Upcoming Projects
 |
 |
 |
|
Prospective postgraduate research students may find a list of upcoming projects under the MOR group on this page. Further inquiries can be made to the respective contact persons.
INTELLIGENT MEDICAL IMAGING
Team: Prof. Ir. Dr. Ahmad Fadzil M. Hani, Dr. Aamir S. Malik, Dr. Raja Kamil, Dileep Kumar Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary: Osteoarthritis (OA) is a serious, painful and life altering joint disease that will lead to permanent disability in elders and it is most common in the knee joint. From WHO, knee OA is most prevalent joint disease in developing countries. More alarmingly is an increasing trend of younger people being diagnosed with having OA as early as at age 25. Therefore, early detection of osteoarthritis becomes extremely important as it allows physicians to begin hyaluronic acid treatment and dispensing weight control advice that would help to slow down the onset of OA. It is known that knee OA is due to the gradual loss of articular cartilage (AC). Research has so far shown that AC features that change with OA progression can be categorised into three classes: (a) morphology, (b) mechanical/electrical properties and (c) molecular composition. Detecting features that are associated with osteoarthritis at early stage is challenging and still under research investigations from the perspectives of non-invasive procedure and the use of a single modality. In addition, measurements of a single feature (and changes) vary from one study to another and result in inconclusive early detection of OA. This warrants the use of more than one unique feature to develop a diagnostic tool for early OA. The literature review leads us to believe that by measuring the cartilage thickness, proteoglycan (PG) and water contents can reveal early changes/degeneration, therefore allowing early detection of OA. Various modalities have been used previously to detect the change in articular cartilage features associated with osteoarthritis. In pursuing a single modality approach to detect early OA features, MRI (with qMRI) is the most promising modality, and it is non-invasive, non-ionizing and in vivo. This project aims to develop a non-invasive diagnostic tool for early detection of in vivo knee osteoarthritis. Diagnostic tool will utilize MRI data acquired for morphological and physiological information on articular cartilage features associated with early OA. Furthermore, measured values of change in features associated with early OA will be used to grade the severity of OA at early stages.
Team: Prof. Ir. Dr. Ahmad Fadzil M. Hani, Dr. Mohamed Nordin Zakaria, Assoc. Prof. Dr. Vijanth Sagayan Asirvadam Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary:
Abnormality of melanin production causes skin pigmentation disorders. While there are many treatments for skin pigmentation disorder and for general skin care, changes to skin surface colour as a response to treatment takes time, affecting the ability to assess the efficacy of different treatment modalities. Moreover, the measurement process under Physician’s Global Assessment framework only refers to visual conditions of skin surface. It does not assess the conditions of the underlying skin layers and pigments which are important to the resulting skin tone. The measurement process is not standardised and thus can lead to inter- and intra-rater variability. From the literature, there is no validated in vivo method that is non-ionizing and non-invasive to measure accurately pigment types and quantities to be used in assessing severity of skin pigmentation disorder and treatment efficacies. Presently, the assessment is invasive; skin biopsy is conducted for chemical analysis of skin samples. Although researches on models and simulations of light interaction with human skin have been reported, none has been specifically developed for pigmentation analysis of melanin types. Simulation bio-optics models of light interaction with human skin are used to compute absorption and scattering coefficients of skin layers, to correlate skin colour with skin optical properties, to estimate spectral reflectance of skin, etc. In general, they are used to characterise the structure and properties of skin but they neither classify melanin types nor measure the concentrations of eumelanin and pheomelanin. In this project, we propose a new approach using digital signal and image analysis of multispectral data obtained from skin for skin pigmentation analysis in a clinical setting. The developed technique can be used in the classification and quantification of pigment types.
