7103 research outputs found
Sort by
Reconstruction of the built environment in post-conflict contexts : a case study of Benghazi after the 2014-2017 conflict
Full text linkThesis author: Saleh Almogrbe --
Degree supervisor: Prof. Branka DimitrijevicThis thesis examines the post-conflict reconstruction of Benghazi, Libya, following the armed conflict of 2014-2017, focusing on the challenges, approaches, and effectiveness of reconstruction efforts from the perspective of householders. Using a mixed-methods approach combining quantitative surveys of 238 householders with qualitative interviews of 10 participants, the study provides a comprehensive analysis of reconstruction processes and their impacts on the urban environment and local communities. The research findings reveal significant disparities between implemented reconstruction projects and community needs. Survey data indicates widespread dissatisfaction with recovery efforts, with only 1.76 out of 5 mean satisfaction score regarding municipal efforts to restore life and activity in the city. The study found that 31.1% of residents were displaced to other cities during the conflict, while 27.3% reported minor damage to their homes and 8.4% experienced complete destruction. Notably, 88% of respondents received no aid or support during the conflict period. The research identifies several critical challenges in Benghazi's reconstruction process: fragmented governance structures, unresolved property rights issues stemming from Libya's socialist era, limited community consultation (71% reporting no consultation about projects), and inadequate coordination among reconstruction actors. Unlike the internationally coordinated reconstruction of Mostar or the private sector-driven rebuilding of Beirut, Benghazi's recovery has been characterized by competing priorities and limited stakeholder coordination. This study contributes to the field in three key ways: first, by providing empirical evidence of reconstruction challenges from a householder perspective; second, by examining the relationship between physical reconstruction and social recovery in the Libyan context; and third, by developing practical recommendations for improving reconstruction processes based on community needs and experiences. The research offers insights into how post-conflict reconstruction can better serve affected populations while promoting sustainable urban development. The findings suggest that successful post-conflict reconstruction requires not only technical expertise and financial resources but also careful attention to social dynamics, property rights, and community needs. The study recommends an integrated approach to reconstruction that balances immediate recovery needs with long-term development goals while emphasizing the importance of community participation and transparent governance.
Key Words: Benghazi – Heritage - Libya - Post-Conflict - Reconstruction - Urban Development- Old cityThis thesis examines the post-conflict reconstruction of Benghazi, Libya, following the armed conflict of 2014-2017, focusing on the challenges, approaches, and effectiveness of reconstruction efforts from the perspective of householders. Using a mixed-methods approach combining quantitative surveys of 238 householders with qualitative interviews of 10 participants, the study provides a comprehensive analysis of reconstruction processes and their impacts on the urban environment and local communities. The research findings reveal significant disparities between implemented reconstruction projects and community needs. Survey data indicates widespread dissatisfaction with recovery efforts, with only 1.76 out of 5 mean satisfaction score regarding municipal efforts to restore life and activity in the city. The study found that 31.1% of residents were displaced to other cities during the conflict, while 27.3% reported minor damage to their homes and 8.4% experienced complete destruction. Notably, 88% of respondents received no aid or support during the conflict period. The research identifies several critical challenges in Benghazi's reconstruction process: fragmented governance structures, unresolved property rights issues stemming from Libya's socialist era, limited community consultation (71% reporting no consultation about projects), and inadequate coordination among reconstruction actors. Unlike the internationally coordinated reconstruction of Mostar or the private sector-driven rebuilding of Beirut, Benghazi's recovery has been characterized by competing priorities and limited stakeholder coordination. This study contributes to the field in three key ways: first, by providing empirical evidence of reconstruction challenges from a householder perspective; second, by examining the relationship between physical reconstruction and social recovery in the Libyan context; and third, by developing practical recommendations for improving reconstruction processes based on community needs and experiences. The research offers insights into how post-conflict reconstruction can better serve affected populations while promoting sustainable urban development. The findings suggest that successful post-conflict reconstruction requires not only technical expertise and financial resources but also careful attention to social dynamics, property rights, and community needs. The study recommends an integrated approach to reconstruction that balances immediate recovery needs with long-term development goals while emphasizing the importance of community participation and transparent governance.
Key Words: Benghazi – Heritage - Libya - Post-Conflict - Reconstruction - Urban Development- Old cit
Application of deep generative models for the design of pharmaceutical manufacturing processes
Full text linkDesigning processes for pharmaceutical product manufacturing is a complex and
resource-intensive task. With increasing research costs and quality standards, the
pharmaceutical industry seeks innovative technologies to enhance productivity and
maintain competitiveness. While a variety of tools exist in the process design domain for
optimizing conditions or selecting materials, options for guiding the selection of
manufacturing operations remain limited.
In this context, deep generative models (DGMs) emerge as a promising approach.
DGMs, known for learning the probability distribution of data, have gained popularity for
their ability to generate realistic examples, commonly applied in text and image
generation. In drug discovery, DGMs have successfully generated new active substances
with desirable properties. However, their application in the manufacturing space remains
unexplored. These models have the potential to assist in operation selection and
experimental targeting, thereby reducing development time.
This thesis aims to investigate the applicability of DGMs in pharmaceutical manufacturing
process design, developing DGMs capable of generating plausible sequences of
operations for product manufacturing, taking input information about the target product.
A significant challenge in developing DGMs is the requirement for large datasets. To
address this, two datasets were constructed using natural language processing (NLP)
applied to primary and secondary manufacturing data extracted from patents. The
primary processing dataset comprises over 385K manufacturing processes, while the
secondary processing dataset includes approximately 9K procedures for various dosage
forms and active ingredients.
The study involved training and comparing several architectures based on generative
adversarial networks (GAN) and variational autoencoder (VAE) using different metrics.
Real and generated sequences were contrasted manually to evaluate how closely the
model outputs resemble typical manufacturing sequences. This research contributes to
the exploration of DGMs’ application in pharmaceutical manufacturing, offering insights
into their potential for operation selection and product development. In the end, DGMs
were successfully trained and their potential for the generation of plausible sequences
was demonstrated. A survey assessed by a panel of experts yielded that the models generated sequences at least as good as the actual procedures in 38% of occasions for
the primary domain. While this shows the potential of generative modelling in this field, it
also remarks there is room for improvement to make it applicable in real-world scenarios.Designing processes for pharmaceutical product manufacturing is a complex and
resource-intensive task. With increasing research costs and quality standards, the
pharmaceutical industry seeks innovative technologies to enhance productivity and
maintain competitiveness. While a variety of tools exist in the process design domain for
optimizing conditions or selecting materials, options for guiding the selection of
manufacturing operations remain limited.
