Search

Chips with tens of billions of transistors have become today's norm. These designs are straining our electronic design automation tools throughout the design process, requiring ever more computational resources. In many tools, parallelisation has improved both latency and throughput for the designer's benefit. However, tools largely remain restricted to a single machine and in the case of RTL simulation, we believe that this leaves much potential performance on the table. We introduce Metro-MPI to improve RTL simulation for modern 10 billion transistor-scale chips. Metro-MPI exploits the natural boundaries present in chip designs to partition RTL simulations and leverage High Performance Computing (HPC) techniques to extract parallelism. For chip designs that scale in size by exploiting latency-insensitive interfaces like networks-on-chip and AXI, Metro-MPI offers a new paradigm for RTL simulation scalability. Our implementation of Metro-MPI in Open-Piton+Ariane delivers 2.7 MIPS of RTL simulation throughput for the first time on a design with more than 10 billion transistors and 1,024 Linux-capable cores, opening new avenues for distributed RTL simulation of emerging system-on-chip designs. Compared to sequential and multithreaded RTL simulations of smaller designs, Metro-MPI achieves up to 135.98× and 9.29× speedups. Similarly, for a representative regression run, Metro-Mpireduces energy consumption by up to 2.53× and 2.91× ., This work has been partially supported by the Spanish Ministry of Economy and Competitiveness (contract PID2019-107255GB-C21), by the Generalitat de Catalunya (contract 2017-SGR-1328), by the European Union within the framework of the ERDF of Catalonia 2014-2020 under the DRAC project [001-P-001723], and by the Arm-BSC Center of Excellence. G. Lopez-Paradís has been supported by the Generalitat de Catalunya through a FI fellowship 2021FI-B00994 and GSoC 2021, and M. Moreto by a Ramon y Cajal fellowship no. RYC-2016-21104. A. Armejach is a Serra Hunter Fellow., Peer Reviewed, Postprint (author's final draft)

The hematopoietic niche is a supportive microenvironment composed of distinct cell types, including specialized vascular endothelial cells that directly interact with hematopoietic stem and progenitor cells (HSPCs). The molecular factors that specify niche endothelial cells and orchestrate HSPC homeostasis remain largely unknown. Using multi-dimensional gene expression and chromatin accessibility analyses in zebrafish, we define a conserved gene expression signature and cis-regulatory landscape that are unique to sinusoidal endothelial cells in the HSPC niche. Using enhancer mutagenesis and transcription factor overexpression, we elucidate a transcriptional code that involves members of the Ets, Sox, and nuclear hormone receptor families and is sufficient to induce ectopic niche endothelial cells that associate with mesenchymal stromal cells and support the recruitment, maintenance, and division of HSPCs in vivo. These studies set forth an approach for generating synthetic HSPC niches, in vitro or in vivo, and for effective therapies to modulate the endogenous niche.

There can be numerous electronic components on a given PCB, making the task of visual inspection to detect defects very time-consuming and prone to error, especially at scale. There has thus been significant interest in automatic PCB component detection, particularly leveraging deep learning. However, deep neural networks typically require high computational resources, possibly limiting their feasibility in real-world use cases in manufacturing, which often involve high-volume and high-throughput detection with constrained edge computing resource availability. As a result of an exploration of efficient deep neural network architectures for this use case, we introduce PCBDet, an attention condenser network design that provides state-of-the-art inference throughput while achieving superior PCB component detection performance compared to other state-of-the-art efficient architecture designs. Experimental results show that PCBDet can achieve up to 2$\times$ inference speed-up on an ARM Cortex A72 processor when compared to an EfficientNet-based design while achieving $\sim$2-4\% higher mAP on the FICS-PCB benchmark dataset., Comment: 7 pages, 6 figures

Self-supervised pretraining has been observed to improve performance in supervised learning tasks in medical imaging. This study investigates the utility of self-supervised pretraining prior to conducting supervised fine-tuning for the downstream task of lung sliding classification in M-mode lung ultrasound images. We propose a novel pairwise relationship that couples M-mode images constructed from the same B-mode image and investigate the utility of data augmentation procedure specific to M-mode lung ultrasound. The results indicate that self-supervised pretraining yields better performance than full supervision, most notably for feature extractors not initialized with ImageNet-pretrained weights. Moreover, we observe that including a vast volume of unlabelled data results in improved performance on external validation datasets, underscoring the value of self-supervision for improving generalizability in automatic ultrasound interpretation. To the authors' best knowledge, this study is the first to characterize the influence of self-supervised pretraining for M-mode ultrasound., Comment: 10 pages, 6 figures, to be published in the 2023 Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops

In this study, we investigated whether self-supervised pretraining could produce a neural network feature extractor applicable to multiple classification tasks in B-mode lung ultrasound analysis. When fine-tuning on three lung ultrasound tasks, pretrained models resulted in an improvement of the average across-task area under the receiver operating curve (AUC) by 0.032 and 0.061 on local and external test sets respectively. Compact nonlinear classifiers trained on features outputted by a single pretrained model did not improve performance across all tasks; however, they did reduce inference time by 49% compared to serial execution of separate fine-tuned models. When training using 1% of the available labels, pretrained models consistently outperformed fully supervised models, with a maximum observed test AUC increase of 0.396 for the task of view classification. Overall, the results indicate that self-supervised pretraining is useful for producing initial weights for lung ultrasound classifiers., Comment: 10 pages, 5 figures, submitted to IEEE Access

