September 09, 2019 –
Summer Undergraduate Research Fellows in the Internet of Things (SURF-IoT) presented their projects Thursday, Aug. 29, at the SURF-IoT Symposium at the CALIT2 Auditorium.
Thirteen students shared the results of their 10-week research efforts by presenting final reports at the closing-day event.
The SURF-IoT program, co-sponsored by UCI’s Undergraduate Research Opportunities Program (UROP) and CALIT2, provides students with a unique experience. Each student has the guidance of a UCI faculty mentor, along with the opportunity to gain experience and advanced training in state-of-the-art facilities and techniques.
To learn more about SURF-IoT visit here.
Here is a brief overview of the 2019 projects and participants.
mHealth App to Assist Caregivers of Pediatric Cancer Patients
Fellow: Shweta Karkee
Mentors: Michelle Fortier, PhD; Lindsay Jibb, PhD; Haydee Cortes, BA; Beverly Mendoza, BA; Lessley Torres, BA
Pediatric cancer pain is common, has negative psychosocial consequences for children and caregivers (impeded cancer recovery, child and family distress, chronic pain in survivorship) and is a significant cost to healthcare systems and families. Additionally, the de-hospitalization of childhood cancer care means that significant proportions of pediatric cancer pain are experienced in environments where treatment options are limited (e.g., at home, school). In response, Pain Squad+, a smartphone-based real-time pain assessment and management clinical decision support system for adolescents with cancer (12-18 years) was developed. Evaluation of Pain Squad+ has demonstrated the feasibility of mobile health (mHealth) use among adolescents. However, young children with cancer are particularly vulnerable to under-managed pain due to their limited ability for pain self-report and reliance on caregivers for pain treatment.
The objective of our research is to collect qualitative data from caregivers and clinicians in order to find the pain points in outpatient care, and tailor a caregiver-led pain management app that caters to these needs.
Understanding the Transmission and the Spread of Antibiotic Resistance through Human Sewage
Fellow: Loan Le
Mentor: Dr. Sunny Jiang
The rapid spread of antibiotic resistance genes in the environment is an emerging threat to global public health. Human sewage has been considered as a major source of antibiotic resistance gene (ARGs) because the interaction of high abundance of microbial population with the residual antibiotics in sewage favor the emergence of antibiotic resistant bacteria. However, the variability of resistant profile in different treatment plants and the efficacy of degrade the resistance gene through sewage treatment have not been well characterized. In this study, the occurrence and abundance of the antibiotic resistance genes were investigated from sixteen sewage samples collected from different wastewater treatment plants (Washington D.C., Los Angeles County, and Orange County) using next-generation sequencing (NGS) and metagenomic analyses. Genes coding for antibiotic resistance were identified using BLAST analysis against MEGARes database. Results showed RpoB, tlrC, TUFAB, OXA, and gyrA as the prominent antibiotic resistance genes in all sewage samples. Despite these similarities in resistant profiles, sewage samples from different treatment plants had significant variability in the abundance of aadA11, MPHA, and VanA genes. Antibiotic resistance gene encoding resistance toward aminoglycoside was more abundant in Orange County samples compared to two others. While the relative abundance of resistant genes was reduced upon the treatment, high prevalence of genes RpoB, TUFAB, OXA, VanA, and gyrA in the final effluents highlighted the need to improve wastewater treatment processes. The results provide an insight of ARGs prevalence in municipal wastewaters for better understanding on the ARGs dissemination through human sewage. Comparison of different treatment processes in relationship to antibiotic resistance removal will help to identify strategies to reduce the spread of antibiotic resistance through sewage discharge.
Pain Assessment with Wearable Electronics
Fellow: Tingjue Yin
Mentor: Emad Kasaeyan Naeini
Professor: Nikil Dutt
Team members (I-Surf): Hyung Jun Gu, Youngwoo Cho
Pain scales and questionnaires are commonly used to assess pain but they are not accurate because pain is subjective and multi-dimensional. Also, these conventional methods are unable to be performed on unconscious patients, such as in deep sleep or with mental disorder. It is a potential solution of using biophysical data from the individual to predict the pain. The study worked on a dataset of biopotentials, classifying pain into five intensity levels and applying the machine learning method and then, evaluated the testing quality. The machine learning method, including support vector machine and random forest classifications, have the similar results – they can predict between no pain and the most severe pain with the highest accuracy score 0.84. When it comes to distinguish pain levels with some pain intensities, the accuracy score decreases to 0.55. The study provides a feasible approach to assess pain but calls for more future work to increase the accuracy of recognizing different pain levels.
