Auckland Bioengineering Institute

Project

Computed tomography (CT) stands out as the predominant imaging modality for evaluating suspected stroke patients, offering widespread availability, swiftness, simplicity, and cost-effectiveness in comparison to Magnetic Resonance Imaging (MRI). CT scans play a pivotal role in delivering crucial insights into the extent of ischemia. Assessing the prognosis of acute ischemic stroke patients relies significantly on stroke lesion volume, a key radiologic measurement. However, the manual segmentation of stroke lesions remains a highly time-consuming task with relatively low inter-rater repeatability.

Role

This project is dedicated to the development of an automated segmentation method, leveraging deep learning techniques to precisely delineate ischemic lesions from multislice CT images of stroke patients. The segmented stroke lesions will subsequently be aligned with the standard brain template/atlas, enabling a quantifiable assessment of the stroke lesion's impact on recovery outcomes.

Ideal student

Ideal candidates will have an interest in image processing, and basic programming skills in Matlab, C++, or Python. Experience in machine/deep learning and/or medical image analysis will be beneficial.

Project

Innovations in wearable technologies are needed to reduce the burden on the healthcare system and improve patient outcomes. Our project aims to advance medical diagnosis and monitoring with innovative wearable and portable devices combined with AI software. This technology will combine multimodality data and offer remote, personalized, and continuous wellness monitoring at an affordable cost.

Role

By utilizing publicly available datasets of physiological sounds (including heart, lung, bowel, ambient, and human sounds), we will develop proprietary AI algorithms for signal denoising, feature extraction, and model training. These algorithms will leverage advanced AI models such as CNNs and RNNs to identify key patterns in temporal signals, optimizing the performance of our diagnostic tools.

The prospective student will gain hands-on experience in data acquisition, AI algorithm development, and model optimization. They will collaborate on creating robust protocols for audio signal processing and feature extraction, contributing to a project that aims to enhance medical diagnostics. This opportunity will provide students with practical skills, technical knowledge, and the chance to work alongside experienced researchers, positioning them at the forefront of advancements in medical technology and AI.

Project aims

The main objective of this project is to develop AI tools to combine multiple organ datasets. The tasks involved, which can be altered based on the student's preference, include:
- Testing signal processing and denoising algorithms on available datasets
- Implementing multi-modality AI tools (CNN/RNN)
- Performing experiments for acoustic data acquisition
- Analysing and presenting data from experiments

Ideal student

Students will join a multidisciplinary team of engineers and clinicians. The project is suitable for students interested in translational research and will involve software development, experiments, and signal processing/data analysis. Prior clinical experience is not required.

Outcome

As part of a long-term plan for diagnostic device development, the outcomes of this project will enable the creation of more efficient and effective strategies for diagnosing and monitoring various health conditions both in clinical settings and at home.

Categories
Computational Science & Engineering
Medical Devices
Signal Processing

Desired skills
Basic knowledge of AI methods and Programming; the ability to work independently and as part of a team.

Project

The heart is a complex organ that relies on a delicate balance of many interacting cellular processes. When a person develops diabetes, this balance can be disturbed, leading to compromised function. To understand the mechanisms driving dysfunction in the diabetic heart cell, we turn to quantitative computational modelling.

Role

By simulating the effects of diabetes on cardiac physiology, we can gain valuable insights into how the disease manifests and progresses, leading to the potential development of novel therapeutic strategies and treatments.

Aim

The overall aim in this project is to develop an integrated model of cardiac excitation-contraction to investigate the mechanisms that are responsible for dysfunction in the diabetic heart.

Skills required
Familiarity with MatLab, Python and cellML
Understanding of cardiac cellular excitation-contraction

Project

The human brain functionality is highly variable across the population. Early-life stimulation and later specialisation play a significant role in shaping humans to adapt to different skills and environmental requirements. This trait of brain specialisation hinders the comparison of brain activity across the population, needing person-specific neurodynamic models to understand how functionality unravels in each of us.

Role

In this project, we propose the development of a workflow to create such models from medical images of brain anatomy (T1-weighted MRI) and function (fMRI).

The student will have access to all the Animus Laboratory resources and the opportunity to participate in group discussions and events to enhance their understanding of computational modelling and neurovascular physiology.

The developed workflow will be applied to the analysis of brain function in healthy and suspected dementia groups, giving the opportunity to explore the direct application of the developed tools in clinical research.

Prerequisites
- Experience in Python or C++ programming;
- Experience in UNIX environments and version control (or interested in learning);
- Good teamwork and communication skills.