In this context, deep generative models (DGMs) emerge as a promising approach.
DGMs, known for learning the probability distribution of data, have gained popularity for
their ability to generate realistic examples, commonly applied in text and image
generation. In drug discovery, DGMs have successfully generated new active substances
with desirable properties. However, their application in the manufacturing space remains
unexplored. These models have the potential to assist in operation selection and
experimental targeting, thereby reducing development time.
This thesis aims to investigate the applicability of DGMs in pharmaceutical manufacturing
process design, developing DGMs capable of generating plausible sequences of
operations for product manufacturing, taking input information about the target product.
A significant challenge in developing DGMs is the requirement for large datasets. To
address this, two datasets were constructed using natural language processing (NLP)
applied to primary and secondary manufacturing data extracted from patents. The
primary processing dataset comprises over 385K manufacturing processes, while the
secondary processing dataset includes approximately 9K procedures for various dosage
forms and active ingredients.
The study involved training and comparing several architectures based on generative
adversarial networks (GAN) and variational autoencoder (VAE) using different metrics.
Real and generated sequences were contrasted manually to evaluate how closely the
model outputs resemble typical manufacturing sequences. This research contributes to
the exploration of DGMs’ application in pharmaceutical manufacturing, offering insights
into their potential for operation selection and product development. In the end, DGMs
were successfully trained and their potential for the generation of plausible sequences
was demonstrated. A survey assessed by a panel of experts yielded that the models generated sequences at least as good as the actual procedures in 38% of occasions for
the primary domain. While this shows the potential of generative modelling in this field, it
also remarks there is room for improvement to make it applicable in real-world scenarios
Modelling of the Saudi Arabia energy system with the future 100% renewable energy systems scenarios for power supply for all sectors coupled with transport sector electrification
Full text linkSaudi Arabia depends on fossil fuels for its energy production. Oil and natural gas are the primary natural resources used. In 2018, the total CO2 emissions were 491.7 Mt, which indicates that Saudi Arabia is one of the top CO2-emitting countries in the world. This work investigated possible pathways to 100% renewable energy supply by 2050 for the power from entire sectors with transport electrification. Three main gaps were identified: lack of a comprehensive dataset representing the current energy system, a validated hourly model, and a modelling study exploring 100% renewable options. To address these gaps, research was divided into two parts. First, comprehensive data gathering, review, and analysis were carried out for all energy sectors in the Kingdom. In this part, data collection and calculations were carried out for each sector in the Kingdom, providing the correct and validated information. In addition, the dataset was used to create a comprehensive energy balance block diagram of the entire Kingdom’s sectors. A comprehensive review and dataset were conducted to fix the lack of data and clarity in several parts of the Saudi energy system. Then, in the second part, the new data set created was used to inform and validate the energy model of Saudi Arabia, allowing future scenarios in 2050 to be assessed from technical and economic perspectives. After the energy model was created and validated with actual data, 100% RES using solar photovoltaic, wind, and battery storage was investigated for power from all sectors. First, each RES technology was studied and simulated solely and assessed from a technical and economic perspective with limitations. Then, different combinations of RES are evaluated based on the same criteria and selected based on the limitations of the previous system in a series of gradual, cumulative improvements to reach the final optimal system. The renewable energy systems are compared and evaluated, and the final optimal system is identified. A green hydrogen power plant was created as a backup system to work in case any shortage occurs due to unusual events and changes in 2050. The backup system used green hydrogen, which was 100% generated from the surplus power of the renewable energy system through electrolysers. The passenger vehicle fleet of the transport sector in the Kingdom was 100% electrified using the surplus power generated from the renewable energy system in 2050. It was found that the combination of photovoltaics located in the Tabuk region, wind turbines located in NEOM city, and battery storage in addition to a green hydrogen backup plant was the optimal solution to supply the entire Kingdom power in 2050, technically and economically. In addition, carbon dioxide emissions were reduced by 60% in 2050 in the Kingdom with the renewable energy
systems scenario compared to the same year in the business-as-usual scenario. Finally, the results are discussed, and the conclusion is carried out. The limitations of this work, future
work, and recommendations were identified. The contributions of the work in directly addressing the identified gaps are discussed.Saudi Arabia depends on fossil fuels for its energy production. Oil and natural gas are the primary natural resources used. In 2018, the total CO2 emissions were 491.7 Mt, which indicates that Saudi Arabia is one of the top CO2-emitting countries in the world. This work investigated possible pathways to 100% renewable energy supply by 2050 for the power from entire sectors with transport electrification. Three main gaps were identified: lack of a comprehensive dataset representing the current energy system, a validated hourly model, and a modelling study exploring 100% renewable options. To address these gaps, research was divided into two parts. First, comprehensive data gathering, review, and analysis were carried out for all energy sectors in the Kingdom. In this part, data collection and calculations were carried out for each sector in the Kingdom, providing the correct and validated information. In addition, the dataset was used to create a comprehensive energy balance block diagram of the entire Kingdom’s sectors. A comprehensive review and dataset were conducted to fix the lack of data and clarity in several parts of the Saudi energy system. Then, in the second part, the new data set created was used to inform and validate the energy model of Saudi Arabia, allowing future scenarios in 2050 to be assessed from technical and economic perspectives. After the energy model was created and validated with actual data, 100% RES using solar photovoltaic, wind, and battery storage was investigated for power from all sectors. First, each RES technology was studied and simulated solely and assessed from a technical and economic perspective with limitations. Then, different combinations of RES are evaluated based on the same criteria and selected based on the limitations of the previous system in a series of gradual, cumulative improvements to reach the final optimal system. The renewable energy systems are compared and evaluated, and the final optimal system is identified. A green hydrogen power plant was created as a backup system to work in case any shortage occurs due to unusual events and changes in 2050. The backup system used green hydrogen, which was 100% generated from the surplus power of the renewable energy system through electrolysers. The passenger vehicle fleet of the transport sector in the Kingdom was 100% electrified using the surplus power generated from the renewable energy system in 2050. It was found that the combination of photovoltaics located in the Tabuk region, wind turbines located in NEOM city, and battery storage in addition to a green hydrogen backup plant was the optimal solution to supply the entire Kingdom power in 2050, technically and economically. In addition, carbon dioxide emissions were reduced by 60% in 2050 in the Kingdom with the renewable energy
systems scenario compared to the same year in the business-as-usual scenario. Finally, the results are discussed, and the conclusion is carried out. The limitations of this work, future
work, and recommendations were identified. The contributions of the work in directly addressing the identified gaps are discussed
Solution techniques for bi-level knapsack problems and application in allocation of healthcare funds problem
