Browsing by Author "Li, Yimei"

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Compressed Sensing in Multi-Hop Large-Scale Wireless Sensor Networks Based on Routing Topology Tomography
    (IEEE, 2018) Li, Yimei; Liang, Yao; Computer and Information Science, School of Science
    Data acquisition from multi-hop large-scale outdoor wireless sensor network (WSN) deployments for environmental monitoring is full of challenges. This is because of the severe resource constraints on tiny battery-operated motes (e.g., bandwidth, memory, power, and computing capacity), the data acquisition volume from large-scale WSNs, and the highly dynamic wireless link conditions in outdoor harsh communication environments. We present a novel compressed sensing approach, which can recover the sensing data at the sink with high fidelity when a very few data packets need to be collected, leading to a significant reduction of the network transmissions and thus an extension of the WSN lifetime. Interplaying with the dynamic WSN routing topology, the proposed approach is both efficient and simple to implement on the resource-constrained motes without motes' storing of any part of the random projection matrix, as opposed to other existing compressed sensing-based schemes. We further propose a systematic method via machine learning to find a suitable representation basis, for any given WSN deployment and data field, which is both sparse and incoherent with the random projection matrix in compressed sensing for data collection. We validate our approach and evaluate its performance using a real-world outdoor multihop WSN testbed deployment in situ. The results demonstrate that our approach significantly outperforms existing compressed sensing approaches by reducing data recovery errors by an order of magnitude for the entire WSN observation field while drastically reducing wireless communication costs at the same time.
  • Thumbnail Image
    Item
    A Networked Sensor System for the Analysis of Plot-Scale Hydrology
    (MDPI, 2017-03-20) Villalba, German; Plaza, Fernando; Zhong, Xiaoyang; Davis, Tyler W.; Navarro, Miguel; Li, Yimei; Slater, Thomas A.; Liang, Yao; Liang, Xu; Computer and Information Science, School of Science
    This study presents the latest updates to the Audubon Society of Western Pennsylvania (ASWP) testbed, a $50,000 USD, 104-node outdoor multi-hop wireless sensor network (WSN). The network collects environmental data from over 240 sensors, including the EC-5, MPS-1 and MPS-2 soil moisture and soil water potential sensors and self-made sap flow sensors, across a heterogeneous deployment comprised of MICAz, IRIS and TelosB wireless motes. A low-cost sensor board and software driver was developed for communicating with the analog and digital sensors. Innovative techniques (e.g., balanced energy efficient routing and heterogeneous over-the-air mote reprogramming) maintained high success rates (>96%) and enabled effective software updating, throughout the large-scale heterogeneous WSN. The edaphic properties monitored by the network showed strong agreement with data logger measurements and were fitted to pedotransfer functions for estimating local soil hydraulic properties. Furthermore, sap flow measurements, scaled to tree stand transpiration, were found to be at or below potential evapotranspiration estimates. While outdoor WSNs still present numerous challenges, the ASWP testbed proves to be an effective and (relatively) low-cost environmental monitoring solution and represents a step towards developing a platform for monitoring and quantifying statistically relevant environmental parameters from large-scale network deployments.
