Project Description

Atrial fibrillation (AF) is an irregular and often rapid heart rhythm that can lead to blood clots in the heart. A blood clot detached from the heart can travel to the brain and result in an ischemic stroke. Many individuals with AF experience sporadic episodes that may go unnoticed during routine or even prolonged monitoring periods. This intermittent nature complicates detection, as traditional diagnostic methods like electrocardiograms (ECGs) or Holter monitors may fail to capture these transient episodes. Additionally, AF can appear with subtle or no symptoms, further delaying diagnosis. The lack of observable symptoms also makes it challenging for patients to recognize the need for medical evaluation and for healthcare providers to initiate timely diagnostic procedures. Furthermore, the diagnostic tools available for AF detection, such as prolonged ECG monitoring devices and implantable cardiac monitors, can be costly, invasive, and require significant patient compliance. These factors contribute to the challenges in achieving widespread and efficient AF screening, especially in populations with limited access to healthcare resources. These challenges underscore the need for an innovative, cost-effective approach to improve AF detection and management. Magnetic resonance imaging (MRI)—particularly diffusion-weighted imaging (DWI)—has become routine soon after a stroke. As a result, an automated MRI-based risk stratification framework could be pivotal for the early identification of AF patients who would benefit most from prolonged cardiac monitoring

Key Outcomes

• A novel deep learning framework to differentiate AF from other stroke etiologies in post-stroke brain MRI (DWI)
• An innovative approach to integrate AF diagnosis into routine stroke imaging protocols, which can aid in more accurate treatment and management strategies.
• A multi-task learning approach that combines segmentation and classification tasks, improving both model performance and interpretability

Collaborators

• Mohammad Javad Shokri (University of Melbourne)
• Dr Nandakishor Desai (University of Melbourne)
• Dr Angelos Sharobeam (Royal Melbourne Hospital)
• Prof Bernard Yan (University of Melbourne)
• Prof Marimuthu Palaniswami (University of Melbourne)

Publications

• ACCV 2024 (Accepted)[PDF]
• DICTA 2024 (Best Paper Award)[PDF]

See Publications page for full details.

Project Description

Stroke is one of the leading causes of disability among the elderly population and is a significant public health problem worldwide. The main impact of stroke is functional disabilities due to motor impairment after stroke. Advances in modern medicine and technology have significantly improved diagnosis and treatment; however, most post-stroke care is based on the effectiveness of rehabilitation. Stroke rehabilitation depends on two main components: (i) training (or therapy) to restore the patient to pre-stroke mobility and (ii) assessing motor functionality of affected patients performing activities to track motor recovery.
This project focuses on combining wearable devices and machine learning (ML) produces new pathways for effective stroke rehabilitation. While wearable devices help capture patient movements at much finer time resolutions, ML allows us to build predictive models from wearable data to assist clinicians in diagnosis and treatments. Specifically, the project will focus on how wearable devices and ML can improve monitoring quality in training intervention, assessment, and remote monitoring.
Furthermore, stroke rehabilitation interventions require multiple training sessions and repeated assessments to evaluate the improvements from training. The training and assessment process incurs time, labor, and cost to determine whether the training produces positive outcomes. Predicting the effectiveness of gait training based on baseline minimum foot clearance (MFC) data would be highly beneficial, potentially saving resources, costs, and patient time.

Key Outcomes

• A comprehensive and deeper analysis of various wearable devices and machine-learning techniques to monitor stroke patients’ motor recovery during rehabilitation
• Developed Short-term Fourier Transform (STFT)-based features from minimum foot clearance (MFC) data (outperform wavelet, histogram, and Poincaré-based features) to predict the effectiveness of biofeedback training

Collaborators

• Nandini Sengupta (PhD Student, University of Melbourne)
• Soheil Bajelan (Victoria University)
• Prof Rezaul Begg (Victoria University)
• Prof Catherine M Said (Victoria University, University of Melbourne)
• Prof Bernard Yan (University of Melbourne)
• Prof Marimuthu Palaniswami (University of Melbourne)

Publications

• IEEE Access, 2024. [Open Access]
• Frontiers in Bioengineering and Biotechnolgy, 2024. [Open Access]

See Publications page for full details.