Project timeline (12 weeks)
- Familiarisation with (i) anatomical segmentation and fMRI tools (pre-existent software) and (ii) anatomical brain reconstruction (segmentation) - 2 weeks;
- Integration of segmentation (CAT12) and fMRI tools (fMRI prep) - 3 weeks;
- Extract and analyse brain function in healthy and suspected dementia cohorts - 3 weeks;
- Couple brain function with a pre-existent whole brain neurodynamic model - 4 weeks.

Project

Just like electrical activity is important to keep your heart beating, the electrical waves of the gut are important to keep the food digesting and moving. However, the rhythmic gastrointestinal myoelectrical events, also known as slow waves, are not comprehensively understood to make reliable interpretations from EGG recordings. Therefore, many passive recording systems are being applied at present to first understand the features of these waves in acute settings.

Aims

This project aims to develop techniques to evaluate the performance of a number of electrical recording systems and determine their suitability for measuring bioelectrical activity. In this project 4 systems will be evaluated: (i) BioSemi (an established 256 channel commercial system), (ii) Shimmer3 (a portable/wearable 3 channel system), (iii) OpenBCI (an open-source portable, battery operated system) and (iv) Redlab 1208LS (a small USB powered 8 channel system). Specifically, there are 4 major aims:

1. Develop a robust benchtop test setup along with a recording interface hardware using existing flexible electrode arrays.

2. Develop a signal processing framework to analyse the signals acquired from the 4 acquisition systems.

3. Measure signals using the benchtop setup as well as from in vivo animal studies.

4. Quantify and compare the signal quality across the 4 acquisition systems.

Ideal student

Ideal candidates should be able to communicate effectively and work independently. No previous experience in hardware or signal processing will be required, but basic operations of MATLAB will be preferred.

Project

Newborn infants, especially preterm babies, are highly susceptible to brain injury following hypoxia-ischemia (HI), which can lead to lifelong neurodevelopmental disabilities. Despite the critical need, there are currently no specific treatments for neuroprotection or neurorepair in these infants. HI can occur at or before birth, with injury evolving over weeks, necessitating treatments tailored to the injury phases, as evidenced by the effectiveness of therapeutic hypothermia initiated within 3 hours of birth.

Biomarkers are crucial for identifying at-risk babies and determining the injury phase.

Role

Our research focuses on using electroencephalography (EEG) to study brain activity post-HI. This project aims to examine EEG variations in preterm and full-term fetal sheep after HI to identify spectral features affected by different interventions.

Objectives

1. Extracting temporal spectral features from EEG recordings up to 48 hours post-HI from four groups: normothermia, hypothermia, MgSO4 treated, and sham, in both preterm and full-term sheep fetuses.
2. Analyzing differences in these features among the groups.
3. Utilizing machine learning classifiers (preferably CNNs or transformers) to determine if spectral features can distinguish between these groups within the latent phase.

This research will enhance our understanding of HI injury development and contribute to the creation of targeted treatments.

Desired skills

We seek a motivated student with a background in engineering, computer science, or biomedical engineering, with a strong interest in medical data analysis and computational neuroscience, and proficiency in Python or Matlab. Experience in machine/deep learning, and/or medical signal/image analysis is highly beneficial.

Project

For this summer project, you will be involved with developing and improving new experimental approaches for making measurements from uterus muscle. You will be working on constructing and improving design of ex vivo tissue measurement systems. You will be working with advanced signal analysis and motion tracking algorithms to analyze data from tissue experiments to improve our understanding of uterine smooth muscle function.

Project

Echocardiography is the mainstay for non-invasive routine clinical assessment of the heart. While 3D echocardiography is increasingly available, 2D images make up the large majority of all echocardiographic examinations.

Aim

The aim of this project is to apply deep learning to generate 3D geometric models of the heart from standard 2D echocardiograms to provide clinicians with more accurate volume estimates, novel kinematic and energetic indices, and to enable further applications, such as cardiac electro-mechanical modelling and analyses.

As part of a NZ Government funded programme on clinical translation of heart modelling, outputs of this project will contribute to more efficient and effective strategies for the diagnosis and monitoring of heart disease at Auckland City Hospital.

Ideal student

You will join a multi-disciplinary team of bioengineers, clinicians, and imaging specialists. This project will suit students with an interest in deep learning, medical image analysis, and translational research. Prior medical imaging experience is not required.

Project

Animals can move in a tremendous number of different ways. Animals with skeletons are limited by those skeletons; skeletons (both endo-, like humans, and exo-, like insects) can be damaged by sudden forces, and enforce size and shape limits on where the animals can go.

Other animals, however, can achieve strength and dexterity without skeletons, overcoming these limits: an octopus can open a jar, take apart a clam, and then hide in a crevice nearly as small as its beak! To date, robots have been designed with skeletons, but can we design them to be soft like an octopus?

