25813 research outputs found
Sort by
Accurate classification of bloodstain spatter patterns using machine and deep learning optimisations.
No full textBloodstain Pattern Analysis is a forensic discipline focused on the examination of bloodstains at crime
scenes with a view to assisting with the reconstruction of events. Analysis of the physical
characteristics of individual bloodstains and the overall pattern can assist with determining the
mechanism that caused them, a task referred to as classification. Currently, the analysis of bloodstain
patterns is still a manual and somewhat subjective task. With error rates of 13-23% for classification
of spatter patterns, a more objective method is recommended.
In order to develop objective methodologies, quantitative data from high-quality and reliable empirical
research is required. As such, a large and varied dataset of bloodstain patterns from three classes of
spatter patterns (186 impact, 162 cast-off and 132 expirated) has been created on vertical paper
targets. To capture the entirety of the patterns that contains numerous small bloodstains spread over
large areas, these images were composed of multiple high-resolution images stitched together,
ranging in size from 7 to 81 megapixels. During the creation of these patterns, key variables such as
object mass, velocity, and type, blood volume, mechanism, and distance to target, were investigated
with regards to their influence on pattern characteristics in relation to stain number and size.
Proposed computer vision and machine learning methods for the classification task of bloodstain
pattern analysis has also been explored. An improved automated bloodstain pattern analysis (ABPA)
method was developed and validated, which proposed a computer vision process to identify
bloodstains on a plain background within a digital image. The method generates metrics relating to
the individual stains as well as the overall pattern, including bloodstain pattern specific metrics such
as the gamma angle, circularity, solidity, area of convergence, stain density and pattern linearity. This
proposed method provides an objective process for the analysis of bloodstains and bloodstain
patterns and can generate a wealth of quantitative data that is currently not obtainable using manual
techniques or other image-based programs currently utilised in the discipline. Validation of the
proposed ABPA method demonstrated that the method has improved stain segmentation compared
to the prior ImageJ and Matlab developed methods and is thus able to detect more stains in the
images.
This novel method was then applied to the laboratory generated dataset and metrics retrieved from
each pattern. Data analysis was undertaken on over 540,000 individual stains to generate several
metrics that were summarised for each pattern. This data was used to train three machine learning
classifiers: random forest, SVM and neural network. Analysis and further improvements of the data
were proposed which increased the accuracy of the three proposed machine learning classifiers from
~94% to ~98%. These results support the view that objective methodologies can be successfully
implemented for the classification task of BPA.
The features within the optimised training dataset were then systematically altered and the changes in classifier accuracies were used to investigate which features were the most influential in providing accurate classification of the three patterns. The proposed linearity ellipse feature which describes
the size and shape of the overall pattern (by defining the values of a fitted ellipse) proved to have the
largest influence on classifier accuracy.
A further proposed classification model based on the analysis of the pattern images themselves using
deep learning with Convolutional Neural Networks (CNNs) was also explored. The large size of the
training images created a challenge for training the CNNs, with novel preprocessing steps proposed
to create appropriate training data. Several pre-processing techniques, including dilation, resizing and
cropping, were evaluated and used to further train two proposed CNN models based on ResNet-50
and VGG-16. Dilated and resized images representing the full extent of the bloodstain patterns
reached accuracies of >95%, but <90% for partial patterns created through cropping
Impact of gamification on engagement and learning in video-based platforms.
No full textVideo-based learning provides flexible and self-paced learning. However, passive video consumption results in shallow learning. To address this, reflective activities such as commenting on videos and reviewing peers' comments are used to increase engagement. Incorporating gamification can further boost engagement. This study investigates the impact of gamification, specifically badges, on student engagement and learning in video-based learning. We conducted a comparative study between a gamified and a non-gamified version of the same platform. Findings show that badges significantly increased learner engagement. Student feedback indicated a positive reception of badges. Also, students who used the gamified platform wrote higher quality reflective comments on videos, leading to increased learning
Development and clinical validation of an ultra low-cost insulin pump, for more equitable diabetes care.
No full textDiabetes Mellitus is a chronic disease and a leading cause of healthcare spending, with 1 in every
4 healthcare dollars in the US being spent caring for a person with diabetes. The prevalence of
diabetes is increasing rapidly, and by 2045 it is predicted that 8.5% of the world's population
will have been diagnosed with some form of diabetes, with care quickly becoming unaffordable
if new solutions are not implemented.