Full text linkIn consideration to the significant healthcare funds given by donors, both rich countries and philanthropic organizations, it is important to allocate this aid money in an effective way. The conventional mechanism of healthcare funds allocation to the projects is based on cost-effectiveness, i.e. the projects are ranked in their profit to cost ratios and are prioritized based on this ranking. An alternative approach of allocating subsidies to projects that are just cost-ineffective to a country rather than funding entire projects that are cost-effective has been proposed in the literature. This approach will not only encourage the country to contribute its resources but it will promote ownership of projects that are subsidized from outside. ABi-level Programming Problem (BLPP) has been used to represent this approach wherein there are two participants- a leader (donor agency) and a follower (recipient country). In this specific BLPP, referred to as Donor-Recipient Bi-level Knapsack Problem (DR-BKP) in our work, the donor has choice to subsidize healthcare projects that are implemented by the recipient country and the country has choice to take these subsidies for project selection using the further funding. There is a set of healthcare projects, each one associated with a certain profit and cost, under consideration by both participants. The cost of every project is common to the participants however the profit values may differ for them. Along with the healthcare projects, the country has an external project that is of no interest to the donor in this setup. Once the donor decides on cost subsidies for the healthcare projects that are within its individual budget, the country solves a knapsack problem with the cost subsidized projects and the external project constrained by its own budget. Starting with an introduction to BLPPs and motivation of application in a healthcare economics problem, we present the DR-BKP in chapter 1. In chapter 2, we give the literature reviewed for the BLPPs and their solution algorithms. In chapter 3, we provide evidence for ΣP 2-hardness by showing that the DR-BKP is both NP-hard and Co-NP hard. After showing the existence of a solution for the DR-BKP, we give a pair of finitely converging exact algorithms, an enumeration algorithm and a branching algorithm, and show the convergence of these algorithms. A set of fifteen differing data sets, each having randomly generated ten instances, have been generated and solved to evaluate the performance of the proposed algorithms. These data sets are generated such that they mimic a range of instances arising in real-life, i.e. starting from the ones that can be trivially addressed to the ones that are complex and difficult to solve. The complexity of some of these data sets is characterized firstly by the external project having higher valuation than the healthcare projects, and then by the discrepancy in the healthcare project valuation done by the two players. From the results of the computational experiments, it can be seen that the branching algorithm performs better when the instances are complex. We give a nested sequential approach in chapter 4 for addressing the DR-BKP, wherein the donor problem is solved using a genetic algorithm and the parameterized country problem is solved using a heuristic or an exact solver. This is followed by computational experiments of solving the fifteen data sets generated in chapter 3 using the genetic algorithm and its results. At the end of chapter 4, we summarize the performance of all the three algorithms over the fifteen data sets. The exact algorithms perform better to solve the data sets with external project having higher valuation than the healthcare projects, whereas the genetic algorithm performs better to solve the data sets having discrepancies in the healthcare project valuation by the two players. In chapter 5, we present generalizations of the proposed solution algorithms to other bi-level optimization frameworks that are closer to realistic scenarios of the application, like a single leader having multiple followers. Finally, we conclude the work done and give directions for future research in chapter 6.In consideration to the significant healthcare funds given by donors, both rich countries and philanthropic organizations, it is important to allocate this aid money in an effective way. The conventional mechanism of healthcare funds allocation to the projects is based on cost-effectiveness, i.e. the projects are ranked in their profit to cost ratios and are prioritized based on this ranking. An alternative approach of allocating subsidies to projects that are just cost-ineffective to a country rather than funding entire projects that are cost-effective has been proposed in the literature. This approach will not only encourage the country to contribute its resources but it will promote ownership of projects that are subsidized from outside. ABi-level Programming Problem (BLPP) has been used to represent this approach wherein there are two participants- a leader (donor agency) and a follower (recipient country). In this specific BLPP, referred to as Donor-Recipient Bi-level Knapsack Problem (DR-BKP) in our work, the donor has choice to subsidize healthcare projects that are implemented by the recipient country and the country has choice to take these subsidies for project selection using the further funding. There is a set of healthcare projects, each one associated with a certain profit and cost, under consideration by both participants. The cost of every project is common to the participants however the profit values may differ for them. Along with the healthcare projects, the country has an external project that is of no interest to the donor in this setup. Once the donor decides on cost subsidies for the healthcare projects that are within its individual budget, the country solves a knapsack problem with the cost subsidized projects and the external project constrained by its own budget. Starting with an introduction to BLPPs and motivation of application in a healthcare economics problem, we present the DR-BKP in chapter 1. In chapter 2, we give the literature reviewed for the BLPPs and their solution algorithms. In chapter 3, we provide evidence for ΣP 2-hardness by showing that the DR-BKP is both NP-hard and Co-NP hard. After showing the existence of a solution for the DR-BKP, we give a pair of finitely converging exact algorithms, an enumeration algorithm and a branching algorithm, and show the convergence of these algorithms. A set of fifteen differing data sets, each having randomly generated ten instances, have been generated and solved to evaluate the performance of the proposed algorithms. These data sets are generated such that they mimic a range of instances arising in real-life, i.e. starting from the ones that can be trivially addressed to the ones that are complex and difficult to solve. The complexity of some of these data sets is characterized firstly by the external project having higher valuation than the healthcare projects, and then by the discrepancy in the healthcare project valuation done by the two players. From the results of the computational experiments, it can be seen that the branching algorithm performs better when the instances are complex. We give a nested sequential approach in chapter 4 for addressing the DR-BKP, wherein the donor problem is solved using a genetic algorithm and the parameterized country problem is solved using a heuristic or an exact solver. This is followed by computational experiments of solving the fifteen data sets generated in chapter 3 using the genetic algorithm and its results. At the end of chapter 4, we summarize the performance of all the three algorithms over the fifteen data sets. The exact algorithms perform better to solve the data sets with external project having higher valuation than the healthcare projects, whereas the genetic algorithm performs better to solve the data sets having discrepancies in the healthcare project valuation by the two players. In chapter 5, we present generalizations of the proposed solution algorithms to other bi-level optimization frameworks that are closer to realistic scenarios of the application, like a single leader having multiple followers. Finally, we conclude the work done and give directions for future research in chapter 6
Data-driven prognostics and health management for maritime systems employing trustworthy digital twins
No full textFrom fronds to forests : applying dynamic energy budget (DEB) theory to individual -based and whole-forest models of laminaria hyperborea
Full text linkKelp forests are dynamic ecosystems that support biodiversity, carbon capture, and primary production while providing critical habitats for marine species. This study uses a Dynamic Energy Budget (DEB) model to simulate the growth of Laminaria hyperborea at individual and forest scales. The model tracks the pathways of carbon and nitrogen from environmental uptake to their assimilation into tissue. By integrating physiological processes with environmental factors, it reveals key insights into growth dynamics that are shaped by population structure, such as recruitment, growth and mortality, and intraspecies competition for light.