Web-based Ground Control Station
Fellow: Lyuyang Hu
Mentor: Professor Peter Burke
The popularity of drone technologies has increased significantly in recent years because of the advancement of piloting software from companies including DJI and Ardupilot. The ground control stations (GCSs), however, still rely on offline software like Mission Planner and DJI Go which imposes limitations on control range, number of pilots and number of connected drones. The project aims to design and build a web-based ground control station that overcomes all three limitations by using cloud technologies. A software architecture that uses Micro Air Vehicle Communication Protocol (MAVLink), WebSocket and message broker is proposed and a preliminary version of the web-based GCS is implemented using Django framework, Django Channels, MySQL database, Redis, pymavlink and Map Box. Such a web-based GCS is suitable for professional film crews, construction teams and individual drone users. The next step of the project is to deploy the software and to continue adding more GCS and security features.
An Experimental Environment for Mobile Sensor Deployment
Fellow: Wenzhuo Wang
Mentor: Solmaz Kia
Ph.D. Mentor: Navid Rezazadeh
Stationary surveillance systems are only able to cover limited area. The number of monitors is restricted by the bandwidth and cost, so it is required to develop a system with mobile sensor. The mobile sensors could cover a relatively large area by deploying them to different position. This project is similar to drone navigation. It requires a system which could maintain stable flight of each quadcopter. Then it uses Optitrack to locate the quadcopter. The main code is written by python. Currently, I am working with stabilizing the quadcopter with the location data. After we could control the stable flight, we will make a user-friendly interface for input. This project is still in starting stage, so there is less outcome now.
Interfacing IoT Data Collection Systems with Smart Applications
Fellow: Leo Peng
Mentor: Professor Sharad Mehrotra
Postdoc Mentors: Dr. Roberto Yus & Dr. Georgios Bouloukakis
A smart space is a space with an embedded system of Internet of Things (IoT) devices
such as sensors and actuators. Using those, useful applications can be developed to improve comfort, productivity, security, and more. Due to the heterogeneity of IoT devices and the lack of a widely used standard, developing applications for smart spaces can be challenging. Smart space application developers must repeatedly overcome the challenges of interoperability (i.e., allowing communication between diverse devices) and reusability (i.e., developing applications that can be reused in different contexts/spaces regardless of the underlying infrastructure of devices). This study relieves these challenges through the development of a framework for IoT
smart spaces called SemIoTic. SemIoTic incorporates relationships between the low-level communication protocols of IoT devices and real-world concepts such as people, spaces, and observable attributes. During this summer project, we focused on the development of the SemIoTic Application Programming Interface (API) that allows developers to easily associate observable attributes with physical entities and collect comprehensible data from sensors in order to develop smart space applications. The design of the API was based on an extensible metamodel used to define different smart spaces and the devices within them. We have tested the API by modeling a smart space (Donald Bren Hall) and including different sensory data captured within it. In the following stages of this study, we will integrate the SemIoTic API with real sensor data from buildings across campus and allow students to develop their own smart applications for UCI.
Exploring the Design of Fertility Tracking Tools
Fellow: Thu Huynh
Mentor: Professor Yunan Chen, Ph.D. candidate Mayara Costa Figueiredo
Fertility mobile applications are getting more and more popular with a vast number of them released to the market, which makes it difficult for the users to navigate around. In addition, most apps lack the option to tailor the interface based on the users’ goals. The purpose of our project is to explore the possible designs of fertility apps to elevate the users’ experience and offer them the ability to customize the app features in a way that suits their needs. To do this, we chose 31 most popular apps on both Android and iOS platform and conducted app evaluation to see the correlation between the goals that they claim to support and the tracking indicators they offer. We also downloaded and analyzed all the reviews of each app then sorted them by goals, issues, positive aspects and keywords. This way, we can gain insight into what the users’ pain points are and which aspects of these apps could be improved. We did come up with a final design and may conduct user testing sessions to get user feedback in the future.
Title: Analyzing Smartwatch Data of Dementia Caregivers (Pilot Study)
Fellow: Anthony Park
Faculty Mentors: Professor Nikil Dutt, Professor Amir Rahmani, Professor Jung-ah Lee
Graduate Student Mentor: Sina Labbaf
I-Surf Fellow: Jeonghwa Jeong
Dementia is a major public health problem affecting 5.7 million people in the United States. Prior research shows that Asian American minority families with Alzheimer patients underuse dementia care services due to the language barrier, cultural characteristics, and stigma related to mental illness in general. Importantly, some minority caregivers expressed severe depression. Therefore, there is an urgent need to improve dementia caregiving skills and stress management skills for the family caregivers including ethnic minorities. With the culturally sensitively designed home visitations to participated caregivers, wearable IoT devices(Garmin smartwatch) was implemented to the caregivers to help the study team monitor their physiological changes. In order to examine the effectiveness of the home visit program, the study team analyzed the caregivers’ stress levels measured by the smartwatches during and after the intervention periods. As a result, the average trend of a slight stress reduction was found during the home visit intervention. The study team, however, faced limitations regarding the feasibility of caregivers wearing the smartwatches consistently, and analyzing processed data provided by the Garmin cloud. Therefore, introducing smart-ring is being considered for the future participating caregivers. The smart-rings are expected to ease the difficulties of wearing and provide raw heart rate signal(PPG) for the objective signal processing.
