Sohn Hearts and Minds PhD Scholarships

AI-Driven Biomarkers for AF-Associated Cardiomyopathy, supervisor Dr Dhani Dharmaprani

Atrial fibrillation (AF) is the most common heart rhythm disorder, affecting millions of Australians and significantly increasing the risk of stroke, heart failure, and death. While traditionally viewed as an electrical condition, emerging evidence suggests that AF may actively contribute to structural damage of the heart muscle, leading to AF-induced cardiomyopathy (AF-CM). Cardiomyopathy refers to a disease of the heart muscle that affects its ability to pump blood effectively, often progressing silently until serious symptoms appear. This form of heart disease is difficult to detect early, and current diagnostic tools fail to identify high-risk patients before irreversible damage occurs.

This project addresses this critical gap by developing and clinically validating AI-driven biomarkers capable of identifying early-stage AF-CM. By integrating multimodal data (including ECG, electrophysiological studies, cardiac MRI, and echocardiography) it will deliver interpretable machine learning models for early diagnosis and personalised risk stratification.

This initiative is supported by a proven interdisciplinary team, including cardiologists at Flinders Private and Lyell McEwin Hospitals and AI/ML experts from the Australian Institute for Machine Learning.

Anticipated outcomes will support clinical decision-making, inform diagnostic guidelines, and enable a proactive, precision-medicine approach to AF care, ultimately improving long-term cardiac outcomes and reducing the burden of heart failure.

Clinically Deployable Virtual Heart Technology for Personalising AF Treatment - supervisor Dr Dhani Dharmaprani

This research program addresses atrial fibrillation (AF) - the world’s most common heart rhythm disorder and a leading cause of cardiac hospitalisation. More than half of the current treatments for persistent cases of AF fail, stemming from strategies that do not account for individual patient variability. The proposed project addresses this gap by pioneering real-time, clinical-grade cardiac ‘virtual hearts’ – called digital twins – designed to significantly improve AF outcomes. These digital constructs are tailored to mirror individual cardiac behaviours and hold great potential for transformative clinical applications.

However, current cardiac digital twins face key limitations: they rely heavily on pre-intervention data and demand intensive computation, which limits their clinical viability. This project will overcome these barriers by integrating award-winning clinical and engineering research to connect real-time bedside measurements of patient-specific AF behaviour with the precision of cardiac digital twins for the first time. This represents a cutting-edge fusion of advanced digital technologies with detailed electrophysiological evaluation, offering a revolutionary approach to AF treatment.

Findings from this project can help drive a future paradigm shift in cardiology, from traditional trial-and-error methods to personalised digital tools. This project is supported by an established interdisciplinary team, including cardiologists at Flinders Private and Lyell McEwin Hospitals, cardiac digital twin leaders at Imperial College and Queen Mary University of London, and AI/ML experts at the Australian Institute for Machine Learning. These longstanding collaborations ensure clinical relevance, technical rigour, and rapid translation from development to real-world impact. The resulting platform will be also primed for human clinical trials.

A novel device to improve health outcomes with pressure support ventilation in intensive care, in motor neurone disease and in other disorders of inadequate breathing during sleep - supervisor Dr Phuc Nguyen

Many people with a critical illness, or with chronic conditions such as motor neurone disease (MND), need pressure support ventilation (PSV) to adequately breathe, particularly during sleep. Although PSV can be lifesaving and life-extending, asynchronous or mistimed breaths are a frequent problem in many patients. Patient-ventilator asynchrony reflects PSV machine triggering increased (inspiratory) pressure and/or cycling to reduced (expiratory) pressure too early or too late relative to patient inspiration. Asynchrony contributes to potentially injurious lung stretch, sub-optimal ventilation, and patient discomfort, intolerance, and reduced long-term usage of PSV.

Through a long-standing collaboration between the Flinders Health and Medical Research Institute (FHMRI) Sleep Health team and Medical Device Research Institute (MDRI) and Medical Device Partnering Program (MDPP), we have developed novel breath-by-breath analytical tracking methods and two medical device prototypes to implement them at the patient bedside. One device uses an oesophageal balloon catheter and mask airflow sensor to directly assess inspiratory effort timing and volume breath-by-breath. A newer device uses a mattress sensor and bedside acoustic sensor specifically designed to translate these new analytical methods into a new non-intrusive approach.