A significant result is the impact of canopy shading on growth. Cohorts under dense canopies experience constrained growth rates until pivotal events, such as mortality or thinning, allow light to penetrate the canopy, triggering rapid growth acceleration. This dynamic interaction between cohort settlement timing and canopy changes profoundly affects individual growth trajectories and forest biomass. The interplay between shading and canopy structure highlights the importance of population dynamics in kelp forest ecosystems.
The DEB model predicts individual growth curves and seasonal biomass fluctuations with accuracy, showing that shading is a critical factor for younger plants, while mature plants reach stable biomass. Simulations under future environmental scenarios, including the IPCC RCP4.5 and RCP8.5 for 2050 and 2100 periods, predict how changing conditions might alter growth and biomass patterns of L. hyperborea in Scottish waters. These findings emphasize the sensitivity of kelp forest ecosystems to environmental changes and the central role of canopy dynamics in mediating these effects.
This study underscores the importance of incorporating canopy dynamics and population structure in modelling kelp forest ecosystems. Unlike traditional approaches that extrapolate individual growth patterns to populations, this model demonstrates how canopy-driven processes influence growth limitations and accelerations. Such insights are essential for understanding the responses of kelp forests to climate change and for guiding conservation and management strategies.
All the data and code used to implement this model and the figures presented here are available at: https://gitlab.cis.strath.ac.uk/spb19186/laminaria-hyperborea-ibmKelp forests are dynamic ecosystems that support biodiversity, carbon capture, and primary production while providing critical habitats for marine species. This study uses a Dynamic Energy Budget (DEB) model to simulate the growth of Laminaria hyperborea at individual and forest scales. The model tracks the pathways of carbon and nitrogen from environmental uptake to their assimilation into tissue. By integrating physiological processes with environmental factors, it reveals key insights into growth dynamics that are shaped by population structure, such as recruitment, growth and mortality, and intraspecies competition for light.
A significant result is the impact of canopy shading on growth. Cohorts under dense canopies experience constrained growth rates until pivotal events, such as mortality or thinning, allow light to penetrate the canopy, triggering rapid growth acceleration. This dynamic interaction between cohort settlement timing and canopy changes profoundly affects individual growth trajectories and forest biomass. The interplay between shading and canopy structure highlights the importance of population dynamics in kelp forest ecosystems.
The DEB model predicts individual growth curves and seasonal biomass fluctuations with accuracy, showing that shading is a critical factor for younger plants, while mature plants reach stable biomass. Simulations under future environmental scenarios, including the IPCC RCP4.5 and RCP8.5 for 2050 and 2100 periods, predict how changing conditions might alter growth and biomass patterns of L. hyperborea in Scottish waters. These findings emphasize the sensitivity of kelp forest ecosystems to environmental changes and the central role of canopy dynamics in mediating these effects.
This study underscores the importance of incorporating canopy dynamics and population structure in modelling kelp forest ecosystems. Unlike traditional approaches that extrapolate individual growth patterns to populations, this model demonstrates how canopy-driven processes influence growth limitations and accelerations. Such insights are essential for understanding the responses of kelp forests to climate change and for guiding conservation and management strategies.
All the data and code used to implement this model and the figures presented here are available at: https://gitlab.cis.strath.ac.uk/spb19186/laminaria-hyperborea-ib
Variability in semi-automatic segmentation from CT images : implications for knee joint modelling
Full text linkThe knee joint is one of the most complex and weight-bearing joints in the body, making it highly susceptible to injury from various activities. Knee surgery often becomes necessary when conservative treatments fail to alleviate pain and other related disorders. In 2020, research indicated that there were nearly 60,000 total knee replacements (TKRs) for women and approximately 50,000 TKRs for men across England, Wales, Northern Ireland, and the Isle of Man. Projections suggest that by 2060, the demand for hip and knee replacements in the UK will rise by 40%. Robotic knee surgery, a minimally invasive and computer-assisted orthopaedic surgery (CAOS), allows for precise surgical movements, leading to quicker recovery and reduced postoperative pain. However, according to NHS Patient Reported Outcome Measures (PROMs), around 4% of patients in England remain dissatisfied with their knee replacement outcomes, primarily due to implant malalignment. Virtual 3D knee models, generated from CT and MRI scans, play a critical role in improving implant alignment before surgery. These models enable preoperative planning by allowing surgeons to virtually model the patient's knee in 3D, optimizing implant selection and simulating postoperative range of motion. However, the mechanical functionality of the knee joint remains poorly understood, and researchers are actively exploring improvements through finite element analysis (FEA). FEA is a valuable tool for simulating the mechanical behaviour of the knee under various conditions, helping surgeons and biomedical engineers analyse stress distribution, implant stability, and soft tissue interactions. Although existing finite element (FE) knee models provide highly detailed meshes of anatomical structures like bones, cartilage, ligaments, and tendons, these models are complex, time-consuming to create, and prone to human error, making them unsuitable for analysing large image datasets. This brings us to our primary research question: What is the impact of using simplified soft tissue models on finite element simulations of subject-specific knee joints? Can we create an FE model that incorporates elastic, homogeneous soft tissue around knee bones instead of modelling individual ligaments and cartilage? This approach is inspired by a study by Arjmand, which replaced soft tissue in the proximal tibia with an incompressible cylindrical medium. However, that model did not adequately represent the joint's volume or surface topology. In our study, we propose a simplified FE model where all soft tissues and bony structures are contiguous, maximizing anatomical accuracy. One of the critical steps in creating subject-specific 3D models for FEA is segmentation, which, as our systematic review revealed, suffers from significant variability. Variability in the segmentation process introduces uncertainty into the quantitative data, affecting the reliability of the resulting