Project Description

Routine assessment of structural conditions (such as defects and cracks) is necessary to ensure structural and operational safety of critical infrastructure. The current practice to detect structural damages such as cracks depends on human visual observation methods, which are prone to efficiency, cost and safety concern. Current mainstream methods of infrastructure assessment involve performing visual inspection periodically to inform management agencies the current stage of infrastructure. Hence maintenance and strengthening works can be carried out timely to assure the operational efficiency and safety of critical infrastructure. For example, the current level-1 and level-2 inspection guidelines heavily rely on visual inspection carried out by qualified inspectors to detect visible cracks on the surface of structures. Manual inspection of large infrastructure such as long-span bridges requires inspectors to enter hazardous areas or inaccessible to physical location limits, which not only affects the reliability and efficiency of the inspection but is also a safety concern for inspector.
Recently vision-based systems appear to be a promising solution for an autonomous inspection system to analyze images and detect defects on structures. The advantages of vision-based methods is that they capture 2D/3D information of the structures. This will enable automated systems to detect superficial defects (cracks, corrosion) as well as add comprehensive information about the structures. In addition, vision-based systems provide accurate information compared to manual inspection and crack detection using contact-based sensors

Key Outcomes

• Classification Model: Presented a non-overlapping window-based approach using Convolutional Neural Network to classify concrete cracks from images at finer level (windows size of 64 x 64) on smaller images mainly targeted toward real-time applications.
• Segmentation Model: A novel deep learning framework for detecting cracks and then estimating crack widths in concrete surface images. It uses Attention Recurrent Residual U-Net (Attention R2U-Net) with Random Forest regressor to predict crack width.
• A vision transformer (ViT)-based framework to detect cracks on asphalt and concrete surfaces

Collaborators

• Prof Tuan Ngo (University of Melbourne)
• Prof Tuan Nguyen (University of Melbourne)
• Prof Daniel Dias-da-Costa (University of Sydney)
• Prof Marimuthu Palaniswami (University of Melbourne)

Funding

• CRC-P for Advanced Manufacturing of High Performance Building Envelope, funded by the CRCP program of the Department of Industry, Innovation and Science, Australia
• Australian Academy of Science, Australia (the Asia Pacific Research Network for Resilient and Affordable Housing (APRAH)

Publications

• Automation in Construction, 2022. [PDF]
• Structural Health Monitoring Journal, 2021. [Link]
• Structural Health Monitoring Journal, 2020. [Link]

See Publications page for full details.

Project Description

The World Health Organization (WHO) estimates that worldwide there are about 285 million vision-impaired people as of August 2014. WHO also estimates that around 39 million of the 285 million are blind, with 90% of the vision-impaired are from developing countries. Many people have vision problems due to birth defects, uncorrected errors, work nature, accidents, and aging. The white cane and guide dog are the most widely used means of navigation for vision-impaired needs. However, there is still a lack of devices to detect potholes and uneven pavements, which inhibits mobility after dark.
This pilot study proposes a computer vision based pothole and uneven surface detection approach to assist vision-impaired people in meeting their mobility needs. Our system uses laser patterns, monocular video and analyzes patterns for providing path cues. Our system provides over 90% accuracy in detecting potholes for assisting in real-time navigation.

Key Outcomes

• A new computer vision system using laser projects and camera to record an analyse patters for detecting potholes
• Detects non-protruding hazards (suchas potholes and drop-offs)
• Portable system that processes and outputs hazards in real-time

Collaborators

• Prof Elaine Wong (University of Melbourne)
• Dr Javavardhana Gubbi (University of Melbourne; current: Tata Concultancy Services Research)
• Prof Marimuthu Palaniswami (University of Melbourne)

Partners

• Melbourne Networked Society Institute (MNSI)
• Vision Australia
• Guide Dogs Australia

Funding

• Disability Research Initiative (DRI), University of Melbourne
• Ian Potter Foundation - Health and Disability Grants

Publications

• IEEE International Conference on Communications (ICC), 2016 [PDF]
• IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), 2016 [PDF]

See Publications page for full details.