Team: Dr. Ibrahima Faye, Dr. Samir Brahim-Belhaouari Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary: Early detection and treatment is the key point in reducing the mortality due to cancer. Computer-Aided Diagnosis (CAD) systems are needed to assist radiologists in the challenging task of cancer diagnosis or treatment follow up. Currently, a significant research effort is being devoted to improve the accuracy of the commercial CAD systems for mammography and also to develop CAD systems in other modalities such as PET or Multi-sequence MRI. The key factors to build an efficient system of automatic detection of cancer rely on the combination of an optimal classification method with the most discriminative image features. Recently many advances have been made in characterisation of functions. Some recent multi-scale representations (e.g. curvelets) have been shown to be highly efficient in detecting different types of singularities of functions. An image of cancer can actually be seen as a mathematical function with singularities. On the other hand, the efficiency of a feature extraction method depends on the decomposition of images that was used and also on the nature of abnormalities that are targeted. This project aims to investigate the implementation of a new multi-scale representations (e.g. Curvelet, Wave Atoms) for cancer detection in images. The methodology consists of the decomposition of images in different bases, followed with different methods of feature extraction. It will then implement the best combination of multi-scale representation and feature extraction methods to build highly efficient systems capable of detecting and differentiating different types of cancer in images.
Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary: Leg ulcer is chronic wound that causes severe pain and discomfort to the patients. The first indication of leg ulcer healing is the growth of granulation tissue which reflects changes in ulcer cavity volume. It is important to assess the leg ulcer accurately and precisely to provide the most appropriate treatment and dressing. In clinical practice, assessment of leg ulcers is based on the skill of dermatologist using simple visual inspection. The assessment is very subjective as it may differ from one dermatologist to the other. In a previous research, the emphasis was on the data acquisition method to collect 3D surface scan of leg ulcer wound. Research is carried out using various shapes and volume of the ulcer for validation. Different algorithms for volume computation were applied to the ulcer wound depending on the shape and size of the leg ulcer. The algorithms were able to determine regular-shaped ulcers, but for ulcers covering a large area of the leg and with irregular shapes, a new algorithm is needed for the volume computation. Solid reconstruction used in our algorithms are based on the edges of the wound whereas other researches use the surrounding healthy surface for solid reconstruction. The aim of this project is to develop a system that gives an accurate, reliable and good repeatability in volume measurement and monitoring of the wound treatment efficacy. In this system, the edges of the ulcer wound will be cropped automatically before proceeding with volume determination. The cropped ulcer will be analyzed early and the best algorithm for volume computation according to the shape and size of the ulcer will be chosen. This may result in quality improvement of the treatment given to ulcer patients. NEUROSIGNAL PROCESSING
Team: Assoc. Prof. Dr. Nidal Kamel, Dr. Aamir S. Malik, Dr. M. Zuki Yusoff, Dr. Nasreen Badruddin Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary: With the advent of 3D HDTV in early 2010, more and more households are embracing the 3D technology. In addition to the 3D HDTV, 3D technology entering the consumer market in the last one year include DSLR cameras, mobile phones, monitors, video cameras, game players and blueray disc players. As a result, lot of 3D content is being generated professionally as well as at the amateur level. As there is explosion of 2D multimedia data on the internet in the last decade due to digital cameras and mobile phone cameras, the same can be expected of 3D data when 3D digital cameras and 3D mobile phones become widespread. In education, multimedia tools have played a great role in educating people of all ages. The current multimedia educational tools are based on existing resources which are 2D in nature. However, as the usage of 3D consumer electronics is becoming widespread, it is just a matter of time for 3D educational tools to become available. Therefore, it is the right time to investigate the effects of 3D educational tools on human memory retention and recall processes, using EEG and to compare it with existing 2D educational tools. Research has been reported in the area of traditional multimedia tools using EEG signals. Memory retention and recall processes have been observed using traditional multimedia tools. In this research, we propose to study the effects of 3D educational tools and correlate them with the memory retention and recall processes using EEG signals. This research is the first of its kind and will help in the understanding of the working of the human brain with respect to 2D and 3D educational tools. We plan to design EEG experiments using 3D contents as well as traditional 2D contents. The results of both approaches will be compared and throroughly analysed to determine the more effective educational tool. This will help researchers in the education sector to improve educational and learning tools.
