Engineering Science

Applications are now closed

Breast MRI registration using biomechanical models


Supervisors

Prof. Martyn Nash
Prof. Poul Nielsen

Discipline

Engineering Science

Project code: ENG062

Breast cancer affects one in nine New Zealand women. Early detection is key to improving the likelihood of survival. We are developing and implementing an automated biomechanical model-based image processing pipeline at Auckland City Hospital to assist clinicians address challenges in detecting and treating breast cancer. This pipeline automatically builds computational models of the breast from diagnostic clinical magnetic resonance images (MRI) acquired in the prone position, and uses these models to predict what an MRI of the breast would look like in the supine position where treatment procedures such as surgery are performed (shown in Figure 1). Before this pipeline can be routinely applied in clinical practice, we need to ensure that the pipeline accurately predicts the deformation of the breast between these two positions. An approach to validating the accuracy of the model predictions involves mapping (or registering) the model predicted supine MRI with a real supine MRI acquired from the same volunteer. 

Figure 1. Procedure for predicting a supine MRI from a prone MRI using a biomechanical model.
Figure 1. Procedure for predicting a supine MRI from a prone MRI using a biomechanical model.

Main aims of the project

  1. develop a biomechanics based framework for registering prone and supine MRI and;
  2. use this framework to validate predictions of breast biomechanical models.
  3. Identify regions within the breast where the model predictions should be improved.

This study will use our unique imaging database where both prone and supine MRI are available for over 100 volunteers. The ultimate goal of this research is to develop reliable computational models that are suitable for use by our clinical collaborators to accurately predict breast deformations that occur during clinical procedures such as imaging and surgery.

The student would ideally be keen on computer modelling and will develop skills in finite element modelling image processing and registration, and code development using python and Matlab.

Project goals

  1. Use existing tools to automatically segment and build FE models from MR images of the torso in the prone position from a population of volunteers.
  2. For each volunteer:
    • use computational biomechanics to deform the prone breast FE model to predict the breast shape in the supine position;
    • use this deformation model to warp the prone MR image to predict a supine MR image
    • register the predicted supine MRI and the individual’s real supine MRI
  3. Assess the accuracy of the model predicted supine MR by analysing the registration results

How stiff is the breast?


Supervisor

Prof. Martyn Nash
Prof. Poul Nielsen

Discipline

Engineering Science

Project code: ENG063

Breast cancer affects 1 in 9 NZ women. Early detection is key to improving the likelihood of survival. We are developing and implementing an automated model-based image processing pipeline at Auckland City Hospital to assist clinicians address challenges in detecting and treating breast cancer. This pipeline automatically builds computational models of the breast from clinical magnetic resonance images (MRI) acquired in the prone position, and uses these models to predict where tumours would move to in the supine position, in which treatment procedures are performed. This is helpful during clinical diagnostic investigations, because supine images are generally not available.

breast MRI

The main aim of this project is to implement a novel statistical approach to estimate the mechanical properties of the breast. This will involve using existing software tools to identify the mechanical properties for a cohort of cases in our breast image database (which includes both prone and supine MRI from healthy participants). Once this database of mechanical properties has been generated, a statistical model based on partial least-squares regression (PLSR) will be used to relate the stiffness values identified from each participant to the specific shape of their breasts. The statistical model will then be used to predict the tissue stiffnesses, for any new subject, directly from the parameters of the personalised prone model shape.

A challenging extension is to correlate the fat/fibroglandular tissue ratio with the stiffness of the breast tissues to strengthen the predictions from the PLSR model. This would involve applying novel machine learning algorithms developed in our group to automatically segment fibroglandular tissue from breast MR images and incorporate this additional information into the PLSR analysis.

The student would ideally be keen on computer modelling and will develop skills in finite element modelling, image processing, nonlinear parameter optimisation, statistical regression, machine learning, and code development using python and Matlab.

Project aims

  1. Identify breast tissue stiffness parameters for a cohort of healthy subjects using nonlinear optimisation.
  2. Construct a PSLR model to relate the tissue stiffnesses identified for each subject to the prone breast shape parameters.
  3. Validate the the PLSR model using a leave-one-out statistical analysis to determine whether tissue stiffness can be accurately identified given only the shape of the prone breast.

Extension

  • Apply/extend our existing machine learning algorithms to automatically segment fibroglandular tissue from MRI images.
  • Relate tissue stiffnesses to the ratio of adipose/fibroglandular tissue in the PLSR model.
  • Quantify improvement in the predictive power of the PLSR model.