It is not only the direct costs of diabetes that result in the financial burden of the disease being
so high. If diabetes is poorly controlled in its early stages, significant complications are likely
to occur in the long term, with cardiovascular disease, stroke, kidney disease, kidney failure,
and vision loss being more commonly associated with poorly controlled diabetes. Maintaining
normoglycemia can be difficult for those with diabetes, with complex management strategies
required to keep blood glucose levels within a normal range.
Currently, insulin pump therapy represents the gold standard of diabetes treatment, with
large benefits demonstrated in terms of blood glucose control and a reduction in negative
psychosocial factors associated with diabetes. However, while insulin pumps currently stand
as the best treatment method for diabetes, their use is limited due to a high initial purchase and
ongoing consumables cost. As such, their use is limited to those who can afford the devices
themselves or pay for an insurance policy that provides one. In countries with socialized
healthcare, the high cost of pumps results in limited access or places a significant burden on the
taxpayer.
The aim of this thesis is to develop a low-cost, open-source insulin pump, which offers
similar performance to commercial models in both lab-based and clinical testing. Using a mix
of low-cost, commonly available components and 3D-printed parts, an initial prototype was developed to ascertain whether the low cost of these components would limit the ability to
match a commercially available device in standards-based testing. Lab-based results showed
both devices exhibited a delivery error of less than 0.2% at an intermediate basal rate, and when
tested at a low basal delivery rate, the low-cost device was shown to be 3X more accurate than
a commercial device. Bolus testing showed similar results, with the low-cost device performing
as well as or better than the commercial device.
These promising initial results warranted the development of a more functional device,
incorporating features from commercial devices to make the updated low-cost device more
accessible and user-friendly. Bluetooth connectivity was added, in addition to hardware to track
insulin deliveries and automate common functions. An Android companion app was developed
to allow control of the insulin pump from a mobile device. The updated pump was subject to
the same lab-based testing as the previous device to assess its delivery accuracy and suitability
for use in an inpatient clinical trial.
During lab-based testing, the updated, low-cost pump showed an overall delivery error of
7.46% above the set rate when tested at a low basal rate, and 3.7% above the set rate at an
intermediate basal rate. Regarding bolus deliveries, the pump generally showed a range of error
of 5% or less and a tendency to under-deliver. An analysis was conducted of a range of current
commercial insulin pumps for comparison, with the low-cost device found to sit centrally in
terms of accuracy when compared to the other devices. Therefore, it was decided to progress to
an inpatient clinical trial and ascertain if the positive results of the lab-based testing could be
translated into effective diabetes control in real patients.
A first-in-human trial of the low-cost device was conducted, testing performance over a period
of 10 hours. In addition, participants took part in a focus group to assess how a low-cost device
might be received by the diabetes community. Results showed all participants' blood glucose,
β-OH-B, and insulin assays were consistent with insulin delivery. Blood insulin and blood glucose
remained steady throughout the trial, suggesting that the ULCIP was delivering the correct
volume of insulin as required by the patient. One issue was identified: insulin delivery failed due
to an underfilled cartridge. Once the cartridge had been properly filled, this participant exhibited
normal blood glucose and blood insulin levels until the trial concluded. Overall, participants showed increased time-in-range when using the ULCIP, even when normalizing participants'
previous data to account for the time window of the trial. Participants' opinions of the ULCIP
were generally positive, although opinions were mixed regarding the simplicity of the device and
the lack of customer support.
In addition to presenting the ULCIP, this thesis also proposes a novel, low-cost testing
methodology for insulin pumps. The equipment required to test an insulin pump to the relevant
standard can cost in excess of US1, meaning patients and clinicians could easily test their own devices, ensuring proper
function and delivery accuracy. This thesis has also proposed modifications to standards-based
testing, as well as providing explanation and solutions to ambiguities present in the methodology.