  • Thumbnail Image
    Item
    NFS-18. Lower Body Surface Area is Associated with Increased Likelihood of Plexiform Neurofibroma Response to MEK Inhibition
    (Oxford University Press, 2024-06-18) Kotch, Chelsea; Dombi, Eva; Gross, Andrea; Weiss, Brian; Mueller, Sabine; Reddy, Alyssa T.; Perreault, Sébastien; Alves, Mélanie; Brown, Symone; Li, Yimei; Widemann, Brigitte C.; Fisher, Michael J.; Pediatrics, School of Medicine
    BACKGROUND: MEK inhibitors (MEKi) are altering the management approach for plexiform neurofibroma (PN), with high rates of treatment response to multiple MEKi. Despite these successes, a subset of PN fail to respond and little is known about the clinical features associated with treatment response. METHODS: We performed a retrospective cohort study integrating clinical trial data (NCT01362803, NCT02407405, NCT02096471, NCT03231306, NCT03363217) to identify baseline clinical features associated with response of PN to MEKi. Partial response (PR) was defined as ≥20 percent reduction in tumor volume from baseline. RESULTS: Of 232 eligible participants, adequate clinical trial and imaging data was available for 223 participants. In the primary analysis of 184 participants with central response evaluation, the median age was 15.2 years with a median tumor volume of 488 milliliters at clinical trial enrollment. One hundred and eighteen (64%) participants achieved a PR with median time to PR of 8 cycles. Thirty-five participants (19%) required a dose reduction prior to 6 cycles of therapy due to toxicity. Younger age and lower body surface area (BSA) were significantly associated with PR in univariate analysis while female sex and typical PN appearance (versus nodular) on imaging approached significance. In multivariable analysis, only lower BSA was significantly associated with response while typical PN appearance approached significance. In the multivariable analysis of pediatric participants treated per BSA-based dosing, lower BSA was the only feature significantly associated with PR. In the expanded analysis of all 223 participants, lower BSA and typical PN appearance were significantly associated with PR. CONCLUSION: Lower BSA and typical appearance of PN were associated with PR to MEK inhibitors. Future studies of MEK inhibitor for PN should integrate tumor pharmacokinetic-pharmacodynamic analyses to prospectively explore the impact of BSA on treatment response.
  • Thumbnail Image
    Item
    A semiparametric likelihood-based method for regression analysis of mixed panel-count data
    (Wiley, 2018-06) Zhu, Liang; Zhang, Ying; Li, Yimei; Sun, Jianguo; Robison, Leslie L.; Biostatistics, School of Public Health
    Panel-count data arise when each study subject is observed only at discrete time points in a recurrent event study, and only the numbers of the event of interest between observation time points are recorded (Sun and Zhao, 2013). However, sometimes the exact number of events between some observation times is unknown and what we know is only whether the event of interest has occurred. In this article, we will refer this type of data to as mixed panel-count data and propose a likelihood-based semiparametric regression method for their analysis by using the nonhomogeneous Poisson process assumption. However, we establish the asymptotic properties of the resulting estimator by employing the empirical process theory and without using the Poisson assumption. Also, we conduct an extensive simulation study, which suggests that the proposed method works well in practice. Finally, the method is applied to a Childhood Cancer Survivor Study that motivated this study.
  • Thumbnail Image
    Item
    Towards Long-Term Multi-Hop WSN Deployments for Environmental Monitoring: An Experimental Network Evaluation
    (MDPI, 2014-12-05) Navarro, Miguel; Davis, Tyler W.; Villalba, German; Li, Yimei; Zhong, Xiaoyang; Erratt, Newlyn; Liang, Xu; Liang, Yao; Computer and Information Science, School of Science
    This paper explores the network performance and costs associated with the deployment, labor, and maintenance of a long-term outdoor multi-hop wireless sensor network (WSN) located at the Audubon Society of Western Pennsylvania (ASWP), which has been in operation for more than four years for environmental data collection. The WSN performance is studied over selected time periods during the network deployment time, based on two different TinyOS-based WSN routing protocols: commercial XMesh and the open-source Collection Tree Protocol (CTP). Empirical results show that the network performance is improved with CTP (i.e., 79% packet reception rate, 96% packet success rate and 0.2% duplicate packets), versus using XMesh (i.e., 36% packet reception rate and 46% packet success rate, with 3%–4% duplicate packets). The deployment cost of the 52-node, 253-sensor WSN is $31,500 with an additional $600 per month in labor and maintenance resulting in a cost of $184 m−2·y−1 of sensed area. Network maintenance during the first four years of operation was performed on average every 12 days, costing approximately $187 for each field visit.