Mathematical Modeling of Water Network Simulations Using Maude Rewriting Logic
Fellow: Lily McBeath
Faculty Mentor: Dr. Nalini Venkatasubramanian
Graduate Mentor: Qing Han
Water networks are particularly vulnerable to disasters such as earthquakes, often causing pipe leaks and breaks, water contamination, and unavailability of drinking water. Additionally, the behavior of water networks is determined by complex, interdependent factors, making it a challenge to identify and respond to failures quickly in a disaster situation. In this project we present the rewriting logic Maude as a possible solution to this problem. We will use the reflective capabilities of Maude to develop a high-level formal model of water network behavior as simulated in the Water Network Tool for Resilience (WNTR). We will also examine the possibility of using a formal model to quickly identify and analyze the effects of large-scale disasters on a water network using critical points.
RFID Localization with a Single Dynamic Antenna
Fellow: Kenzo Spaulding
Mentors: Professor Solmaz Kia and Mr. Navid Rezazadeh
Most RFID-based localization systems require an array of RFID antenna. In order to reduce cost, we utilized a tilt-yaw servo mount to implement an algorithm for localizing a target RFID tag in a grid of reference RFID tags with a single antenna. This means that the system will be capable of localizing a target tag in a two-dimensional space that was accessible before searching. Additionally, the system gathers data and then finds the target tag in post processing, allowing it to track more than one target tag simultaneously and estimate location accurately despite a changing environment.
A More Interactive Texera: Real-time Engine Status Inspection Using WebSocket
Fellowt: Yinan Zhou
Mentors: Ph.D. Shengquan Ni; Professor Chen Li
Texera is a web-based big data analytics system that has been developed by a group of students under the instruction of Professor Chen Li. One of the features we envisioned is to allow users to visualize and manipulate the status of the backend engine easily on the frontend user interface. To achieve such goal, implementing a WebSocket is necessary since a lot of information will need to be communicated back and forth between the frontend and the backend in real-time. My work this summer consists of three main parts. First, I give the two ends with the ability to send and accept a WebSocket request so that a consistent connection can be established. Secondly, I extracted a set of useful engine information such as operator status and processing speed to be sent to the frontend. Last but not least, I displayed such information on the frontend in a way that the users can easily access it. In the future, more information will be provided to the users and users will be given more power over manipulating the backend during its execution.
Sensor GraphX, a User Interface for Indoor Mapping Localization System on Android Platform
Fellow: Anthony Skoury
Mentors: Professor Andrei Shkel, Chi-Shih Jao, Sina Askari
Navigation technology is becoming more advanced by the day, both in phones and specialized systems. Although our mobile devices have sensors embedded into them, they are not as accurate as specialized systems. Such systems however, need a mobile and inexpensive platform for the end user to display data from these devices. The development of the Android application Sensor GraphX solves this issue by connecting to the existing navigation system through the system’s Wi-Fi and graphs all the necessary data to display your position. With numerous customization features and options, the user can make the most out of the system’s intended purpose: indoor navigation on a three-dimensional axis. Users can upload a floor plan of the building as a background and scale the graph to fit their needs. Although this Wi-Fi connection is versatile and allows communication between the devices, it also has a downside, limiting the frequency the application can pull data from the sensor device. Despite this, the app proves to be effective, holding the same results as the original simulation program developed for desktops. Accessibility are mobility are important for practical usage in systems intended for first responders and having an Android application for mobile devices fits that need.
Automated Cardiac MRI Segmentation
Fellow: Yilei Wu
Mentors: Saeed Karimi Bidhendi
Faculty Advisor: Professor Hamid Jafarkhani(Dept of EECS)
Professor Arash Kheradvar(Dept of Biomedical Engineering)
Congenital heart disease (CHD) is the most common class of birth defects and is responsible for 30-50% of congenital infant mortality. Over the past decade, the Cardiac MRI (CMR) imaging technique is extensively used to monitor structural and functional heart condition for CHD patients to avoid late diagnosis and mistreatment. Precise chamber segmentation is the main step toward analyzing the CMR data; however, manual segmentation by expert physicians and radiologists is a tedious and time-consuming task. Motivated by the recent success of deep learning in computer vision task, we proposed a deep learning framework for an automated segmentation of heart chambers and provided a baseline result on a private set of CMR data.