Project Description

Crowd behavior analysis is an important problem in video surveillance applications. Often times, security personnel manually scan each video feed from several sources to monitor safety and security of general public. However, this is a cumbersome process, can cause fatigue and prone to human errors. Video analytics allows for automatic detection of events of interest, but it faces many challenges because of non-rigid crowd motions and occlusions. The algorithms developed for rigid objects are ineffectual in managing crowds. This study developed optical flow-based video analytics for crowd analysis and applications include people counting, density estimation, event detection, and abnormal event detection.

Key Outcomes

• Detect suspicious people (loitering behavior) from videos. [Loitering Behavior], [Anomalous Behavior]
• Automatically count number of people in crowded videos. [Publisher Link].
• Estimate density of people in densely crowded scenarios. [Estimate Density]
• Track people movements and their behavior. [Tracking People], [Crowd Movement]
• Identify and detect crowd activities (events) [Crowd Event Detection], [Crowd Activities], [Dimensionality Reduction], [Anomalous Event], [Non-linear Dimensionality Reduction]

Collaborators

• Dr Javavardhana Gubbi (University of Melbourne; current: Tata Concultancy Services Research)
• Dr Slaven Marusic (University of Melbourne; current: Digital Insights Leader, Aurecon)
• Prof Marimuthu Palaniswami (University of Melbourne)

Partners

• Melbourne Cricket Club
• ARUP
• SenSen Networks

Funding

• Australian Research Council (ARC) Linkage Project (LP100200430)

Publications

• Structural Health Monitoring Journal, 2022
• Automation in Construction, 2021

See Publications page for full details.

Project Description

Stroke is a major cause of morbidity and mortality in Australia. There is an annual incidence of 48,000 new strokes and the risk of death is 25 to 30%. Of those who survive, stroke contributes to 25% of all chronic disabilities in Australia. Acute stroke is caused by a blockage of one of the arteries in the brain resulting in interrupted blood supply. The National Institute of Health Stroke Scale (NIHSS) is an international initiative to systematically assess stroke and provide a quantitative measure.
This project focuses on monitoring stroke survivors during the "hot" hours which is usually the first 24 hours after the onset of stroke. We propose a wireless accelerometer-based system to monitor the motor recovery in acute-stroke patients.

Key Outcomes

• Designed and demonstrated end-to-end wireless stroke monitoring system to monitor motor recovery
• Proposed novel measures of time (and frequency) domain coherence between accelerometer data from two arms at different lags (and frequency bands).
• Automated classification of wrist-worn accelerometry data into different levels of hemiparesis
• Demonstrated that the quality of upper limb motion can characterize and identify hemiparesis in stroke survivors
• Analyse the severity of lower limb hemiparesis using gait data

Collaborators

• Dr Javavardhana Gubbi (University of Melbourne; current: Tata Concultancy Services Research)
• Dr Shreyasi Datta (University of Melbourne; current: Tata Concultancy Services Research)
• Dr Chandan Karmakar (Deakin University)
• Prof Bernard Yan (University of Melbourne, Royal Melbourne Hospital, Melbourne Brain Centre)
• Prof Marimuthu Palaniswami (University of Melbourne)

Partners

• Eoxys

Funding

• Australian Research Council (ARC) Discovery Project (DP19010248)

Publications

• Physiological Measurement, 2021. [Online]
• IEEE Engineering in Medicine and Biology Society (EMBC), 2021. [PDF]
• IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2020. [PDF]
• Biomedical Engineering Online, 2013. [Open Access]
• IEEE Engineering in Medicine and Biology Society (EMBC), 2013. [PDF]

See Publications page for full details.