In summary, this thesis presents a new approach to low-cost diabetes care. Results presented
here have shown low-cost hardware is not a barrier to effective insulin pump therapy, and the
high cost of these devices is not a result of the hardware itself. It is hoped the design presented in
this thesis, backed by clinical results in the first trial of its kind, can be used as a starting point
to be built upon by the open-source diabetes care community, providing an affordable option
for those currently unable to afford an insulin pump. In addition, this thesis has also presented
a novel and arguably simpler method of insulin pump testing, which has further demonstrated
that a low-cost solution does not need to compromise quality. While there is understandably
always resistance to unregulated, open-source medical care, it is hoped the results and design
presented in this thesis will be enough to attract further development, resulting in an alternative
option for those with diabetes who wish to commence insulin pump therapy but cannot afford
or access a device
Future roles for our former natural forests: Comparing the benefits of Ginkgo and Kahikatea within a changing climate for Waitaha Canterbury. Report prepared part of the GEOG309 Research for Resilient Communities and Environments course, University of Canterbury, 2025.
No full textAs climate change continues to affect our environment, the roles of trees in rural and urban environments are being more crucial.• In our changing climate, where in Waitaha Canterbury do ginkgo and kahikatea trees provide more benefit?• Literature review of research of Mātauranga Māori and ginkgo and kahikatea history.• “Right tree, right place, right purpose” framework was used to inform our selection of tree species after our community partner suggested investigating ginkgo.• Used satellite images, aerial images, and deep learning models in ArcGIS Pro to create maps on tree coverage and diversity in Waitaha Canterbury.• Soil samples were collected from Christchurch and Banks Peninsula to measure soil moisture, soil organic matter and soil carbon around each tree species.• Leaf area index measurements taken to measure shade potential of each tree.• Both ginkgo and kahikatea have medicinal and cultural uses to different groups.• Areas in the city centre, Eastern suburbs, and Canterbury Plains are lacking canopy coverage. Eastern and Western suburbs have the lowest tree diversity.• Kahikatea trees generally have greater soil moisture, soil organic matter and soil carbon compared to the ginkgo trees.• Ginkgo provides more shade in summer and less in winter compared to kahikatea. It is also tolerant to many common urban pollutants.• Moving forward, where we plant each species should be informed by mana whenua, community and science.• Limited peer reviewed literature available on kahikatea, not enough time in the project to develop effective relationships with subject matter experts.• Limited data in rural areas for accuracy of canopy coverage.• Seasonality and project timeframe prevented some primary data collection and reduced reliability of other data.• Recommend expanding data and research into the lack of shelter in dairy farms and other livestock.• Expand scope to collect direct measurements of impacts on UHI over a longitudinal study.• Other factors that impact soil health such as prior and current land management, climate, soil parent material and natural features.• Further research is needed into evergreen native trees, such as the totara
Tire-wear particles and tire-related emerging contaminants: Characteristics, occurrence, and toxicity in the environment
No full textThe discovery of 6PPD-quinone as a toxic transformation product and aqueous contaminant from a common tire antioxidant has catalyzed extensive research into its environmental occurrence and other tire-related emerging contaminants (TRECs). However, many other tire-derived chemicals remain understudied, yet may pose significant environmental risks. This review provides a critical overview of the characteristics, occurrence, and fate of TRECs, as well as their toxicological impacts. While advancements in analytical techniques have enhanced the detection of these contaminants, knowledge gaps exist regarding their persistence, transformation, and long-term effects in various environmental compartments. Recent findings suggest that several TRECs, including diphenylguanidine (DPG) and hexamethoxymethylmelamine (HMMM), are frequently detected in road runoff, wastewater, sediments, and even biota. Given their widespread occurrence and potential toxicity, further investigations into their environmental behavior, regulatory implications, and mitigation strategies are urgently needed. This review highlights the pressing need for expanded research beyond 6PPD-quinone to address the broader risks associated with TRECs
The effects of reducing design drift limits for structures
No full textObservations from past earthquakes and experiments have shown that stiffer structures experience less deformation, leading to less damage to the structure as well as drift-sensitive nonstructural components such as partitions, facades, and infill walls. These observations suggest that reducing the design storey drift limit would be a simple and effective method of reducing structural and nonstructural damage in the event of earthquake. This study aims to quantify the effects of reducing design drift limits for buildings by (1) compiling quantitative field evidence of buildings after past earthquake, 2) performing nonlinear dynamic analyses of archetypal structures designed to different limits, 3) conducting full-scale experimental tests of suspended ceiling and plasterboard walls subjected to floor motions, and 4) estimating the upfront cost and seismic loss for structures designed to different drift limits.