models. To assess this variability, we conducted inter- and intra-observer variability tests, which are commonly performed in various fields but are notably lacking in the literature for knee joint surgeries. Our secondary aims included determining the intra- and inter-examiner variability in semi-automatic segmentation performed by one operator and 15 operators, respectively. Additionally, we sought to determine the optimal threshold values for knee joint tissues during segmentation, using thresholding techniques. We segmented the tibia at various thresholds and compared the results to a reference tibia segmented at 205 HU. The effect of thresholding proved significant, impacting the final model by causing under- or over-segmentation. The optimal threshold values were identified as 205 HU for the tibia, 160 HU for the femur, 200 HU for the patella, and 232 HU for the fibula. In a pilot study, intra-observer variability was assessed by having one participant segment the knee five times, with the results compared using the Cloud-to-Cloud (C2C) method. The highest similarity (93.39%) was observed between the fourth and fifth segmentations, indicating that operator experience influences the segmentation process. Following ethical approval, 15 volunteers were trained to segment the femur, tibia, and patella five times using ITK-Snap software. Graphical assess intra-observer variability in MATLAB. Inter-observer variability for DSC was calculated using the intraclass correlation coefficient (ICC) in IBM SPSS. The ICC for DSC was 0.975 for the femur, 0.981 for the tibia, and 0.959 for the patella, indicating excellent reliability in the segmentation process. The femur and patella exhibited high DSC and Jaccard Index values, while the tibia had the highest Hausdorff Distance. After confirming the segmentation process’s reliability, we segmented the knee twice more, including the soft tissues, making the model subject-specific. These models were imported into Ansys for FEA, where the soft tissue was modelled as isotropic, homogeneous, and hyperplastic with a neo-Hookean material model (shear modulus: 1 MPa, Poisson's ratio: 0.45). The von Mises strain in the soft tissue following an applied force on the tibia was 1.42 µm for the first knee and 2.43 µm for the second, reflecting a 71% difference. The von Mises stress was 637 Pa and 728 Pa, respectively, showing a 14.2% difference. The articular cartilage experienced the highest stress and strain. Our study successfully simplified the modelling of soft tissue in knee FE models while achieving convergence. The results demonstrated that simulation outcomes are highly sensitive to even minor variations in segmentation. Despite the tibias lower similarity (higher Hausdorff distances), the overall agreement between operators remained consistent. Our findings show good to excellent reliability for segmenting the tibia, patella, and femur in 4D CT images of the knee joint across multiple observers. comparisons were performed using CloudCompare, and quantitative metrics, including Hausdorff Distance, Dice Similarity Coefficient (DSC), and Jaccard Index, were computed to assess intra-observer variability in MATLAB. Inter-observer variability for DSC was calculated using the intraclass correlation coefficient (ICC) in IBM SPSS. The ICC for DSC was 0.975 for the femur, 0.981 for the tibia, and 0.959 for the patella, indicating excellent reliability in the segmentation process. The femur and patella exhibited high DSC and Jaccard Index values, while the tibia had the highest Hausdorff Distance. After confirming the segmentation process’s reliability, we segmented the knee twice more, including the soft tissues, making the model subject-specific. These models were imported into Ansys for FEA, where the soft tissue was modelled as isotropic, homogeneous, and hyperplastic with a neo-Hookean material model (shear modulus: 1 MPa, Poisson's ratio: 0.45). The von Mises strain in the soft tissue following an applied force on the tibia was 1.42 µm for the first knee and 2.43 µm for the second, reflecting a 71% difference. The von Mises stress was 637 Pa and 728 Pa, respectively, showing a 14.2% difference. The articular cartilage experienced the highest stress and strain. Our study successfully simplified the modelling of soft tissue in knee FE models while achieving convergence. The results demonstrated that simulation outcomes are highly sensitive to even minor variations in segmentation. Despite the tibias lower similarity (higher Hausdorff distances), the overall agreement between operators remained consistent. Our findings show good to excellent reliability for segmenting the tibia, patella, and femur in 4D CT images of the knee joint across multiple observers.The knee joint is one of the most complex and weight-bearing joints in the body, making it highly susceptible to injury from various activities. Knee surgery often becomes necessary when conservative treatments fail to alleviate pain and other related disorders. In 2020, research indicated that there were nearly 60,000 total knee replacements (TKRs) for women and approximately 50,000 TKRs for men across England, Wales, Northern Ireland, and the Isle of Man. Projections suggest that by 2060, the demand for hip and knee replacements in the UK will rise by 40%. Robotic knee surgery, a minimally invasive and computer-assisted orthopaedic surgery (CAOS), allows for precise surgical movements, leading to quicker recovery and reduced postoperative pain. However, according to NHS Patient Reported Outcome Measures (PROMs), around 4% of patients in England remain dissatisfied with their knee replacement outcomes, primarily due to implant malalignment. Virtual 3D knee models, generated from CT and MRI scans, play a critical role in improving implant alignment before surgery. These models enable preoperative planning by allowing surgeons to virtually model the patient's knee in 3D, optimizing implant selection and simulating postoperative range of motion. However, the mechanical functionality of the knee joint remains poorly understood, and researchers are actively exploring improvements through finite element analysis (FEA). FEA is a valuable tool for simulating the mechanical behaviour of the knee under various conditions, helping surgeons and biomedical engineers analyse stress distribution, implant stability, and soft tissue interactions. Although existing finite element (FE) knee models provide highly detailed meshes of anatomical structures like bones, cartilage, ligaments, and tendons, these models are complex, time-consuming to create, and prone to human error, making them unsuitable for analysing large image datasets. This brings us to our primary research question: What is the impact of using simplified soft tissue models on finite element simulations of subject-specific knee joints? Can we create an FE model that incorporates elastic, homogeneous soft tissue around knee bones instead of modelling individual ligaments and cartilage? This approach is inspired by a study by Arjmand, which replaced soft tissue in the proximal tibia with an incompressible cylindrical medium. However, that model did