Project Description

Water resources are being polluted at alarming rates. In Victoria (Australia), 80% of the waterways are in poor to moderate condition. A biological early-warning system can aid in forestalling the impact of the immediate pollution. A prototype sensor system as a part of the Autonomous Live Animal Response Monitor (ALARM) system, CAPIM, Victoria. This system is designed to measure a suite of biologically relevant physiochemical parameters in freshwater. The system measures temperature, light intensity, pH, electrical conductivity (EC), total dissolved solids (TDS), salinity (SAL), dissolved oxygen (DO) and oxidation reduction potential (ORP). The project focuses on identifying possible pollutants by monitoring physical properties in water streams, improve the health of water-dependent aquatic and terrestrial lives, and provide environmental benefits.

Key Outcomes

• Design and demonstration of a low-cost, continuous water-quality monitoring system
• Low-cost sensors and open-source hardware for water quality monitoring at reduced cost
• Delivers reliable, continuous water physiochemistry data, allowing catchment managers to improve spatial and temporal resolution of water quality surveillance.

Collaborators

• Dr Javavardhana Gubbi (University of Melbourne; current: Tata Concultancy Services Research)
• Mr Steve Marshall (University of Melbourne)
• Prof Marimuthu Palaniswami (University of Melbourne)

Funding

• Centre for Anthropogenic Pollution Impact and Management (CAPIM), Victorian Government.

Publications

• International Conference on Advances in Computing, Communications and Informatics (ICACCI), 2013. [PDF]

See Publications page for full details.

Project Description

Exposure to excessive noise levels is known to negatively impact quality of life. Exposure to excessive noise levels is known to have detrimental health impacts (at sound pressure levels above 65 dBA). Some of these effects include (a) stress, anxiety contributing to mental illness; (b) pain (at 120 dB); hearing damage (at 85dB); (c) sleep disorders, hypertension; heart diseases. A 2007 social survey by Australian state Victoria’s EPA found that almost half of its people (49 per cent) were disturbed or annoyed by environmental noise and one-quarter (24 per cent) of respondents reported sleep disturbance at some stage in the previous 12 months.
A three dimensional urban noise map is an ideal solution. This requires continuous monitoring. Armed with wireless sensor networks and IoT, this pilot project proposes a new architecture and reports development of necessary infrastructure for this objective. The framework also allows collection of temperature, humidity and light in addition to noise intensity levels in urban environments. Furthermore, this project models the conditions of a city-wide distribution of sensors and data collection applications to model in real time the functioning urban sensing elements of a smart city, translating vast amounts of sensor data into meaningful information and ultimately action.

Key Outcomes

• Intrdoduced a novel, scalable multi-tier WSN architecture for continuous monitoring of noise in urban environment
• Designed a new A-weighted filter to capture only the sound levels without retaining any sound data (privacy-preserving)
• Introduced a new 3D noise mapping architecture using hierarchical structure of expensive and low-cost noise sensors

Collaborators

• Dr Javavardhana Gubbi (University of Melbourne; current: Tata Concultancy Services Research)
• Dr Slaven Marusic (University of Melbourne)
• Prof Marimuthu Palaniswami (University of Melbourne)

Partners

• Melbourne City Council
• ARUP

Funding

• Australian Research Council (ARC) Linkage Infrastructure, Equipment and Facilities (LIEF) Project (LE120100129), 2012
• ARC Linkage Project (LP120100529), 2012

Publications

• International Conference on Advances in Computing, Communications and Informatics (ICACCI), 2013. [PDF]

See Publications page for full details.

Project Description

Knowledge of vehicle locations is of utmost importance for taxi services, car rentals, postal/courier services and emergency services that provide services based on locations or that needs to keep track of its vehicle fleet. However, the astounding cost of the system and the lack of ability to incorporate other sensor devices to the same platform restrict the applicability of such system for wide range of applications. This project develops an embedded system to acquire, process and send GPS locations to a centralized server through text messages or Short Message Service (SMS). The platform is scalable to generic sensor networking platform having numerous applications ranging from medical to military and from aerospace to underwater applications.