Reports and surveys from 1639 buildings after 15 different earthquakes were compiled and structural indices were used to compare the frequency of severe damage for structures. Structures with a Wall Index larger than 0.2% (wall area normalised by the total floor area) were approximately 10 times less likely to sustain severe damage when compared with structures with a Wall Index of less than 0.1%. Surveys and reconnaissance data of 111 Hospital building, 85 reinforced concrete (RC) wall structures, and 26 RC frames, were compiled after the 2010 Maule Earthquake. From surveys of each building, frame structures were approximately five times more likely to sustain damage to nonstructural components and/or “acceleration-sensitive” components, suggesting that stiffer RC wall structures do not lead to more damage to “acceleration-sensitive” components.
To quantify the effect of reducing design drift limits on the structural performance, 16 RC wall-frame structures and 16 RC frame structures were designed to different drift limits ranging from 2.5% to 1.0% for a nominal 500 year return period earthquake. Nonlinear dynamic analyses was performed and the structures were evaluated in terms of the estimate collapse risk, peak floor acceleration demands, and peak storey drift demands. For a large intensity ground motion with a nominal 2500 year return period, reducing the drift limit from 2.5% to 1.0% led to a 10-fold reduction in the likelihood of collapse. Reducing the drift limit from 2.5% to 1.0% led to modest increases in floor acceleration demands of approximately 40%.
To investigate whether or not structures designed to lower drift limits have less drift capacity, data from experimental tests of 341 RC columns subjected to cyclic loading were compiled. A machine learning model was developed to obtain reliable estimates of drift capacity and used to identify which column properties influence drift capacity. While it was observed that drift capacity reduced for RC sections with an aspect ratio of less that 2, this reduction in drift capacity can be offset by (a) the reduction in axial load ratio related to larger sections, and (b) providing additional transverse reinforcement. For structures with RC walls, the design drift can be reduced by increasing the quantity or thickness of walls while providing adequate detailing and not reducing the aspect ratio. The results suggest that structures can be designed to lower drift limits without reducing the drift capacity of the structures.
Reducing the design drift limit also has implications for contents and nonstructural elements. It is well understood that components such as partitions, facades, and infill walls are sensitive to excessive drift, and reducing the drift limit will have a positive effect. Nevertheless, the current consensus within the profession is that suspended nonstructural elements such as piping, ceilings, and services are sensitive to acceleration demands. It is thought that reducing the drift limit will increase floor accelerations and therefore lead to more damage to these elements. Nevertheless, anecdotes and reports from the field emphasize that stiff RC wall buildings are less susceptible to nonstructural damage. Experimental dynamic tests of suspended ceilings and plasterboard walls were conducted to understand how the reducing drift limit influences the behaviour of nonstructural components. A two-storey dynamic floor motion simulator was designed and constructed to simulate floor motions representative of flexible and stiff structures to examine the effect of floor motions on nonstructural components. For the configurations of nonstructural components considered in the experimental programme, both suspended ceilings and plasterboard walls were observed to be sensitive to drift demands.
Current estimates of material and labour costs were used to quantify changes to the upfront cost associated with reducing the design drift limit of hypothetical buildings located in Wellington, New Zealand. The increase in upfront building cost related to reducing the drift limit from 2.5% to 1.0% was approximately 1.0%. Seismic loss assessments were also conducted to evaluate changes in the expected loss for the different structures. Accounting for the expected loss and the upfront cost, it was observed that stiffer RC wall and frame structures designed to a drift limit of 1.0% or 1.5% led to a cost saving of 5-10 over the lifespan of the building.
This study highlights the benefits of reducing the design drift limit and raises concern regarding structures designed to an allowable drift limit of 2.5%, considering the approximate 10-fold increase in collapse risk and 3-fold increase in the expected annual loss
Navigating the sustainability landscape : an exploration of sustainability performance and reporting in Indonesian companies
No full textIndonesia's rapid economic growth presents a compelling context for exploring the intersection of sustainability practices and development challenges. This thesis examines the current state of sustainable development practices in Indonesia, focusing on the complex interplay between sustainability reporting, performance, and broader development goals. It draws upon five individual papers, each offering a unique perspective on the evolving landscape of corporate sustainability in Indonesia. This thesis employs a mixed-method approach, combining interpretivist principles with qualitative data collection and analysis. Semi-structured interviews with 27 managers from eight Indonesian listed companies and three representatives of two institutional investors provide insights into stakeholder perspectives on sustainability performance and reporting practices. Additionally, a content analysis of sustainability reports from eight Indonesian listed companies across various industries sheds light on sustainability communication strategies and materiality considerations. Thematic analysis techniques are employed to identify key themes and patterns within the collected data.