not adequately represent the joint's volume or surface topology. In our study, we propose a simplified FE model where all soft tissues and bony structures are contiguous, maximizing anatomical accuracy. One of the critical steps in creating subject-specific 3D models for FEA is segmentation, which, as our systematic review revealed, suffers from significant variability. Variability in the segmentation process introduces uncertainty into the quantitative data, affecting the reliability of the resulting models. To assess this variability, we conducted inter- and intra-observer variability tests, which are commonly performed in various fields but are notably lacking in the literature for knee joint surgeries. Our secondary aims included determining the intra- and inter-examiner variability in semi-automatic segmentation performed by one operator and 15 operators, respectively. Additionally, we sought to determine the optimal threshold values for knee joint tissues during segmentation, using thresholding techniques. We segmented the tibia at various thresholds and compared the results to a reference tibia segmented at 205 HU. The effect of thresholding proved significant, impacting the final model by causing under- or over-segmentation. The optimal threshold values were identified as 205 HU for the tibia, 160 HU for the femur, 200 HU for the patella, and 232 HU for the fibula. In a pilot study, intra-observer variability was assessed by having one participant segment the knee five times, with the results compared using the Cloud-to-Cloud (C2C) method. The highest similarity (93.39%) was observed between the fourth and fifth segmentations, indicating that operator experience influences the segmentation process. Following ethical approval, 15 volunteers were trained to segment the femur, tibia, and patella five times using ITK-Snap software. Graphical assess intra-observer variability in MATLAB. Inter-observer variability for DSC was calculated using the intraclass correlation coefficient (ICC) in IBM SPSS. The ICC for DSC was 0.975 for the femur, 0.981 for the tibia, and 0.959 for the patella, indicating excellent reliability in the segmentation process. The femur and patella exhibited high DSC and Jaccard Index values, while the tibia had the highest Hausdorff Distance. After confirming the segmentation process’s reliability, we segmented the knee twice more, including the soft tissues, making the model subject-specific. These models were imported into Ansys for FEA, where the soft tissue was modelled as isotropic, homogeneous, and hyperplastic with a neo-Hookean material model (shear modulus: 1 MPa, Poisson's ratio: 0.45). The von Mises strain in the soft tissue following an applied force on the tibia was 1.42 µm for the first knee and 2.43 µm for the second, reflecting a 71% difference. The von Mises stress was 637 Pa and 728 Pa, respectively, showing a 14.2% difference. The articular cartilage experienced the highest stress and strain. Our study successfully simplified the modelling of soft tissue in knee FE models while achieving convergence. The results demonstrated that simulation outcomes are highly sensitive to even minor variations in segmentation. Despite the tibias lower similarity (higher Hausdorff distances), the overall agreement between operators remained consistent. Our findings show good to excellent reliability for segmenting the tibia, patella, and femur in 4D CT images of the knee joint across multiple observers. comparisons were performed using CloudCompare, and quantitative metrics, including Hausdorff Distance, Dice Similarity Coefficient (DSC), and Jaccard Index, were computed to assess intra-observer variability in MATLAB. Inter-observer variability for DSC was calculated using the intraclass correlation coefficient (ICC) in IBM SPSS. The ICC for DSC was 0.975 for the femur, 0.981 for the tibia, and 0.959 for the patella, indicating excellent reliability in the segmentation process. The femur and patella exhibited high DSC and Jaccard Index values, while the tibia had the highest Hausdorff Distance. After confirming the segmentation process’s reliability, we segmented the knee twice more, including the soft tissues, making the model subject-specific. These models were imported into Ansys for FEA, where the soft tissue was modelled as isotropic, homogeneous, and hyperplastic with a neo-Hookean material model (shear modulus: 1 MPa, Poisson's ratio: 0.45). The von Mises strain in the soft tissue following an applied force on the tibia was 1.42 µm for the first knee and 2.43 µm for the second, reflecting a 71% difference. The von Mises stress was 637 Pa and 728 Pa, respectively, showing a 14.2% difference. The articular cartilage experienced the highest stress and strain. Our study successfully simplified the modelling of soft tissue in knee FE models while achieving convergence. The results demonstrated that simulation outcomes are highly sensitive to even minor variations in segmentation. Despite the tibias lower similarity (higher Hausdorff distances), the overall agreement between operators remained consistent. Our findings show good to excellent reliability for segmenting the tibia, patella, and femur in 4D CT images of the knee joint across multiple observers
Cognitive pattern recognition models for computational musicology
Full text linkMusic Information Retrieval (MIR) is essential for comprehending and analysing music, and it has various applications in music education, music creation, music recommendation, and other related areas. Conventional music processing techniques heavily depend on human derived characteristics and regulations, which hinders the comprehensive exploration of the abundant information embedded in music. This thesis aims to utilise artificial intelligence approaches, specifically modelling methods rooted in music knowledge and cognition, to address three objectives: automatic music transcription, predominant instrument detection, and music shape evaluation. Automatic music transcription (AMT) is the process of effectively identifying notes from audio signals. Predominant musical instrument recognition (PMIR) involves determining the dominant instrument in a musical section. Music shape evaluation (MSE) shows performance qualities and styles. This thesis introduces a cognitionguided framework for AMT, achieving F-measures of 76.3% on the MAPS dataset (an 8% improvement over the baseline), 80.17% on the BACH10 dataset (second-best performance), and 67.63% on the TRIOS dataset (leading performance). For PMIR, an innovative HHT-DCNN framework is proposed, achieving an 84% F-measure on the IRMAS dataset, which represents a 6% improvement over state-of-the-art methods. Finally, a new dataset is created for the MSE task, and a novel S-ResNN architecture is introduced, achieving an average accuracy of 93.78% across different training ratios. The experimental findings indicate that the suggested approaches may greatly improve current technical standards and achieve outstanding performance. Moreover, the findings of this thesis have the potential to be applied in several aspects of music education, such as the creation of curriculum, the