Key Outcomes

• Design and demonstration of a real-time asset monitoring system
• Integrates with sensor nodes for large-scale dpleoyment
• Updates moving asset locations in real-time via Google Maps.

Collaborators

• Mr Rubin Tellis (Deakin University)
• Dr Davood Izadi (Deakin University)
• Dr Samitha Ekanayake (Deakin University)
• Prof Pubudu Pathirana (Deakin University)

Funding

• Australian Research Council (ARC) Linkage Infrastructure, Equipment and Facilities (LIEF) Project (LE00883073), 2008

Publications

• Intelligent Sensors, Sensor Networks and Information Processing Conference (ISSNIP), 2009. [PDF]

See Publications page for full details.

Project Description

Industry 4.0 is characterised by the integration of advanced technologies and digitalisation, and is reshaping the global business landscape across many industries. To date, the building industry has lagged behind manufacturing industries and has failed to realise the key benefits of Industry 4.0.
Industry 4.0 involves integrating advanced technologies like the Internet of Things (IoT), cloud computing, artificial intelligence (AI), robotics, and automation. These technologies promise to improve sustainability, efficiency, productivity, and project timelines. They can also help enhance safety practices and risk management through real-time monitoring and data analysis.
In contrast to the traditional industry’s data paucity, Industry 4.0 enables data-driven insights, predictive analytics, and collaborative ecosystems that connect stakeholders in the building process. By adopting Industry 4.0 technologies, techniques and processes, the Australian building industry can drive substantial enhancements, empowering both industry professionals, customers, and end-users alike.
This project highlights the transformative implications of Industry 4.0 for the Australian building industry. Implementation of key principles and technologies are summarised along with an action plan and roadmap for the building industry.

Key Outcomes

• Recommendations to government in skills development, vocational training and upskilling strategies
• Technology insights and role of Industry 4.0 for building industry
• Recomentations for industry:
  • Transitioning to Industry 4.0 via collective efforts
  • Embracing future operating models considering the entire life-cycle of assets
  • Digitisation and automation of the design, manufacturing, and assembly stages
  • Establishing building-specific Industry 4.0 processes
  • Embracing relevant technologies such as smart cities, cloud computing, AI, machine learning, and GIS-enabled mapping
  • Leveraging digitisation for holsitic performance enhancement (including digital tools and platforms, digitising supply chains)
  • Integrating data and leevraging AI to gain deeper insights into o project performance, identify trends and patterns, and make more informed decisions
• Recommendations for policy makers to aid in the accelerated adoption of Industry 4.0 in the construction industry

Collaborators

• Dr Philip Christopher (University of Melbourne)
• Dr Duncan Maxwell (Monash University)
• Dr Sahar Soltani Monash University)
• Dr Siddhesh Godbole (University of Melbourne)
• Dr Ali Rashidi (Monash University)
• Dr Tendai Makasi (Queensland University of Technology)
• Dr Francisca Leonard (Queensland University of Technology)
• Dr Priyadarshini Das (Monash University)

Partners

• Bentley Homes
• BlueScope
• Exergenics
• Fleetwood Australia
• Hyne Timber
• Lendlease Digital
• prefabAUS
• Space Platform
• XLam Australia
• Ynomia

Funding

• Building 4.0 CRC

Publications

• Building 4.0 CRC Final Report, 2023. [PDF]

Project Description

The construction supply chain is a critical enabler for the construction industry but also poses challenges and risks. This is mainly due to the typical make-to-order nature of the construction supply chain, which is often unstable, highly fragmented and geographically dispersed. The ability to track and trace, called traceability, is becoming increasingly important as it contributes to building compliance, project efficiency, safety, sustainability and performance.
This study aimed at understanding the state-of-the-art traceability in the construction industry and key stakeholders’ perspectives and recommend future research. The longer-term objective is to demonstrate how sensor networks can be used to provide live streamed data to improve project management and validate building compliance through measures used to guarantee the provenance of the supply chain. Systems to be developed will have the further capability of integration with the building digital twin.