The five individual papers explore specific aspects of sustainability performance, with four focusing on the Indonesian context. The first paper (published in Corporate Governance and Sustainability Review, hXps://doi.org/10.22t95/cgsrv8i3p1) proposes a new theoretical framework of ethical sustainability governance to guide organizations toward a more holistic approach to sustainability. The second paper (published in the Journal of Risk and Financial Management, hXps://doi.org/10.3390/jrfm17070296) investigates the inconsistencies in environmental, social, and governance (ESG) ratings assigned to palm oil companies in Indonesia, highlighting the need for improved standardization and transparency. The third paper (published in Qualita@ve Research in Financial Markets, hXps://doi.org/10.1108/QRFM-09-202t-02t1) explores the key drivers of sustainability performance in Indonesian banks, identifying factors that influence their sustainability initiatives. The fourth paper (published in Pacific Accounting Review, hXps://doi.org/10.1108/PAR-11-202t-0308) examines management and investor perspectives on sustainability practices, focusing on motivation for reporting and materiality considerations. Finally, the fiqh paper (published in Meditari Accountancy Research, hXps://doi.org/10.1108/MEDAR-11-202t-2738) examines the implementation of double materiality in Indonesian companies, examining the role of internal interactions between accounting and sustainability functions, and external ESG risk ratings in identifying material sustainability issues.
This thesis offers an analysis of sustainability performance and reporting in Indonesia, providing valuable insights for academics, companies, investors, and policymakers. It provides several contributions to the literature. First, this thesis proposes a novel theoretical framework of ethical sustainability governance and its practical implementation guidance to enable organizations to focus on sustainability performance rather than solely on reporting. Second, this thesis provides a nuanced understanding of the current state of sustainability practices in Indonesia, considering the unique challenges and opportunities faced by companies in this developing market. Third, this thesis highlights the importance of sector- specific ESG scoring methodologies, stakeholder engagement, materiality, sustainability strategy, external assurance, financing strategy, sustainability certification, and Green Taxonomy implementation as key drivers of sustainability performance. Fourth, this thesis reveals additional factors influencing sustainability integration, including customer-centric approaches, internal interaction between accounting and sustainability functions, and the combined pressures of regulatory compliance, investors, and lenders as a primary motivation of sustainability integration. Fiqh, this thesis provides empirical evidence of the benefits of sustainability by examining the potential positive relationship between sustainability performance, firm performance, and financial valuation, and highlighting the long-term benefits of sustainable practices.
By addressing these key areas, the thesis contributes to the development of new theoretical insights as well as practical guidance for ethical sustainability governance. Such insights can foster better sustainability performance and a deeper understanding of sustainable development practices in Indonesia. The aim is to offer both theoretical contributions to the academic literature and actionable recommendations for companies to promote sustainable development
Ngā hoe reo, kia whakatere i te waka. Navigating the journey : the value-creation stories of teachers’ experiences of learning te reo Māori and using te reo Māori in their educational settings.
No full textMy thesis presents my investigation into teachers’ experiences as both learners of te reo Māori—the indigenous language of Aotearoa New Zealand—and as practitioners integrating the language into their practice. Research suggests that tamariki (children) in English-medium schools have limited exposure to te reo Māori. To address this disparity, the Labour-led government’s policy agenda on revitalising te reo Māori and improving Māori student achievement required both Māori and non-Māori teachers to learn and use the language in their early childhood education settings and school classrooms.
Despite this agenda, a shortage of teachers with the requisite knowledge, skills, and confidence to integrate te reo Māori into their teaching practice has made policy implementation difficult. Teachers face challenges in interpreting policy mandates let alone implementing them. Many lack even basic knowledge of language-acquisition or language-revitalisation policy. These challenges affect teachers’ ability to use or integrate te reo Māori effectively, which undermines their confidence and investment in learning te reo Māori. This situation raises important questions about teachers’ experiences as they navigate the process of learning te reo Māori and integrating it into their daily practice.
