cells.

Genome information, which is three-dimensionally organized within cells as chromatin, is searched and read by various proteins for diverse cell functions. Although how the protein factors find their targets remains unclear, the dynamic and flexible nature of chromatin is likely crucial. Using a combined approach of fluorescence correlation spectroscopy, single-nucleosome imaging, and Monte Carlo computer simulations, we demonstrate local chromatin dynamics in living mammalian cells. We show that similar to interphase chromatin, dense mitotic chromosomes also have considerable chromatin accessibility. For both interphase and mitotic chromatin, we observed local fluctuation of individual nucleosomes (~50 nm movement/30 ms), which is caused by confined Brownian motion. Inhibition of these local dynamics by crosslinking impaired accessibility in the dense chromatin regions. Our findings show that local nucleosome dynamics drive chromatin accessibility. We propose that this local nucleosome fluctuation is the basis for scanning genome information.