This project will extend this grant-funded work to further develop and test these novel technologies in clinical practice towards evidence-based, commercialisable and clinically deployable future methods for improved respiratory tracking and treatment outcomes in patients requiring PSV. Tests will include direct breath-by-breath measures from intrusive compared to indirect non-intrusive device methods at the bedside in patients requiring PSV in intensive care. Other tests will involve working closely with patients with MND, and their carers and clinical care teams to further test and implement non-intrusive device methods to evaluate when PSV may be required, and how new breath-by-breath technologies can be used to improve PSV setup and timing to reduce asynchrony to improve PSV tolerance, outcomes and usage

Novel photosensitiser-enhanced nature-derived wound dressings for managing Epidermolysis Bullosa, supervisor Professor Youhong Tang

Dermatological disorders represent a huge burden globally within the context of science and healthcare. Among these, Epidermolysis bullosa (EB) is an heterogenous group of inherited mechanobullous conditions characterized by skin fragility disorders. One of these significant complications of EB is the development of chronic wounds that are associated with infections where its susceptibility to different bacteria colonization led to substantial morbidity. Therefore, there is an urgent need for curative strategies to replace existing palliative care interventions. Current dressings for EB applications have shown efficacy in infection control although still face some limitations. The high frequency of dressing changes, potential cytotoxic effects, allergenicity, and traumatic adherence for fragile skin patients are notable disadvantages reported.

In light of these challenges, this project will emphasize the advantages of nature-derived polymers for wound care due to their chemical structures, biocompatibility, low cost, and their secondary effects minimization when referring healing. For example, the amphiphilic structure of casein, a milk-derived protein, can be a promising candidate for wound dressings. Casein has demonstrated anti-inflammatory effects by reducing key pro-inflammatory cytokines such as TNF-α and TGF-β, thereby supporting the immune and inflammation response, offering cell adhesion properties and the ability to support cell proliferation for the EB demanding obstacles.

Meanwhile, aggregation-induced emission photosensitiser (AIE PS) in conjunction with various techniques have been implemented for antibacterial coatings and skin wounds healing. One of the essential points regarding AIE PS and wound care is their low toxicity and the ability to generate reactive oxygen species (ROS) under visible light, characterized for being high-efficient in the aggregated state. The integration of AIE PS coating layer on the outsurface of wound dressing, such as casein/PVA nanofibrous dressings is a promising alternative for the wound management and bacterial prevention for EB patients, characterized by chronic wounds and high risk of infection.

Advancing signal processing to discriminate motor unit activity non-invasively from respiratory muscles, supervisor Dr Anna Hudson

 Respiratory diseases are common, ∼16% of the total disease burden worldwide in 2021, and are unlikely to decline given our ageing population, climate change, increased risk of global pandemics, and a new generation of nicotine addiction (electronic cigarettes/vapes). We need accurate, safe and practical ways to assess respiratory muscle activity for research and clinical care.

The gold standard method to measure respiratory muscle electromyographic activity (EMG) to assess neural respiratory control is single motor unit recordings with intramuscular (e.g. needle) electrodes. However, the risks associated with needle electrodes near the lung include pneumothorax. As such, our understanding of respiratory pathophysiology has stagnated, and we need non-invasive methods to measure respiratory muscle single motor units, especially in vulnerable populations (e.g. chronic lung disease and spinal cord injury). Worldwide, high-density EMG electrodes and blind source separation algorithms are used to record single motor unit activity from non-respiratory skeletal muscles, typically during isometric (i.e. static) contractions. However, breathing is rhythmic, and current algorithms cannot decompose single motor units using high-density recordings from respiratory muscles. We need to advance signal processing to discriminate single motor units using these non-invasive electrodes for respiratory muscles during breathing. This will allow safe and accurate measures of neural respiratory control in health and disease without the risk of serious complications. This project will advance our understanding and clinical care of respiratory health and impact the field of motor control more broadly.

This project will develop and test signal processing algorithms to discriminate single motor units from high-density EMG electrodes in respiratory muscles in health. This knowledge will be applied to examine fundamental physiology in health and pathophysiology in chronic lung disease, currently impossible with invasive needle electrodes.

Informing decision making in the development of orthopaedic implants, Associate Professor Egon Perilli

In the design of orthopaedic implants, additive manufacturing (for example, selective laser melting) is an increasingly used process. However, defects can arise in the structure (e.g., pores or unmelted powder in lattice structures), which may impact the mechanical behaviour of the implant. X-ray micro-computed tomography imaging can be employed to capture and quantify defects in the structures under investigation. In our preliminary studies we have identified pores and unmelted powder in 3D printed titanium structures, which may impact both the mechanical behaviour as well as the longevity of the implant once implanted.

There has been a shift towards cementless orthopaedic implants in recent years, for which additive manufacturing is being increasingly used. The success of cementless long term fixation critically depends on primary fixation and micromotion, the relative movement between the implant and the surrounding bone. In the development of a new implant, both the built quality (as flawless as possible after manufacturing) as well as the biomechanical behaviour of the bone-implant construct once implanted, have to be considered. In a recent study, we were able to quantify the interference fit and the mechanical environment of tibial knee implants post implantation. The developed technique can be adapted to implant types from various manufacturers, including to prototypes, aiding in the decision making process during implant development.