Using a qualitative narrative methodology, I collected narratives from 14 registered teachers in early childhood, primary, and secondary educational settings in a semi-rural region of Te Wai Pounamu (the South Island). The participants, predominantly wāhine (female) and non-Māori, shared their experiences of learning and using te reo Māori, a process that revealed both positive and negative experiences. My particular interest was in understanding how these teachers were experiencing and creating value through their engagement with te reo Māori, within the context of social learning spaces. Drawing on the Wenger-Trayner’s value-creation framework as an analysis tool, I analysed the participants’ narratives with the aim of identifying the types of value created across different stages of engagement. My overall intention in applying this framework was to determine whether and how the teachers’ te reo learning supported them in understanding and implementing the New Zealand Government’s te reo policies.
The value creation framework approach for data analysis provided a deeper understanding of how teachers’ language-learning journeys contribute to their professional development and practice, even when they may not explicitly articulate these experiences as ‘valuable’. The value-creation framework was applied during the data-analysis phase to interpret teachers’ reflections on their experiences of learning and using te reo Māori in practice. Using the value-creation framework enabled identification of immediate benefits (immediate value), potential for future application (potential value), integration into practice (applied value), and broader professional impact (realised value). This analytical process highlighted how learning te reo Māori can generate value for teachers on their reo Māori improvement journeys across both policy implementation and classroom practice. The findings also suggest that social learning spaces play a critical role in supporting engagement, knowledge capital, and confidence building.
Although the teachers also experienced challenges such as limited proficiency and student resistance when attempting to use or integrate te reo Māori, they recognised the importance of te reo Māori to Aotearoa New Zealand’s identity. They valued the survival of te reo Māori, understood their role in its revitalisation, and considered this awareness contributed to their professional growth and a reshaping of their professional identities.
Overall, the study highlights the need for te reo Māori mentors to guide teachers, helping them draw on the value created from their learning experiences in ways that support sustained language-in-practice. With this consideration in mind, I also propose in this thesis a new conceptual model, Toia Mai Te Waka, to enhance existing mentoring programmes for both new and existing teachers. I furthermore suggest that teachers use Kāhui Ako (communities of learning) or a similar model to help them focus on collective, sustainable language planning, as well as authentic and robust measures of accountability for policy, legislation, and Teaching Council Aotearoa New Zealand mandates relating to te reo Māori learning across the Aotearoa New Zealand education sector and integration in pedagogical practice
Portraying protection : newspaper depictions of the smallpox vaccine in Washington, D.C., 1800–1850, and the evolution of vaccine activism and discourse.
No full textHistoriography regarding American vaccination activism often focuses on the post-Civil War era, as this was an introductory period for multiple American anti-vaccine societies and vaccine movements. This research essay focuses on tracking the depictions of the smallpox vaccine through newspapers published in Washington, D.C., from 1800 to 1840, and investigates the way vaccine activism and discourse changed over this period. This paper uses keyword searches in the Library of Congress Digital Catalogue to navigate articles and their contents in order to analyse a selection of articles from Washinton D.C. publications. These methods made it possible to catalogue individual references to vaccination discussions and activism, resulting in a coherent narrative throughout the early period. By tracking the trends of activism shown in each decade of this earlier period, a more holistic understanding of vaccine activism can be garnered. This research supports the position that historiography focused on American vaccine activism should refer to this introductory period, to contextualise later positive and negative vaccine activist arguments and shows how quickly the public image of vaccination evolves. My findings reveal significant shifts regarding the methods of vaccine advocacy and in how vaccination was represented. A significant shift between 1800 and 1840 was how vaccination was interacted with by physicians. The vaccine was initially characterised by physicians as completely safe and unquestionably beneficial, whereas physicians in later decades recognised that improper procedure could render vaccines ineffective and dangerous. This recognition led to the adoption of specific measures to help restore public confidence. A mixture of articles, advertisement, letters to the editor and Governmental minutes reveal a range of themes and trends regarding how vaccination was advocated and discussed in the media between 1800 and 1840