Team: Dr. Aamir S. Malik, Assoc. Prof. Dr. Nidal Kamel, Dr. M. Zuki Yusoff, Dr. Nasreen Badruddin Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary: One in 80 suffers from some sort of epilepsy in the world. However, over 40 types of different epilepsy conditions have been identified which displays a variety of outward reactions which may or may not be very visible, such as seizures, absence-of-seizures, tonic clonic, myoclonic, atonic etc. The epileptic attacks can be controlled through medications especially before the epilepsy attack. Therefore, it becomes extremely important to detect the conditions leading to epilepsy in a predictive manner so that appropriate medical action could be taken. Various modalities including EEG have been used previously to identify abnormal brain activity during epilepsy. In this project, EEG sensors will be used for epileptic monitoring. This project aims to design a mobile EEG monitoring system that can record and assess the abnormal brain activity and predict possible epilepsy attack. The system will include mobile EEG hardware and corresponding epilepsy assessment and prediction algorithms. The sensor array will keep monitoring the conditions and as soon as a known epileptic signature from the brain-activity is detected, appropriate signal is generated and sent to the related clinical staff. Initially, known brain signatures will be used to formulate the algorithm which will be further enhanced by more clinical trials for more realistic signatures from real patient-monitoring.
Team: Dr. Mohd Zuki Yusoff, Dr. Aamir Saeed Malik, Dr. Nidal Kamel Contact Person: Dr. Mohd Zuki Yusoff Summary: Brain Computer Interface (BCI) designs are very useful for motor-disabled (e.g., completely paralyzed) individuals to communicate with external surroundings using their brain thoughts. BCI designs are also suitable for use in simple hands off menu selection on the screen. The main motivation of today's independent BCIs is to develop not only replacement communication and control medium for severely disabled people, but also to explore BCIs as novel input devices for able-bodied subjects in household or automotive applications, entertainment and gaming. In a nutshell, the “direct communication and control channel” of the BCI system and technology can give benefits to the following applications: interface devices between human and computers – mental typewriter; interface devices between human and machines – wheelchair control, household appliance controls, robot arm controls, automobile safety gadgets, video games; bioengineering assist devices for disabled people, etc. EEG-based BCIs utilize some selected aspects of the human brain’s electrical activity; however, this definition does not dictate a specific control methodology that needs to be followed. In other words, despite all advances in brain science experiments attained so far, it is still very little that people know about the EEG behavior and underlying process in the brain. This gap creates opportunities for researchers to explore and investigate further the dynamical properties underlying the EEG in human brains. A lot of efforts have been put to improve the information throughput of BCI systems. Recent literatures report that EEG-based BCI systems information transfer rate is still low, relatively lower than 100 bits per minute. The aims of this research are to improve the overall transfer rate of an independent BCI system, and to neutralize the effects of non-stationary and variable properties of EEG. Also, a focus will be given to mechanisms that enable the human brain to trigger consistent computer commands within a short period of time. The main tasks for this study will be on the acquisition and preprocessing of useful EEG signals from pertinent scalp areas, the discrimination between meaningful and unwanted EEG signals, the careful analysis, correlation and selection of unique features corresponding to thoughts, and on the design and development of a feature detection/translation algorithm. In order to remove unwanted EEG noise and other artefacts, statistical blind signal processing techniques will be employed. The deliverables for the project are thought translation algorithms and novel machine learning schemes that can recognize useful EEG patterns and translate them into control signals.