Role

In this project, you will work on a new idea for soft robot actuation using powerful magnets, liquid metals, and electrical currents, a technique called magnetohydrodynamics. Specifically, you will investigate the design of magnetohydrodymanic (MHD) pumps, and how they can be used to drive soft robots.

Ideal student

We are looking for students with experience in engineering science, biomedical engineering, mechanical engineering, mechatronics engineering, and/or electrical engineering, but could also accommodate students with a hands-on chemistry or physics background.

Tasks

The specific project tasks will be tailored to student experience, but some of the possible research questions are as follows:

– What can we learn from animal anatomy to help design systems powered by MHD pumps? For instance, squid and octopuses use fluid pressure to move, as do spiders. Can we use their anatomy to inspire robot designs?
– How can we model these pumps, when soft and/or driven with power pulses? What can such models tell us about how to optimize their designs?
– What happens when we drive MHD pumps with pulsed power? How much current can we deliver to a small pump, and does this result in the expected behaviour?
– How can we coat the materials used in MHD pumps to improve pump performance? Can electroplated coatings help reduce resistance?

Project

Breast cancer affects 1 in 9 women in Aotearoa New Zealand. A key challenge for clinicians is determining where tumours move as the breast deforms due to changes in patient positioning during treatment procedures, such as surgery or radiotherapy.

Tracking the deformations of the patient's skin surface provide useful information to biomechanical models that predict the position of tumours within the breast during repositioning. Cutting edge machine learning motion tracking methodologies, such as 'Track everything everywhere all at once' (as illustrated) are capable of achieving high tracking performance on traditional tasks such as tracking images of buildings, cars, or crowds. However, the performance of these techniques has yet to be evaluated on human skin.

Aims

This project aims to compare the performance of these cutting edge methods against traditional methods of feature tracking. To this end, the student will compare the performances of the methods using breast imaging datasets collected by our research group.

Ideal student

You will join a multi-disciplinary team of bioengineers, clinicians, computer vision experts and instrumentation specialists. This project will suit students with an interest in computer vision, machine learning and computer programming.

Project

Cardiovascular disease (CVD) affects millions worldwide and is the leading cause of mortality. Devices that can efficiently and non-invasively provide early and clinically useful diagnostic information may improve patient quality of life and help reduce CVD morbidity. Our group has created a camera-based imaging system that estimates the carotid artery and jugular venous pressure waveform by measuring the deformation of the skin on the neck due to the vessel pulsations. The results to date highlight the potential for this system to be used as a non-invasive diagnostic tool for cardiovascular disease.

Role

This summer studentship project will focus on key developments for the current system. The project will be based on evaluating the techniques with mobile high-speed video recording capabilities including spatial video and stereoscopic imaging. These investigations will be conducted alongside quantitative extraction of blood pressure-related biomarkers. Overall, these advancements will enable the translation of this research toward a home-based healthcare tool.

Project aims
The main objectives of this project will be to:
- Implement and acquire videos from a mobile phone
- Perform experiments for video and blood pressure (BP) acquisition
- Analysis of 3D video/deformation and identify BP imaging biomarkers
- Analyse and present data from experiments.

Ideal student

Students will join a multi-disciplinary team of engineers and clinicians. The project will suit students interested in translational research and will involve hardware development with cameras and optics, bioinstrumentation, imaging experiments, and signal processing/data analysis. Prior medical imaging experience is not required.

Outcome

This project will enable the development of more efficient and effective strategies for the diagnosis and monitoring of cardiovascular diseases cardiac patients in the clinical setting and at home.

Categories
Computational Science and Engineering
Medical Devices
Medical Imaging
Signal Processing

Project

The lymphatic system includes a complex network of vessels and lymph nodes that transports excess fluid from the body's tissues back into the bloodstream. Unfortunately, up to 20% of patients treated for cancer will develop lymphoedema, an incurable disease, caused by lymphatic system dysfunction. When lymphoedema develops, the lymphatic vessels do not work properly and lymph fluid moves in a retrograde direction back into the skin, causing chronic skin thickening and hardening.

Currently it is impossible to predict if lymphoedema will develop after cancer treatment or the altered route that lymph fluid will take during its onset and progression. To improve our understanding of this complex disease, our team is working to develop a state-of-the-art AI models and computational fluid dynamic models of the lymphatic system.

Role

In this summer project, you will work with our team to analyse images of the lymphatic system, to help inform our computational models. This may include fluorescence imaging using our team's novel imaging device, optical coherence tomography (OCT) imaging and CT imaging data. This will contribute to developing the world's most state-of-the-art models of the lymphatics, which could inform prevention strategies and therapies for lymphoedema.