Mammogram classification is directly related to computer-aided diagnosis of breast cancer. Traditional methods rely on regions of interest (ROIs) which require great efforts to annotate. Inspired by the success of using deep convolutional features for natural image analysis and multi-instance learning (MIL) for labeling a set of instances/patches, we propose end-to-end trained deep multiinstance networks for mass classification based on whole mammogram without the aforementioned ROIs. We explore three different schemes to construct deep multi-instance networks for whole mammogram classification. Experimental results on the INbreast dataset demonstrate the robustness of proposed networks compared to previous work using segmentation and detection annotations. 1
Three-dimensional (3D) cell culture provides an effective way over conventional two-dimensional (2D) monolayer culture to more closely imitate the complex cellular organization, heterogeneity, and interactions as well as tissue microenvironments in vivo. Here we present a novel droplet-based 3D cell culture method by using droplet array attached on the sidewall of a PDMS piece. Such an arrangement not only avoids cells from adhering on the chip surface for achieving 3D cell culture in nanoliter-scale droplets, but also facilitates performing multiple operations to cells in droplets, including cell suspension droplet generation, drug treatment, and cell staining with a capillary-based liquid handling system, as well as in situ observation and direct scanning with a confocal laser scanning microscope. We optimized the system by studying the effects of various conditions to cell culture including droplet volume, cell density and fabrication methods of the PDMS pieces. We have applied this system in the 3D culture of HepG2 cells and the stimulation testing of an anticancer drug, doxorubicin, to 3D cell spheroids.
Establishing cell migration assays in multiple different microenvironments is important in the study of tissue repair and regeneration, cancer progression, atherosclerosis, and arthritis. In this work, we developed a miniaturized and massive parallel microfluidic platform for multiple cell migration assays combining the traditional membrane-based cell migration technique and the droplet-based microfluidic technique. Nanoliter-scale droplets are flexibly assembled as building blocks based on a porous membrane to form microdroplet chains with diverse configurations for different assay modes. Multiple operations including in-droplet 2D/3D cell culture, cell co-culture and cell migration induced by a chemoattractant concentration gradient in droplet chains could be flexibly performed with reagent consumption in the nanoliter range for each assay and an assay scale-up to 81 assays in parallel in one microchip. We have applied the present platform to multiple modes of cell migration assays including the accurate cell migration assay, competitive cell migration assay, biomimetic chemotaxis assay, and multifactor cell migration assay based on the organ-on-a-chip concept, for demonstrating its versatility, applicability, and potential in cell migration-related research.
Abstract. Mammogram classification is directly related to computer-aided diagnosis of breast cancer. Traditional methods requires great effort to annotate the training data by costly manual labeling and specialized computational models to detect these annotations during test. Inspired by the success of using deep convolutional features for natural image analysis and multi-instance learning for labeling a set of instances/patches, we propose end-to-end trained deep multiinstance networks for mass classification based on whole mammogram without the aforementioned costly need to annotate the training data. We explore three different schemes to construct deep multi-instance networks for whole mammogram classification. Experimental results on the INbreast dataset demonstrate the robustness of proposed deep networks compared to previous work using segmentation and detection annotations in the training.
Herein, we developed an automated and flexible system for performing miniaturized liquid-liquid reactions and assays in the femtoliter to picoliter range, by combining the contact printing and the droplet-based microfluidics techniques. The system mainly consisted of solid pins and an oil-covered hydrophilic micropillar array chip fixed on an automated x- y- z translation stage. A novel droplet manipulation mode called "dipping-depositing-moving" (DDM) was proposed, which was based on the programmable combination of three basic operations, dipping liquids and depositing liquids with the solid pins and moving the two-dimensional oil-covered hydrophilic pillar microchip. With the DDM mode, flexible generation and manipulation of small droplets with volumes down to 179 fL could be achieved. For overcoming the scale phenomenon specially appeared in picoliter-scale droplets, we used a design of water moat to protect the femtoliter to picoliter droplets from volume loss through the cover oil during the droplet generation, manipulation, reaction and assay processes. Moreover, we also developed a precise quantitative method, quantitative droplet dilution method, to accurately measure the volumes of femtoliter to picoliter droplets. To demonstrate its feasibility and adaptability, we applied the present system in the determination of kinetics parameter for matrix metalloproteinases (MMP-9) in 1.81 pL reactors and the measurement the activity of β-galactosidase in single cells (HepG2 cells) in picoliter droplet array. The ultrasmall volumes of the droplet reactors avoided the excessive dilution to the reaction solutions and enabled the highly sensitive measurement of enzyme activity in the single cell level.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.