Key Outcomes

• Helped to understand of the state-of-the-art of traceability technology solutions and their current usage, development and challenges in the construction industry.
• Used a mixture of research methods (e.g., interviews, literature review and case studies) to assess existing and emerging tracking technologies (e.g., sensors, visual tracking, information systems, data collection) for sectoral and issue appropriateness.
• Future Research:
• Roadmap for sector-wide transformation
• Digitalisation traceability solution development
• Pilot study and living lab
• Education and training.

Collaborators

• Dr Wen Li (University of Melbourne)
• Dr Guilherme Luz Tortorella (University of Melbourne)
• Prof Robin Drogemuller (Queensland University of Technology)
• A/Prof Joseph Liu (Monash University)
• Dr Yihai Fang (Monash University)
• A/Prof Tim Rose (Queensland University of Technology)
• Dr Sara Omrani (Queensland University of Technology)
• Prof Alistair Barros (Queensland University of Technology)
• Prof Tuan Ngo (University of Melbourne)
• Mr Declan Cox (Monash University)
• Ms Negar Adebi (Queensland University of Technology)
• Ms Noor E Karishma Shaik (University of Melbourne)

Partners

• BlueScope Steel Limited
• Holmesglen Institute
• Lendlease Digital Australia Pty Limited
• Salesforce.com, Inc.
• Sumitomo Forestry Australia Pty Ltd
• The Master Builders Association of Victoria
• Victorian Building Authority
• Ynomia Pty Ltd

Funding

• Building 4.0 CRC
• Advanced Manufacturing Growth Centre Ltd (AMGC)

Publications

• Building 4.0 CRC Final Report, 2022. [PDF].

Project Description

The lack of accurate and timely information from building sites makes project management difficult. Often, data and information is collected manually, which can be sporadic, costly and sometimes biased. This makes it difficult to track project progress and deliver projects on time and within budget. Advances in sensing technologies make it practical and economically viable to monitor construction objects and activities automatically and in real time. The evolution of data to information and knowledge increases data’s value and reduces the risks of making decisions based on incomplete data.
The project aims to understand the opportunities and obstacles associated with using new sensing technologies to support project management and decision making on building sites. The main objectives were:

• Understand how passive data collection can improve the management of on-site activities.
• Analyse state-of-the-art sensing and analytics technologies.
• Assess and validate key assumptions underlying an implementation roadmap in the field.

Key Outcomes

Following recommendations were identified based on a literature review, market analysis and an onsite pilot using tracking technology to characterise and optimise waste disposal processes:

• Many technologies can track progress on building sites, but often they do not capture design changes. Tracking design changes and progress could provide more accurate data to guide future projects.
• Technology must support work practices and be appropriate for businesses of all sizes (including sole traders). User interface design is very important.
• It may be worth offering subcontractors incentives to use technologies.
• Technologies proposed for projects and systems trialled should have clear and communicated pathways for future use.

Collaborators

• Dr Yihai Fang (Monash University)
• A/Prof Mehrdad Arashpour (Monash University)
• Dr Robert Moehler (University of Melbourne)
• Dr Duncan Maxwell (Monash University)
• Dr Ivana Kuzmanovska (Monash University)
• Dr Ali Rashidi (Monash University)
• A/Prof Kourosh Khoshelham (University of Melbourne)
• Prof Tuan Ngo (University of Melbourne)
• Prof Robin Drogemuller (Queensland University of Technology)

Partners

• Lendlease Digital
• Ynomia
• Standards Australia

Funding

• Building 4.0 CRC
• Advanced Manufacturing Growth Centre Ltd (AMGC)

Publications

• Building 4.0 CRC Final Report, 2022. [PDF].