Team: Dr. Mohd Zuki Yusoff, Dr. Aamir Saeed Malik, Dr. Nidal Kamel Contact Person: Dr. Mohd Zuki Yusoff Summary: Brain Computer Interface (BCI) designs are very useful for motor-disabled (e.g., completely paralyzed) individuals to communicate with external surroundings using their brain thoughts. BCI designs are also suitable for use in simple hands off menu selection on the screen. The main motivation of today's independent BCIs is to develop not only replacement communication and control medium for severely disabled people, but also to explore BCIs as novel input devices for able-bodied subjects in household or automotive applications, entertainment and gaming. In a nutshell, the “direct communication and control channel” of the BCI system and technology can give benefits to the following applications: interface devices between human and computers – mental typewriter; interface devices between human and machines – wheelchair control, household appliance controls, robot arm controls, automobile safety gadgets, video games; bioengineering assist devices for disabled people, etc. EEG-based BCIs utilize some selected aspects of the human brain’s electrical activity; however, this definition does not dictate a specific control methodology that needs to be followed. In other words, despite all advances in brain science experiments attained so far, it is still very little that people know about the EEG behavior and underlying process in the brain. This gap creates opportunities for researchers to explore and investigate further the dynamical properties underlying the EEG in human brains. A lot of efforts have been put to improve the information throughput of BCI systems. Recent literatures report that EEG-based BCI systems information transfer rate is still low, relatively lower than 100 bits per minute. The aims of this research are to improve the overall transfer rate of an independent BCI system, and to neutralize the effects of non-stationary and variable properties of EEG. Also, a focus will be given to mechanisms that enable the human brain to trigger consistent computer commands within a short period of time. The main tasks for this study will be on the acquisition and preprocessing of useful EEG signals from pertinent scalp areas, the discrimination between meaningful and unwanted EEG signals, the careful analysis, correlation and selection of unique features corresponding to thoughts, and on the design and development of a feature detection/translation algorithm. In order to remove unwanted EEG noise and other artefacts, statistical blind signal processing techniques will be employed. The deliverables for the project are thought translation algorithms and novel machine learning schemes that can recognize useful EEG patterns and translate them into control signals. REHABILITATION AND BIOMECHANICS
Team: Assoc. Prof. Dr. Irraivan Elamvazuthi, Dr. M. Zuki Yusoff Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary: The number of lower limb disability amongst the people has been increasing globally in an alarming rate due to war, accidents and disease. Hence, the demand for the production of BK prosthesis is on the increase every year. Medical Implants or popularly known as sockets serve the important role of interfacing the residuum and the BK prosthesis, allowing an individual to not only bear weight comfortably, but to have control over the movement of the prosthesis. The challenges associated with the socket relate primarily to prosthetic fit, which relies in large part on techniques that are employed during fabrication. The socket is the most individual and custom-made part of the prosthesis and its quality has a major influence on the function of the overall prosthesis and on comfort of the user. The most common approach to socket fabrication in both developed and developing countries is to take a plaster wrap cast of the limb residuum, from which a modified plaster model is made. The plastic socket is then formed around this plaster model. However, the drawbacks are that the fabrication of the socket is time consuming and complicated task where it takes on the average of 2 to 6 months to fit a definitive of prosthesis, shortage of skilled technicians leads to delay in getting a prosthetic fitted and unskilled technicians would produce socket of low quality that would compromise on comfort for the patient. The proposed research would yield faster and more consistent production of prosthetic sockets to service a larger number of individuals by using an integrated digital system that comprises of optical 3D digital scanner to acquire the data, CAD software to digitally model the precise shape of the limb, CAE to carry out the analysis and Fused Deposition Modeling (FDM) to fabricate the socket. Potential benefits of this technology include increased quality of socket structure and socket fit, reduction of costs of long distance fittings (due to travel costs), shorter times required with the patient, improved cost-effectiveness compared to conventional techniques, and ease of storing and manipulating socket shapes as needed by clinicians for future use.
Team: Dr Hasan Fawad, Assoc. Prof. Dr. Irraivan Elamvazuthi Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary:
In the standard definition, cardioplegia is the purposeful stopping of the heart using generally cold temperature and chemicals, so that difficult surgeries can be performed on it. During surgery the blood is passed through this cardio-pulmonary device where blood is oxygenated and cooled to maintain the body organs at low temperature. The efficient heat transfer, low volume of blood and reduced pressure drop are the core requirements for an efficient design. Computational Fluid Dynamics has been used effectively in the development and optimization of a blood cardioplegia device, predicting heat transfer and flow characteristics before prototypes are fabricated. Based on numerical simulations, water and blood flow paths in the heat exchanger are optimized for better heat transfer performance with a minimal energy exchange surface.
