A Wearable Vital Signs Monitoring Device Using Artificial Intelligent-enabled Internet of Medical Things

La Trobe University – Sheffield Hallam University Joint PhD Scholarship project

The purpose of this study is to devise new Internet of Things (IoT) technologies and integrate them into a miniaturized pulse oximetry device to monitor patient's vital signs such as the blood oxygen saturation level and respiration rate. Pulse oximetry allows blood's oxygenation to be monitored in a continuous, accurate, and non-invasive manner and from its signal a number of vital sign signals such as respiratory signal can be extracted. Respiration rate is one of the most important indicators of health deterioration in critically ill patients. Oximetry operates by using light emitting diodes to shine red and infrared light through the tissue and measuring the proportion of light absorbed by the blood. Normal oxygen saturation is usually between 96% and 98% and reduction from this level is considered dangerous and requires urgent medical attention. An IoT integrated, wearable, miniaturised pulse oximetry device will be designed, and its effectiveness will be evaluated. The novel aspects of the device include: innovations based on the latest IoT technologies to provide safe, reliable and robust communication between the device and clinicians, miniaturisation of the pulse oximetry device to a size that can be comfortably worn by patients from infants to adults and developments in data analysis using latest developments in artificial intelligence.


Prof Wei Xiang, Dr Robert Ross (La Trobe); Prof Reza Saatchi, Prof Heather Elphick (Sheffield Hallam)

D.Start Spring 2020 program “Seeing in Deep Water”

For undersea operations, poor visibility is due to water current, suspended particles, low illumination, water absorption of spectrum, and imaging sensor noise. Undersea image enhancement technicals are essential in many applications including search and rescue, underwater equipment inspection, scientific survey and remote undersea surveillance. This project aims to develop a real-time underwater video enhancement technology which will be a stand-alone low-power embedded system. The system can be carried by divers or fitted onto an underwater UAV. The system will greatly help divers to quickly identify underwater objects and recognize patterns. It will also aid the onshore office in decision making.

Learn more

Climate Smart Sugarcane Irrigation Partnerships (CSSIP)

Investigators: Wei Xiang, Bronson Philippa, Yvette Everingham, Eric Wang, Stephen Attard in collaboration with Geoff Inman-Bamber, Marian Davis, Andrew Schepen, Brock Dembowski, Peter Larsen and Andres Jaramillo

CSSIP will minimise nutrient runoff, improve soil health and increase wetlands water quality by facilitating the adoption of world-class irrigation practices in sugarcane farming systems. Currently, best practice irrigation is assisted by an Irrigation Decision Support Tool (IDST) that provides evidence-based advice. However, IDSTs have not reached their full potential. Firstly, they do not integrate short to medium term weather forecasts (e.g. weekly to multi-weekly forecasts). Secondly, IDSTs do not operate at a spatial scale relevant to farmers. CSSIP will incorporate the Bureau of Meteorology’s new high-resolution climate model into the Irrigation Decision Support Tool. Thirdly, IDSTs require substantial time in manual data entry, which can be alleviated using real-time monitoring via Internet of Things technologies. This will increase irrigation efficiency, reducing excessive runoff into river systems and onto the Reef, and, will help farmers save water and energy costs.

Applying new technologies to enhance biosecurity and cattle quality

Investigators: Ian Atkinson, Wei Xiang, Ron White, Stephanie Duce, Mohan Jacob and Karen Joyce

The vast natural environment of Northern Australia feeds the cattle industry; however, biosecurity threats have negatively impacted this. Conventional management of such threats such as weeds are not suited to such broad, harsh landscapes. The project will use an Internet of Things network with low-cost environmental sensors, drone mapping and big data analytics to develop and test data-driven, strategic pest management programs - ultimately improving both cattle industry and natural assets.

Vibration analysis of mining industry conveyor belt systems: validation of method and pathways to improvement

Investigators: Bronson Philippa, Bruce Belson, Lei Lei and Wei Xiang

Equipment failures in the mining industry can cause serious safety hazards and substantial financial losses. An automated, cost-effective monitoring system that could be retrofitted to existing equipment would provide advance warning to operators and reduce the likelihood of unscheduled outages. This project will test and validate a vibration-based monitoring system for conveyor belts and associated equipment. It will also identify improved methods to analyse the vibration data to increase the sensitivity and/or accuracy of the alerts that are generated.

Internet of Things and Big Data Analytics for Cairns Marine

Investigators: Wei Xiang, Tao Huang, Lei Lei and Mostafa Rahimi Azghadi

This project will assist to analyse the large repository of data held by Cairns marine in relation to its full operations, i.e., from harvest to husbandry, inventory tracking and sales fulfilment. The physical systems currently in use at Cairns Marine are extensive and complex, involving a range of activities in Northern Australia, all activities on site (including R&D) in Cairns and full product (marine animals) stewardship to destination. A deep and informed understanding of data is required by the business to meet its growing global product commitments.