development of interactive learning tools, and the design of personalised music training programmes. This thesis focuses on computational music comprehension and offers substantial contributions to automatic music transcription, instrument recognition, and performance analysis. It highlights the importance and potential applications of research in computational musicology.Music Information Retrieval (MIR) is essential for comprehending and analysing music, and it has various applications in music education, music creation, music recommendation, and other related areas. Conventional music processing techniques heavily depend on human derived characteristics and regulations, which hinders the comprehensive exploration of the abundant information embedded in music. This thesis aims to utilise artificial intelligence approaches, specifically modelling methods rooted in music knowledge and cognition, to address three objectives: automatic music transcription, predominant instrument detection, and music shape evaluation. Automatic music transcription (AMT) is the process of effectively identifying notes from audio signals. Predominant musical instrument recognition (PMIR) involves determining the dominant instrument in a musical section. Music shape evaluation (MSE) shows performance qualities and styles. This thesis introduces a cognitionguided framework for AMT, achieving F-measures of 76.3% on the MAPS dataset (an 8% improvement over the baseline), 80.17% on the BACH10 dataset (second-best performance), and 67.63% on the TRIOS dataset (leading performance). For PMIR, an innovative HHT-DCNN framework is proposed, achieving an 84% F-measure on the IRMAS dataset, which represents a 6% improvement over state-of-the-art methods. Finally, a new dataset is created for the MSE task, and a novel S-ResNN architecture is introduced, achieving an average accuracy of 93.78% across different training ratios. The experimental findings indicate that the suggested approaches may greatly improve current technical standards and achieve outstanding performance. Moreover, the findings of this thesis have the potential to be applied in several aspects of music education, such as the creation of curriculum, the development of interactive learning tools, and the design of personalised music training programmes. This thesis focuses on computational music comprehension and offers substantial contributions to automatic music transcription, instrument recognition, and performance analysis. It highlights the importance and potential applications of research in computational musicology
P2Y1-P2Y12 receptor heterodimers : novel insights from microglial cell signalling and computational approaches
No full textP2Y1 and P2Y12 receptors are class A G protein-coupled receptors and previous studies in the
Kennedy lab suggested that they may physically interact to form functional heterodimers or
possibly higher order oligomers. Little is known, however, about their properties. The aim of
this project, therefore, was to characterise the pharmacological and signalling properties of the
putative heterodimer and to investigate the possible molecular nature of the interaction.
First, Ca2+ imaging was used as a bioassay to characterise the pharmacological properties
of native P2Y receptors in BV-2 microglial cells. The P2Y1 and P2Y12 receptor agonist,
adenosine-5'-diphosphate (ADP) and the selective P2Y1 receptor agonist, MRS2365, evoked
concentration-dependent mobilisation of intracellular Ca2+ stores, which was abolished by the
P2Y1 receptor antagonist, MRS2179. In contrast, P2Y12 receptor antagonists, including ARC69931MX and AZD1283, had no effect against MRS2365, but partially reduced the response
to ADP. Canonically, P2Y1 receptors elicit Ca2+ mobilisation via the Gq/11 G proteins, whereas
P2Y12 receptors couple to Gi/o. Pretreatment with the Gi/o inhibitor, pertussis toxin, had no
effect on the action of MRS2365, but partially inhibited the response to ADP. The inhibitory
action of MRS2179 against both agonists was unaffected by pertussis toxin, but that of
AZD1283 against ADP was abolished. Thus, the Ca2+ mobilisation evoked by ADP in BV-2
cells involves both pertussis toxin-sensitive and -insensitive mechanisms.
The nature of the physical interaction of the receptors was then investigated using the
protein structure prediction tool, AlphaFold2. Initial benchmarking studies modelled
monomeric β2-adrenoceptors, opioid and CB1 cannabinoid receptors accurately and
confidently. Next, AlphaFold2-Multimer modelled monomeric β2-adrenoceptors bound to Gs,
and dimeric GABAB receptors at the Venus flytrap lobes with high confidence, but it was less
successful at modelling the GABAB receptor at the transmembrane regions and opioid receptor
homo- and heterodimers. Finally, monomeric P2Y1 as well as monomeric and homodimeric
P2Y12 receptors were modelled with high confidence, as were the interactions between the
monomeric receptors and G proteins, but the P2Y1-P2Y12 receptor heterodimer was modelled
with lower confidence, both in the absence and presence of G proteins.
These data are consistent with my hypothesis that P2Y1-P2Y12 receptor heterodimers
contribute to ADP-induced release of Ca2+ stores in BV-2 cells. They revealed that pertussis
toxin-sensitive G proteins contribute to this response, whilst the modelling studies provided
insight into the structure of monomeric and homodimeric P2Y1 and P2Y12 receptors and
produced a potential model for the P2Y1-P2Y12 receptor heterodimer. Thus, these studies form
a basis for future studies to investigate the physiological relevance of P2Y1-P2Y12 receptor
heterodimers.P2Y1 and P2Y12 receptors are class A G protein-coupled receptors and previous studies in the
Kennedy lab suggested that they may physically interact to form functional heterodimers or
possibly higher order oligomers. Little is known, however, about their properties. The aim of
this project, therefore, was to characterise the pharmacological and signalling properties of the
putative heterodimer and to investigate the possible molecular nature of the interaction.
First, Ca2+ imaging was used as a bioassay to characterise the pharmacological properties
of native P2Y receptors in BV-2 microglial cells. The P2Y1 and P2Y12 receptor agonist,
adenosine-5'-diphosphate (ADP) and the selective P2Y1 receptor agonist, MRS2365, evoked
concentration-dependent mobilisation of intracellular Ca2+ stores, which was abolished by the
P2Y1 receptor antagonist, MRS2179. In contrast, P2Y12 receptor antagonists, including ARC69931MX and AZD1283, had no effect against MRS2365, but partially reduced the response
to ADP. Canonically, P2Y1 receptors elicit Ca2+ mobilisation via the Gq/11 G proteins, whereas
P2Y12 receptors couple to Gi/o. Pretreatment with the Gi/o inhibitor, pertussis toxin, had no
effect on the action of MRS2365, but partially inhibited the response to ADP. The inhibitory
action of MRS2179 against both agonists was unaffected by pertussis toxin, but that of
AZD1283 against ADP was abolished. Thus, the Ca2+ mobilisation evoked by ADP in BV-2
cells involves both pertussis toxin-sensitive and -insensitive mechanisms.