Team: Assoc. Prof. Dr. Irraivan Elamvazuthi, Patrick Sebastian, Dr. Anis Suhaila Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary:
Cranioplasty is a surgical procedure in which part of the skull is reconstructed. This procedure can be performed for aesthetic or medical reasons, and sometimes a blend of both. For example, it is done due to deformity or defect in the skull, injuries to the skull sustained as a result of trauma or cancers can sometimes cause abnormalities of the skull which need to be corrected. Computed imaging is the fundamental tool in the treatment of various craniofacial disorders in medicine. Critical to the success of craniofacial surgery is the surgeon’s accurate perception of the anatomy of the deformity, which is the foundation for the careful planning, required for proper reconstruction of the defect. Both imaging technologies, computed tomography (CT) and magnetic resonance imaging (MRI), acquire data in a planar fashion, but the subtle 3-dimensional features of complex craniofacial deformities can be difficult to appreciate in 2-dimensional representations. Recent advances in computer aided design and manufacturing have made it possible to capture details of the anatomical structures of the human body that aid surgeons in the reconstructive procedure. The bio-model is represented either in the form of 3-dimensional computer model or in the form of actual physical prototype. The latter has been successfully used as a planning tool guide, in a reconstructive surgery, in neurosurgery, orthopedic surgery and craniofacial surgery. This research is directed at using digital technology (CAD/CAM/RP) in the reconstruction of the complex craniofacial malformations to achieve almost optimal results. Surgery could thus become less invasive and results would be more predictable.
Team: Dr. Hasan Fawad Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary:
Titanium and its alloys are used extensively in bio-medical and aerospace applications because of their excellent properties at room and elevated temperatures, high strength-to-weight ratio and its corrosion resistance. However, the cost of conventional machining titanium-alloy components is extremely high compared to that of machining aluminum-alloy components because of the relatively low machining speed and short tool life. The poor machinability of titanium and its alloys is due to their inherent properties, such as low thermal conductivity leading to high cutting temperature, high hardness at elevated temperature, and low modulus of elasticity. Therefore, conventional machining of titanium alloys is a time consuming and consumable (for example, cutting tool and coolant) intensive process. Thermally enhanced machining, in which an external heat source, such as a laser beam, heats and softens the workpiece locally in front of the cutting tool and allows difficult-to-machine materials to be machined with greater ease, this project will develop a system for Laser assisted machining of Titanium alloy with application to bio-medical implants and also to develop process parameters so as to optimize the production process. PERVASIVE HEALTHCARE
Team: Dr. Fawnizu Azmadi Hussin Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary:
Sudden infant death syndrome is defined as the sudden death of an infant younger than 1 year of age. If the child's death remains unexplained after a formal investigation into the circumstances of the death (including performance of a complete autopsy, examination of the death scene, and review of the clinical history), the death is then attributed to SIDS. Sudden infant death is a tragic event for any parent or caregiver.
SIDS is not predictable, therefore making it difficult for parents or caregivers to take preventive measures. Post birth, the first two leading causes of infant deaths, namely congenital malformations and short gestation and low birth weight, are factors that cannot be controlled. On the other hand, SIDS, which contributes to about 7-10% of the total infant deaths, can be caused by several factors, among others: Infant development: A leading hypothesis is that SIDS may reflect a delay in the development of nerve cells within the brain that are critical to normal heart and lung function. Rebreathing asphyxia: Air movement around the infant’s mouth may be impaired during the early stages of post natal. Improper air circulation around the infant’s mouth may cause deprivation of oxygen causing asphyxia, which can cause death if the condition is prolonged. Apnea (cessation of breathing) of prematurity and apnea of infancy are felt to be clinical conditions that are distinct from SIDS. Infants with apnea may be managed with electronic monitors prescribed by doctors that track heart rate and respiratory activity. Apnea monitors will not prevent SIDS, but can be used to alert caregivers when the condition happens. Hyperthermia (increased temperature): Increasing temperature in infants’ body may lead to an increased metabolic rate in these infants and eventual loss of breathing control. In addition to SIDS, hyperthermia can also cause other complications such as brain damage, etc. These can be measured by measuring the changes in · Heart rate · Body temperature · SpO2 level – blood oxygen level · Respiratory pattern/rhythm · Blood pressure · Motion / sound More research needs to be done to ascertain the correlation between the causes of SIDS and the above body vital signs. In this project, we will develop a wearable body sensor systems for continuous infants’ health monitoring to enable a more proactive and effective healthcare. Continuous monitoring and logging of physiological data can assist medical providers to provide an accurate diagnosis of the root cause of the problems. The continuous monitoring data can be made accessible by the doctors in real time by integrating the BSN to the secure medical network on the Internet.