JCU Mosquito Trap Development

Investigators: Kyran Staunton, Tom Burkot and Wei Xiang

To design and validate traps that are low cost and sensitive enough for Aedes aegypti and Aedes albopictus mosquitoes that they can be deployed for both SIT release surveillance during suppression and elimination operations, and also for sentinel surveillance after elimination.

Smart Ear Tag for Livestock

Investigators: Ian Atkinson, Wei Xiang, Bronson Philippa, Ed Charmley, Greg Bishop-Hurley, Nigel Bajema, Scott Mills, Gordon Foyster and Richard Keaney

CeresTag is investing in the development of a smart ear tag for livestock to enable near real-time geo-location and health monitoring. The developed ear tag will be compliant with the current NLIS identification system and cost only marginally more than existing tags. This technology will revolutionise the industry through enhanced animal welfare, improved land management practices and increased profitability. It will form the starting point of block chain traceability that will underpin the continued success of this important component of the Australian economy and help maintain the premium status of Australian livestock products.

Improving water quality for the Great Barrier Reef and wetlands by better managing irrigation in the sugarcane farming system

Investigators: Yvette Everingham, Wei Xiang and Bronson Philippa in collaboration with Stephen Attard

This project will work in partnership with industry, extension, NRM, research and government organisations to develop and deploy an irrigation system that is automatically controlled by remotely accessing feedback from the IrrigWeb decision support tool. Irrigweb provides optimal irrigation schedules on a paddock-by-paddock basis by linking information about climate, soils and management regimes. If new water quality targets as specified in the revised Burdekin Water Quality Improvement Plan are to be met by 2025, it will be critical to establish pathways that enable industry partners to capitalise on new technologies.

Janco Enterprise Pty Ltd

Investigators: Wei Xiang in collaboration with Kang Han

Machine learning sensor network to maintain optimum conditions and collect data for Molten Oxygen Electrolysis (MOE) reactions. Sensor data collection:  Supervisory capacity Machine learning logic for maintaining optimum reaction conditions on a provided hardware network Supervisory capacity Suggestion of industry standards for data collection and fail-safe implementation Suggestion of hardware components and configuration JCU’s scope in this project: Supervise on sensor quality before purchase and calibration logic once installed into PCB Supervise on any errors in the output value and possible causes Supervise on provided network structures for reliably and continuously recording data from multiple sensors to multiple SBC’s to an academic standard Supervise on methods for machine learning collection of data Supervise on methods provided for protecting sensor hardware in extreme conditions Check failsafe methods and data collection failsafes for industry standards requirements Consult on final hardware design for industrial conditional requirements (Industrial standards reports will be gathered prior to consultation, this is a 2nd check precaution) All mathematical modelling, programming, purchasing of hardware and construction shall be completed by JE Pty Ltd

Council improving the water quality of the Great Barrier Reef through the use of smart sensors and the IoT for urban water management

Investigators: Wei Xiang, HanShe Lim and Niels Munksgaard in collaboration with Lynne Powell

The primary aim of this grant application is to bring smart city technology into urban water management to improve urban water quality discharging to the Great Barrier Reef by: 1). Developing IOT technology to manage large data sets obtained from existing smart meters and water quality monitoring probes to make effective management decisions; and 2) Supporting the development of new cost effective, real time water quality monitoring technology. This grant application is for purchase of commercially available water quality monitoring probes suitable for a tropical urban stormwater environment, for supporting the development of new real time monitoring technology for nutrients; for the development of data analysis tools using IOT technology for both smart meter water consumption data, sewer pump station overflow data and stormwater water quality data so that the data is available in real time and can be used for effective decision making.

Securing Critical Data in the IoT – A Privacy-Preserving Biometric System

Investigators: Song Wang, Dennis Deng and Wencheng Yang

Nowadays IoT devices and smartphones are equipped with face ID or fingerprint sensors, making biometric-based authentication on a sharp growth curve. Trust and confidence in cyberspace are becoming all the more important. To ensure that sensitive information or critical data is accessed by genuine users or handled by authorised system operators, it is critical to protect users’ biometric information, which is stored as templates in databases or on smartcards. In this project, we apply machine learning algorithms to protecting biometric template data. The new biometric system will be equipped with strong security while rendering good recognition performance. Such a privacy-preserving biometric system will make access to critical data in the IoT safer and more reliable.

The application of AI to predict strategy outcomes and strategy proposals in team sports

Investigators: Zhen He, Aiden Nibali, Paul Gastin, David Carey

In this project we are using AI algorithms to help coaches of team sports to estimate the likely outcome of new strategy proposals. To achieve this we are developing novel trajectory prediction algorithms on real world team sports trajectory data.