The nature of the physical interaction of the receptors was then investigated using the
protein structure prediction tool, AlphaFold2. Initial benchmarking studies modelled
monomeric β2-adrenoceptors, opioid and CB1 cannabinoid receptors accurately and
confidently. Next, AlphaFold2-Multimer modelled monomeric β2-adrenoceptors bound to Gs,
and dimeric GABAB receptors at the Venus flytrap lobes with high confidence, but it was less
successful at modelling the GABAB receptor at the transmembrane regions and opioid receptor
homo- and heterodimers. Finally, monomeric P2Y1 as well as monomeric and homodimeric
P2Y12 receptors were modelled with high confidence, as were the interactions between the
monomeric receptors and G proteins, but the P2Y1-P2Y12 receptor heterodimer was modelled
with lower confidence, both in the absence and presence of G proteins.
These data are consistent with my hypothesis that P2Y1-P2Y12 receptor heterodimers
contribute to ADP-induced release of Ca2+ stores in BV-2 cells. They revealed that pertussis
toxin-sensitive G proteins contribute to this response, whilst the modelling studies provided
insight into the structure of monomeric and homodimeric P2Y1 and P2Y12 receptors and
produced a potential model for the P2Y1-P2Y12 receptor heterodimer. Thus, these studies form
a basis for future studies to investigate the physiological relevance of P2Y1-P2Y12 receptor
heterodimers
Modelling of pulsed electric field treatment on microorganisms : transient electric field and forces acting on cell membrane, local thermal effects
Full text linkIrreversible or reversible pores could be generated in the cell membrane of
microorganisms by pulsed electric field (PEF) treatment, which is generally called
electroporation. Such process could be used for inactivation of microorganisms or
bio-medical extraction. Both reversible and irreversible pores can be generated in
bio-membranes this changes the permeabilization of the cell membrane, with the
former allowing for transfection (DNA, RNA, etc.) and the latter for cell inactivation
and bio fuel extraction. However, the PEF treatment is generally considered as ‘nonthermal’ due to the lesser significance of the thermal effect among the treatment
samples. Although PEF is reported to be a non-thermal method, local heating
effects which were not reported before does occur among biological cells during
PEF treatment, different level of thermal excitation will be investigated in this study.
Besides, the exact mechanism between the pulsed electric field and microorganisms
were not fully understood. This study aimed to investigate the interaction between
pulsed electric field and microorganisms, with thermal effects (local heating effects)
also taken into account.
Three different novel analytical models were developed in this study: a linear
model, a QuickField model and a COMSOL model. ‘Hot spots’ (due to local heating
effects) were observed in the models and the characteristics of local heating effects
were also investigated. The contribution of induced electric field strength in cell
membrane and local heating effects were evaluated for electroporation process
during PEF treatment. The results suggest that the significant induced electric field
strength in cell membrane made the main contribution to electroporation.
However, local heating effects could be significant when the treatment samples
were highly conductive. The thermal force and electromagnetic force on the cell
membrane were also investigated. Finally, the situation of penetrated membrane
(pore was included in the cell membrane) was also modelled and it was found that the local heating effects in the penetrated membrane were significant and could
enhance the expansion of pores.
The cell nucleus was also included in the novel QuickField and COMSOL models,
which were used to investigate the interactions between microorganism and
external electric field, both electric field strength in membranes (cell membrane
and nuclear membrane) and thermal effects were investigated. It was observed
that, with nano-second PEF treatment, the induced electric field strength in the cell
nucleus was strong enough to cause electroporation. Thermal effects could also be
generated in cytoplasm.
The experimental works were performed using a self-built HV Blumlein generator.
Different test cells were used to investigate the inactivation process of PEF
treatment with different number of impulses. An alternative plasma treatment was
also implemented to compare the inactivation effects between PEF treatment and
Plasma treatment with the same Blumlein generator. It was found that the plasma
treatment in metallic dish test cell could achieve stronger inactivation compared
with PEF treatment with the same number of impulses.Irreversible or reversible pores could be generated in the cell membrane of
microorganisms by pulsed electric field (PEF) treatment, which is generally called
electroporation. Such process could be used for inactivation of microorganisms or
bio-medical extraction. Both reversible and irreversible pores can be generated in
bio-membranes this changes the permeabilization of the cell membrane, with the
former allowing for transfection (DNA, RNA, etc.) and the latter for cell inactivation
and bio fuel extraction. However, the PEF treatment is generally considered as ‘nonthermal’ due to the lesser significance of the thermal effect among the treatment
samples. Although PEF is reported to be a non-thermal method, local heating
effects which were not reported before does occur among biological cells during
PEF treatment, different level of thermal excitation will be investigated in this study.
Besides, the exact mechanism between the pulsed electric field and microorganisms
were not fully understood. This study aimed to investigate the interaction between
pulsed electric field and microorganisms, with thermal effects (local heating effects)
also taken into account.
Three different novel analytical models were developed in this study: a linear
model, a QuickField model and a COMSOL model. ‘Hot spots’ (due to local heating
effects) were observed in the models and the characteristics of local heating effects
were also investigated. The contribution of induced electric field strength in cell
membrane and local heating effects were evaluated for electroporation process
during PEF treatment. The results suggest that the significant induced electric field
strength in cell membrane made the main contribution to electroporation.
However, local heating effects could be significant when the treatment samples
were highly conductive. The thermal force and electromagnetic force on the cell
membrane were also investigated. Finally, the situation of penetrated membrane
(pore was included in the cell membrane) was also modelled and it was found that the local heating effects in the penetrated membrane were significant and could
enhance the expansion of pores.
The cell nucleus was also included in the novel QuickField and COMSOL models,
which were used to investigate the interactions between microorganism and
external electric field, both electric field strength in membranes (cell membrane
and nuclear membrane) and thermal effects were investigated. It was observed
that, with nano-second PEF treatment, the induced electric field strength in the cell
nucleus was strong enough to cause electroporation. Thermal effects could also be
generated in cytoplasm.
The experimental works were performed using a self-built HV Blumlein generator.
Different test cells were used to investigate the inactivation process of PEF
treatment with different number of impulses. An alternative plasma treatment was
also implemented to compare the inactivation effects between PEF treatment and
Plasma treatment with the same Blumlein generator. It was found that the plasma
treatment in metallic dish test cell could achieve stronger inactivation compared
with PEF treatment with the same number of impulses
Access Repository Dashboard
Do you manage STAX (Strathclyde Repository)? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!