In a state of unstress


Supervisors

Poul Nielsen

Martyn Nash

Amir HajiRassouliha

Thiranja Prasad Babarenda Gamage

Discipline

Engineering Science

Project code: ENG064

In a state of unstress

Soft materials, such as elastomers, gels, and tissues, deform under the influence of gravity. Because of this, it is very difficult (on earth) to determine the reference stress-free geometry. One approach to address this problem in incompressible materials is to counter the distorting effects of gravity by imaging the material immersed in fluid of the same density. In this case, the body forces will be cancelled by the pressure field of the surrounding fluid, so the phantom should be in its unloaded, deviatoric-stress-free configuration.

This project will calibrate and test an existing multicamera stereoscopic system to measure breast shape underwater. An incompressible silicone phantom will be constructed with a known geometry and well-defined deviatoric-stress-free state. Tests will be performed using the silicone phantom stereo-imaged both underwater and in air while subjected to gravity loading from a variety of directions. The measured shapes will be compared with those predicted by a finite element model of the phantom. This project will suit someone interested in imaging and 3D geometric modeling.

A challenging extension to this project would involve the construction of a silicone phantom that has no stress-free state (by, for instance, setting a layer of silicone gel over an existing pre-tensioned membrane). Here, again, tests will be performed using the silicone phantom stereo-imaged both underwater and in air while subjected to gravity loading from a variety of directions. The measured shapes will be compared with those predicted by a finite element model of the phantom, with pre-tension as an unknown parameter. We would like to see if the pre-tension can then be estimated by finding the value that best matches the measured and predicted shapes.

Aims

  • Calibrate cameras and stereoscopic system, both underwater and in air, to quantify the 3D measurement accuracy of the device;
  • Measure the geometry of a soft silicone phantom, both underwater and in air while subjected to gravity loading from a variety of directions.
  • Compare the phantom geometries measured under different loading conditions with those predicted by a finite element model.

Extension

  • Measure the geometry of a soft silicone phantom with no stress-free state, both underwater and in air while subjected to gravity loading from a variety of directions.
  • Compare the phantom geometries measured under different loading conditions with those predicted by a finite element model. Hence estimate that degree of pre-tension of the phantom.

Stretching light


Supervisor

Poul Nielsen

Andrew Taberner

David Budgett

Discipline

Engineering Science

Project code: ENG065

Stretching light

Most stretch/displacement sensors are electronic, stiff, bulky, expensive, and/or have limited range. These limitations pose a significant barrier to integrating stretch sensors into clothing or creating cheap disposable single use devices. We have developed a new type of stretch sensor, based on photonic technology, that is very compliant, scalable, cheap, and operates over a very wide range of lengths and speeds. The technology promises to dramatically overcome traditional limitations, enabling many new applications.

This project will characterise the properties (accuracy, repeatability, reproducibility, hysteresis), and investigate the capabilities, of this new class of sensor. There is an opportunity to significantly expand the scope of the project to look at new configurations that can measure, for example, pressure distributions over a very flexible disposable membrane.

This project would suit a student who is interested in instrumentation, and understanding and applying the physical properties of materials to create fundamentally new sensors.

Aims

  • Use an existing testbed to measure the mechanical and optical properties of existing elastomeric materials.
  • Experiment with the mechanical and optical properties of modified elastomeric materials to optimise the sensitivity and signal to noise ratio for length sensing.
  • Create a range of stretch sensors optimised for a selection of materials tested in the second aim.
  • Design, construct, and test a pressure sensor based on this new technology.

Travelling around NZ in an electric car


Supervisor

Andrea Raith
Ext 81977

Discipline

Engineering Science

Project code: ENG066

Map of new zealand with charging stations

One major concern with electric vehicles is that people suffer from what is called range anxiety, the fear of running out of energy before reaching the destination. On the one hand, drivers underestimate how far an electric vehicle with a full battery can drive. On the other hand, there are only about 300 public charging stations in New Zealand (see plugshare).

In daily life, more and more electric vehicles are used, especially for short trips within cities. Looking at websites of rental car companies, only few electric vehicles are offered. We want to find out if electric vehicles could be used by tourists for their journey around New Zealand.

Tourists want to see the best sights in New Zealand which they want to visit exactly once during their trip. In Operations Research, this problem is known as the ‘Travelling Salesperson Problem (TSP)’. In the classical formulation, the goal is to minimise the travel distance (or time). This problem gets more difficult if cars must be recharged. We will investigate the following questions:

  • Would it possible and reasonable to do a round trip to the sights of New Zealand with an electric car? To answer this question a routing model with charging will be developed
  • Would it make economic sense for customers to hire electric cars, and thus for a rental company to offer a fleet of electric rental cars?
  • Where would charging stations have to be placed to make NZ more accessible?

Note that this project is suitable for an Engineering Science OR student who is confident in (python) programming as this is required in all aspects of the project (in modelling and solving the vehicle range problem, working with Open Street Map data, and visualisation of results).