The application of deep learning for estimating 3D pose and other applications

Investigator: Zhen He

3D human pose is an important step in understanding people in images and videos. In this project we are developing new 3D human pose estimation models from a single 2D image from an camera. This is a difficult task since in a 2D image you do not get an explicit value for the depth of each pixel. Using deep learning algorithms we are able to estimate the 3D locations of each joint using via cues in the 2D image. We are developing novel algorithms that use motion information from video to enhance the accuracy of the estimated human pose. We have also created a new 3D human pose dataset which contrasts from existing datasets by being marker less and also shot in real world conditions instead of in front of a green screen.

A software system for automated annotation of swimming videos using deep learning

Investigators: Zhen He and Stuart Morgan

This project is an extension of our previous project in which we developed a prototype deep learning algorithm to automatically annotate swimming video data. After our successful prototype system, we won an AIS competitive innovation grant to develop a production version of our prototype to be used at all swimming competitions, including the Olympic Games. The software automatically annotates all 8 or 10 lanes of races at near real time, replacing the hand annotation of many people. Achieving high accuracy at near real time annotation required us to overcome many challenges in terms of both deep learning algorithm design and very fast video data handling. In addition, we wrote a react based front end to the system. This end to end production quality system demonstrates the strong capability of our deep learning research group.

Classification methods for providing personalised and class decisions

Investigators: Hien Nguyen, Geoff McLachlan, Sharron Lee

This project provides a novel approach to the clustering of multivariate samples on entities in a class that automatically matches the sample clusters across the entities, allowing for inter-sample variation between the samples in a class. The project aims to develop a widely applicable, mixture-model-based framework for the simultaneous clustering of multivariate Isamples with inter-sample variation in a class and for the matching of the clusters across the entities in the class. The project will use a statistical approach to automatically match the clusters, since the overall mixture model provides a template for the class. It will provide a basis for discriminating between different classes in addition to the identification of atypical data points within a sample and of anomalous samples within a class. Key applications include biological image analysis and the analysis of data in flow cytometry which is one of the fundamental research tools for the life scientist.

Latent analysis, adversarial networks, and dimensionality reduction (LANDER)

Investigators: Hien Nguyen, Florence Forbes, Inria Grenoble-Rhone Alpes, Luke Prendergast, Kai Qin, Geoff McLachlan, Darren Wraith, Sharon Lee, Faicel Chamroukhi, Julyan Arbel, Stephane Girard, Antoine Usseglio-Carleve,

LANDER is a project-team started in January 2019 with French and Australian partners. The team federates researchers from Queensland University of Technology and Univ. of Queensland, Brisbane; La Trobe Univ. and Swinburne Univ., Melbourne; Univ. of Adelaide; and Univ. of Caen, France. The collaboration is based on three main points, in statistics, machine learning and applications:

- clustering and classification (mixture models),

- regression and dimensionality reduction (mixture of regression models and non-parametric techniques) and

- high impact applications (neuroimaging, satellite and radar imaging, and planetology).

Intelligent wireless access for Internet of Things

Investigator: Peng Cheng

This project aims to develop and validate fundamental theories and techniques for a novel intelligent wireless access paradigm to enhance the efficiency in frequency usage. This innovative approach will be one of the critical enablers for massive device access necessary for future wireless network evolution to support the growing Internet-of-Things. It will enable co-working devices to autonomously sense the local radio frequency landscape, determining how to avoid interference, and exploiting opportunities to intelligently and efficiently access the available radio resources. This will lead to enhanced efficiency in radio resource usage. The project will significantly improve the efficiency of current radio resource utilisation and offer solutions to a challenge of national significance.

Applying AI Technologies to Enhance Autism Diagnosis

Investigators: Lianhua Chi (La Trobe University), Haishuai Wang (Harvard Medical School, Fairfield University) and Zongyuan Ge (Monash University)

Supporter: Wei Xiang (La Trobe University)

Autism spectrum disorder (ASD) is a developmental disorder that affects communication and behavior. It’s critically important to diagnose ASD as early in a child’s development as possible. The investigators have been creating an effective ASD prediction model that utilize individuals’ characteristic and behavioural data as well as the facial images. They also provided a web-based preliminary ASD screening tool that greatly reduces the waiting time (only takes a few seconds) to return the prediction based on the uploaded facial image with acceptable accuracy, which helps medical researchers and ASD specialist for a preliminary ASD diagnosis. This study has the power to help parents quickly and accurately identify children at potential risk for autism, which led to new insights about ASD and enhances how doctors evaluate patients for other behavioral disorders.