Microfluidic technologies have in conjunction with the field of biomedical research for their utility in making highly precise measurements of biological and chemical specimens. These measurements help us search for specific analytes that can be associated with biological states such as disease in a process called biomarker discovery. The analytes in question are the biomarkers of the states they represent, and these associations become part of the greater body of knowledge that guide other researchers in the recognition and study of other biological phenomena. Eventually, this work helps guide clinicians in making diagnostic and prognostic decisions. In this dissertation, I will describe work done to engineer and implement two microfluidic technologies designed to perform biomarker discovery via different modalities, albeit with both analyzing single-cell phenotypes. The first of these devices, called mechano-node-pore sensing, focused on the mechanical properties of cells, and how the mechanics of the cell as a material can reveal information about drug resistance in acute promyelocytic leukemia. The second of these devices instead analyzed cells for surface proteins using a novel scheme for chemical immobilization of antibody probes. In both cases, I demonstrate how mechanical properties and surface proteins can be used as biomarkers to discriminate between cell phenotypes, and how such measurements can help us more deeply understand a variety of biological phenomena, and eventually make advancements in biomedicine.

Viscoelastic properties of cells provide valuable information regarding biological or clinically relevant cellular characteristics. Here, we introduce a new, electronic-based, microfluidic platform-visco-node-pore sensing (visco-NPS)-which quantifies cellular viscoelastic properties under periodic deformation. We measure the storage (G) and loss (G″) moduli (i.e., elasticity and viscosity, respectively) of cells. By applying a wide range of deformation frequencies, our platform quantifies the frequency dependence of viscoelastic properties. G and G″ measurements show that the viscoelastic properties of malignant breast epithelial cells (MCF-7) are distinctly different from those of non-malignant breast epithelial cells (MCF-10A). With its sensitivity, visco-NPS can dissect the individual contributions of different cytoskeletal components to whole-cell mechanical properties. Moreover, visco-NPS can quantify the mechanical transitions of cells as they traverse the cell cycle or are initiated into an epithelial-mesenchymal transition. Visco-NPS identifies viscoelastic characteristics of cell populations, providing a biophysical understanding of cellular behavior and a potential for clinical applications.

Viscoelastic properties of cells provide valuable information regarding biological or clinically relevant cellular characteristics. Here, we introduce a new, electronic-based, microfluidic platform-visco-node-pore sensing (visco-NPS)-which quantifies cellular viscoelastic properties under periodic deformation. We measure the storage (G') and loss (G″) moduli (i.e., elasticity and viscosity, respectively) of cells. By applying a wide range of deformation frequencies, our platform quantifies the frequency dependence of viscoelastic properties. G' and G″ measurements show that the viscoelastic properties of malignant breast epithelial cells (MCF-7) are distinctly different from those of non-malignant breast epithelial cells (MCF-10A). With its sensitivity, visco-NPS can dissect the individual contributions of different cytoskeletal components to whole-cell mechanical properties. Moreover, visco-NPS can quantify the mechanical transitions of cells as they traverse the cell cycle or are initiated into an epithelial-mesenchymal transition. Visco-NPS identifies viscoelastic characteristics of cell populations, providing a biophysical understanding of cellular behavior and a potential for clinical applications.

Advancements in microfluidic technologies have led to the development of many new tools for both the characterization and sorting of single cells without the need for exogenous labels. Label-free microfluidics reduce the preparation time, reagents needed, and cost of conventional methods based on fluorescent or magnetic labels. Furthermore, these devices enable analysis of cell properties such as mechanical phenotype and dielectric parameters that cannot be characterized with traditional labels. Some of the most promising technologies for current and future development toward label-free, single-cell analysis and sorting include electronic sensors such as Coulter counters and electrical impedance cytometry; deformation analysis using optical traps and deformation cytometry; hydrodynamic sorting such as deterministic lateral displacement, inertial focusing, and microvortex trapping; and acoustic sorting using traveling or standing surface acoustic waves. These label-free microfluidic methods have been used to screen, sort, and analyze cells for a wide range of biomedical and clinical applications, including cell cycle monitoring, rapid complete blood counts, cancer diagnosis, metastatic progression monitoring, HIV and parasite detection, circulating tumor cell isolation, and point-of-care diagnostics. Because of the versatility of label-free methods for characterization and sorting, the low-cost nature of microfluidics, and the rapid prototyping capabilities of modern microfabrication, we expect this class of technology to continue to be an area of high research interest going forward. New developments in this field will contribute to the ongoing paradigm shift in cell analysis and sorting technologies toward label-free microfluidic devices, enabling new capabilities in biomedical research tools as well as clinical diagnostics. This article is categorized under: Diagnostic Tools > Biosensing Diagnostic Tools > Diagnostic Nanodevices.

Analysis of hematopoietic stem cell function in nonhuman primates provides insights that are relevant for human biology and therapeutic strategies. In this study, we applied quantitative genetic barcoding to track the clonal output of transplanted autologous rhesus macaque hematopoietic stem and progenitor cells over a time period of up to 9.5 months. We found that unilineage short-term progenitors reconstituted myeloid and lymphoid lineages at 1 month but were supplanted over time by multilineage clones, initially myeloid restricted, then myeloid-B clones, and then stable myeloid-B-T multilineage, long-term repopulating clones. Surprisingly, reconstitution of the natural killer (NK) cell lineage, and particularly the major CD16(+)/CD56(-) peripheral blood NK compartment, showed limited clonal overlap with T, B, or myeloid lineages, and therefore appears to be ontologically distinct. Thus, in addition to providing insights into clonal behavior over time, our analysis suggests an unexpected paradigm for the relationship between NK cells and other hematopoietic lineages in primates.

Motivated by geometric origami and the straight skeleton construction, we outline a map between spaces of phylogenetic trees and spaces of planar polygons. The limitations of this map is studied through explicit examples, culminating in proving a structural rigidity result., Comment: 17 pages, 12 figures