Team: Dr. Nor Hisham Hamid Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary:
As CMOS transistors scaled down to nano-scale dimensions, they become functionally less reliable. The circuit built using these nano-scale transistors inherently become less reliable. Consequently, to come up with a reliable circuit with nano-scale transistors has become very challenging for circuit designers. Several reliability evaluation schemes have been proposed to estimate circuit’s reliability early in the design cycle. The limitation in the so-far developed reliability-evaluation schemes is that the value of gate-error probability i.e. eg has been arbitrarily assumed for all the techniques that come up under this category. The exact calculation of the value of eg is necessary as it the major parameter used in mathematical models of these techniques and without its exact calculation, the results obtained by the reliability-evaluation schemes does not reflect the actual picture of the circuit’s fault-tolerance capability. In order to calculate the exact value of eg, this parameter has to be modelled according to the physical structure of logic gates that include MOSFETs. A thorough literature review on the MOSFET’s reliability shows that its error-probability depends upon four factors namely gate-oxide reliability, hot-carrier reliability, negative bias temperature instability and power supply noise. Upon modelling these four parameters, we will be able to obtian the actual value of eg, which upon injection into the reliability-evaluation schemes give us the actual reliability of the digital circuit. This research will solve a major bottleneck for circuit designers who want to use reliability-evaluation schemes to calculate the reliability of nano-CMOS technologies that will be introduced in the near-future i.e. below 10 nm.
HEALTH INFORMATICS AND MODELLING
Team: Dr. Brahim Belhaouari Samir, Dr. Ibrahima Faye Contact Person: This e-mail address is being protected from spambots. You need JavaScript enabled to view it Summary: Cancer is the leading cause of death around the globe. It accounted for 7.4 million deaths or 13% of all deaths in 2004. There are many types of cancer that affect the population, however, the top common five types of cancer are breast, colorectal, lung, cervix and nasopharynx. According to the national cancer registry (NCR), 21,773 cases were registered in 2006 affected with cancer. Cancer cause is still unknown to medical professionals; however, if it is detected at the early stages and treated, the number of deaths could be reduced. One of the main approaches that help in early detection of cancer is computer aided diagnosis (CAD) system that assists radiologists in identifying abnormalities in medical images. CAD systems are able to identify cancerous cells in subtle regions that radiologists cannot see them. Although there have been many CAD system, yet, there are still challenges remaining unsolved such as developing better enhancement and segmentation algorithms, designing better feature extraction and selection algorithms and developing high accuracy classifier.. In addition to that, current CAD systems deals with single organ / cancer type and this results in high cost of operation on medical centers. We propose a system for prediction and detection of multi-cancer types. The system provides high detection accuracy due to the techniques that have been utilized to overcome the enhancement and segmentation problem, as well as the feature extraction and selection. In addition to that, the system includes a novel classification algorithm that takes advantage of both the k-Nearest Neighbor (k-NN) which is highly accurate and the K-means cluster which is able to reduce the classification time. We called this technique as Cluster-k-Nearest Neighbor. The system consists of 3 stages namely, organ localization, automatic segmentation and feature selection/extraction and classification. The system provides a detection rate of up to 99%.
|
Powered by Joomla!. Designed by: Joomla 1.5 Template, com.tw domain hosting. Valid XHTML and CSS.