Towards automated knowledge-based planning for radiotherapy treatment


Supervisor

Andrea Raith
Ext 81977

Discipline

Engineering Science

Project code: ENG067

scan of radiation to the tumour

Radiotherapy treatment (RT) for cancer patients involves delivering a therapeutic dose of radiation to the tumour while avoiding damage to the nearby healthy structures. To manage these conflicting trade-offs, the planning process requires iterative adjustment of the planning parameters which can be time consuming. Furthermore, as the effect of tuning cannot be known a priori, it is difficult to know if a plan considered satisfactory can be further improved.

In this project we will investigate if the planning process can be automated using a reference set of existing treatment plans while guaranteeing the plan quality. Data envelopment analysis (DEA) will be used to identify target reference plans, taking anatomical geometric variations into account, and the planning parameters of the target plans will be incorporated in the planning of the current plan.

In this summer project we will work on clinical plan data currently being collected to establish the relationship between planning parameters and plan quality by applying data analysis and machine learning techniques. Ultimately, knowledge-guided optimisation techniques can be used to find a patient-specific (locally) optimal set of parameter for the planning problem.

Skills required (or strong motivation to learn): python, statistics/data analysis/machine learning.

Backyard Synthetic Aperture Radar (SAR)


Supervisor

John Cater
Andrew Austin

Discipline

Engineering Science
Electrical Engineering

Project code: ENG068

Our team are developing lightweight radar systems for small satellites. This is a design and build project to create a small test system that can be used to image small objects. The work will involve hardware and software integration and the development of a moving phased array.

CubeSat Project Manager


Supervisor

John Cater

Nicholas Rattenbury

Discipline

Engineering Science
Physics

Project code: ENG069

New Zealand is entering the space age, and the University of Auckland is going to be a part of it! This project will involve project management and planning for the assembly of a small CubeSat to be launched by RocketLab in late 2018. We are looking for an enthusiastic new member with an interest in space systems to join our team.

Origami in Space


Supervisors

John Cater

Nicholas Rattenbury

Discipline

Engineering Science
Physics

Project code: ENG070

One of the challenges of space-based hardware is power consumption, which usually depends on the size and number of deployed solar panels. These panels are usually stowed during launch. We need to design a robust deployment mechanism that maximizes the size of the deployed array. This project involves some creativity and mechanical engineering design and building skills; the aim is to create a prototype mechanism for our satellite platform.

Wave impact on coastal structures


Supervisor

Primary: Mark Battley

Discipline

Engineering Science

Project code: ENG071

The overall goal of this work is to develop experimental and numerical methods to investigate the effect of wave impacts on coastal structures and landscapes. This topic is of increasing global importance due to climate change induced sea level rises and extreme weather events, and is also very important to communities exposed to under-sea seismic risks.

Specific aims of this project

  • Implement a Laser based Particle Image Velocimetry (PIV) system in an existing lab scale wave impact facility to provide full-field fluid flow velocity measurements
  • Undertake water impact experiments and collect and analyse the PIV data
  • Compare the measured wave profiles and motion to those predicted by classical analytical wave theory, and numerical modelling methods such as Combined Eulerian-Lagrangian Explicit FEA
  • Investigate how the impactor object shape, angle and velocity profile can be systematically varied to achieve the required wave profiles and motion
  • Experimentally and numerically characterise the deformations of a fixed object (such as a vertical cylinder) being impacted by a wave

This project would suit a student who has a good background in computational mechanics and has an interest in developing an understanding of experimental methods and full-field continuum imaging techniques.

Adding Multi-criteria Optimisation to OpenSolver


Supervisor

Andrew Mason

Andrea Raith

Discipline

Engineering Science

Project code: ENG097

This project will extend the popular open source optimiser, OpenSolver, so that users can model and solve optimisation problems with more than one objective using Excel. Solving these problems requires showing users a set of possible solutions, and letting them easily understand the trade-offs in these solutions. We will need to develop good solution algorithms and new visualisation tools to let the user explore the solution set.

The student needs good skills in programming (preferably with some VBA experience), experience using OpenSolver, and an understanding of linear and integer programming.

A new Simulation and Scenario Manager tool for Excel


Supervisor

Andrew Mason

Discipline

Engineering Science

Project code: ENG098

Through our work on the popular open-source optimiser, OpenSolver, we have developed a good understanding of the needs of Excel users and the technologies available to meet these needs. An important area that OpenSolver does not yet address is simulation in Excel. This project will develop an open-source simulation environment for Excel that is easy to use, tightly integrated with OpenSolver, and includes new visualisation tools to help users visualise and understand their simulation results. This project will also continue work on a scenario manager tool that is currently in beta development

The student needs good skills in programming (preferably with some VBA experience), experience using Excel (and ideally OpenSolver), and a willingness to experiment with and learn new software tools. This work involves collaboration with other open-source developers in this space.