This project will expand on such techniques to 1) assess the manufacturing defects of prototype implants and 2) quantify experimentally the mechanical environment of the host bone once it is inserted with such an implant, to obtain a deeper understanding of implant primary stability. The results of this project will guide the decision making during orthopaedic implant development, with the aim to improve implant manufacturing and implant stability. The results can further be used to validate computer simulations (finite element models) of implant biomechanical behaviour.

The Virtual Human Hand, Dr Rami Al Dirini

The human hand enables exceptional dexterity through a complex combination of joint movements, including flexion/extension, radial/ulnar deviation, and rotational motions. These coordinated actions underpin our ability to perform essential daily tasks that directly affect our wellbeing. Biomechanically, the hand and wrist are intricate systems comprising multiple bones, ligaments, and articulating surfaces. Despite their importance, our understanding of how anatomical structures (e.g., bone shape and ligament orientation) and mechanical properties (e.g., ligament stiffness or the effects of injury) contribute to dexterity and stability remains limited.

This PhD project aims to develop a high-fidelity computational model of the human hand and wrist, capable of capturing anatomical and biomechanical variation across a healthy population. Unlike previous models that have relied heavily on quasi-static cadaveric data—limited by the absence of live muscle activity and often driven by artificial mechanisms—this research leverages modern imaging techniques such as 4D CT and dynamic MRI. These modalities allow for detailed, non-invasive assessment of in vivo wrist and hand biomechanics during real movement.

Advanced modelling approaches, including statistical shape modelling and finite element analysis, will be used to analyse the data and explore how variations in anatomy and mechanics influence functional outcomes. This research will provide deeper insights into the mechanisms that enable hand dexterity and joint stability, laying the groundwork for improved clinical interventions, surgical planning, and rehabilitation strategies. Furthermore, the resulting model holds significant potential to inform next-generation prosthetic design and robotic systems, bridging the gap between human biomechanics and technological innovation.

Improving Pedicle Screw Placement in Spinal Surgery, Professor Karen Reynolds

Back pain affects 619 million people and is the leading cause of disability worldwide. When conservative treatments are not viable, surgical intervention with pedicle screws is necessary. Pedicle screws are metal screws placed within the pedicles of the vertebrae to provide fixation and stability to the spine. 

While pedicle screws show good biomechanical performance, the accuracy of their insertion remains a concern, with misplacement rates as high as 42%. In addition to compromising the stability of the construct, a breached screw can damage the spinal nerves or surrounding tissues, leading to pain, numbness, or weakness. Breaches can also damage blood vessels, leading to bleeding or other complications.

Emerging acoustical and torsional surgical measurement techniques show promise for rapid detection of dangerous breaches and have potential to enhance screw insertion safety for future patients. This study aims to build on the work undertaken at the Medical Device Research Institute to characterise screw insertion (breach and non-breach) using acoustic emission when drilling and torque when inserting screws.

In our previous work, we have demonstrated that acoustical and torsional responses are significantly different in breached and non-breached screw insertion scenarios. Realtime monitoring of these signals could aid in breach avoidance and improve pedicle screw placement accuracy.

The proposed PhD student project will undertake further experimental work to characterise screw insertion and develop algorithms able to identify when pedicle screws are being inserted incorrectly and are likely to breach.

Listen to Your Heart—in-ear cardiovascular monitoring for adults with hearing loss, Professor Mridula Sharma

This project aims to develop a cardiovascular monitoring technology embedded within hearing aids to improve awareness and management of high blood pressure (BP) in adults with hearing loss.

High BP, also known as hypertension, kills more Australians than any other disease, yet over half of those with high BP are either unaware of it or do not have it under control even with treatment. Older adults (≥ 65 years) are particularly at risk of hypertension, putting them at higher risk of heart attack and stroke. Regular BP monitoring is thus important for detection and ongoing management of hypertension. However, conventional BP monitoring can be tedious, inconvenient and uncomfortable for individuals. Wearable BP monitoring technology holds great potential to overcome barriers of conventional BP monitoring given the ease of use and inconspicuous capability to monitor BP continually. Besides BP, other cardiovascular parameters can also be monitored with this technology. Although this form of wearable technology has mostly been integrated into watches or rings, it can also be embedded into devices worn in the ear, a location which provides more stable signals for derivation of BP and other cardiovascular parameters (e.g. heart rate).

Given one in three Australian adults aged ≥ 65 years have hearing loss, by embedding BP monitoring and other cardiovascular monitoring into hearing aids, older adults with hearing loss can benefit from both treatment of hearing loss and monitoring of vital health parameters with the one device. This allows for increased self-awareness of high BP and other cardiovascular risk factors in already at-risk adults, which older adults may not otherwise engage with.

This project will be supported by industry partners specialising in in-ear monitoring technology and cardiovascular monitoring, ensuring scalability, implementation and commercialisation.