SolverStudio for Google Sheets: A New Online Optimisation Tool


Supervisor

Andrew Mason

Discipline

Engineering Science

Project code: ENG099

SolverStudio is an Excel add-in that allows optimisation models to be built using Excel and modern modelling languages such as the AMPL clone GMPL. This project will build on our experience moving OpenSolver to the online Google Sheets framework to create a new add-in that allows GMPL models to be built and solved using Google Sheets.

The student will need experience programming in JavaScript and developing web sites with client-side code. Experience with the Google API’s would be a bonus. 

Enhancement of SolverStudio for Excel


Supervisor

Andrew Mason

Discipline

Engineering Science

Project code: ENG100

SolverStudio is an Excel add-in that allows optimisation models to be built using Excel and modern modelling languages such as PuLP and JuMP. This project will extend SolverStudio by adding support for more modelling languages (including SCIP), adding new visualisation capabilities, and creating a new interface for creating and editing OpenSolver models within SolverStudio.

The student will need good programming skills, and will ideally have experience with Python and perhaps Visual Basic. Experience in C# would give more scope for the range of improvements that could be implemented, but is not essential.

Advanced tools for Analytics Visualisations in Excel


Supervisor

Andrew Mason

Tony Downward

Discipline

Engineering Science

Project code: ENG101

Microsoft have recently released a new coding framework for extending Excel with browser-based tools. We have some limited experience in using these new capabilities to create new visualisations and interactive tools within Excel. This project will build on this preliminary work to create new opens-source Excel add-ins based on this new technology. Possible applications include adding new plotting capabilities to SolverStudio, displaying interactive visualisations of OpenSolver’s Branch and Bound process, building interactive Gantt charts and staff scheduling displays, building 3D visualisations for the Simplex algorithm, etc.

The student will need experience programming in JavaScript and developing web sites with client-side code. Experience with the Microsoft Office Javascript API’s would be a bonus.

OpenSolver for Libre Office


Supervisor

Andrew Mason

Discipline

Engineering Science

Project code: ENG102

OpenSolver is currently only available for Excel, which makes it inaccessible to those who choose to use open source spreadsheets such as that in Libre Office. This project will develop a first release of OpenSolver for Libre Office.

The student will need experience with Linux, good programming skills, and a willingness to learn much more about the inner workings of Libre Office than is known by their supervisor!

Analytics for Identifying Māori Shareholders


Supervisor

Andrew Mason

Andy Philpott

Discipline

Engineering Science

Project code: ENG103

We are commencing a National Science Challenge project that involves developing analytics and optimisation tools that can help identify missing shareholders. Many Māori organisations have a growing number of shareholders and face the very real challenge of keeping track of these shareholders. This is not easy, as demonstrated, for example, by the over $3,000,000 of dividends that the Taranki-based organisation, Parininihi ki Waitotara (PKW), has owing to missing shaeholders. Similar problems are faced by the Māori Trustee who is tasked with managing Māori land, where each land parcel can have 100’s of owners. We are developing analytics tools that use public information to help organisations such as PKW better understand and locate its missing shareholders. We are seeking a student with good programming and mathematical modelling skills to work with us in developing these tools.

Advanced non-linear optimisation for OpenSolver – the open source optimiser for Excel


Supervisor

Andrew Mason

Discipline

Engineering Science

Project code: ENG104

This project will develop a new enhanced non-linear tool for OpenSolver – a popular open source optimiser for Excel that has been downloaded over 280,000 times. OpenSolver currently solves non-linear models by translating the Excel spreadsheet formula into various internal formats before writing out a file suitable for a non-linear solver. This leads to many problems, including memory overflows, crashes, poor error messages, and slow solving. We are seeking a student with good VBA skills to create a new approach for handling non-linear problems.  These changes will address the most common complaints from our OpenSolver community, and make a real difference to our thousands of users.

Prototyping Container-based Simulation Modelling


Supervisor

Assoc Prof Cameron Walker
Dr Michael O’Sullivan

Discipline

Engineering Science

Project code: ENG105

Containers are light-weight, virtualised computers that exist in a virtualisation environment. By simulating parts of a system on individual containers, the overall system simulation can be developed in a modular, hence agile, process. However, a knowledge broker needs to coordinate the simulation “pieces” to ensure the entire system works as intended. This project will prototype this approach on a developmental virtualisation environment to simulate multiple health departments working together to provide patient treatment.

Bioremediators for Clean Water


Supervisor

Richard Clarke

Discipline

Engineering Science

Project code: ENG115

This project will look at designing and building portable bioremediation units for cleaning NZ's rivers and lakes. It will involved setting up and executing Computational Fluid Dynamics simulations to determine suitable unit designs, and then constructing a working template for physical testing.