Abstract:

Microfluidic systems are widely used for application in biological assays as they provide many advantages. However, patterning of microbeads in a microfluidic device has not been done before, especially with a specific aim of enabling multiplex biological assays to be performed in the device.

In this project, bead based bioassays were performed using encoded microbeads on a microfluidic device. The microfluidic device was fabricated using soft lithography techniques to obtain very unique done-shape structures . The structures allowed the encoded beads to be patterned orderly unto the device, facilitating individual analysis of beads and enabling bioassays to be multiplexed.

The protocols to efficiently pattern the microbeads unto the array in the device and performing bead based bioassays were optimized. This led to reduced total analysis time and the reagent volume used in the experiments. Also, 3D computer simulation software FLUENT and GAMBIT were used to study the effect of the device design on the patterning of microbeads and how changes to the design may affect the patterning of the encoded microbeads.

All these leads to enhance the advantages of microfluidic device and provides many potential applications in the area of cell-based, protein, DNA and RNA assays.

Abstract:

Abstract:

Water pollution, caused by toxic heavy metals, has been a major cause of concern due to their potential environmental and human health impacts. As a consequence, various environmental protection agencies have imposed more stringent environmental legislations on wastewater discharges. Removal of toxic metals from contaminated water samples through adsorption is considered to be a practical and effective treatment technology. However the commercially available adsorbents are expensive. In recent years, the search for low-cost adsorbents that have good metal-binding capacities has intesified.

 

 

 

 

 

 

Abstract:

This project is a novel method to recognize human facial expressions using a 2D morphable model. In this method, a reference face model, is first synthesized, on which key feature points associated with expression-related deformations are marked (e.g. eye and lip boundaries). “Feature-based image metamorphosis” is used to mathematically measure feature point deformations from the neutral face to all the expressive faces in a database of 3D expressive faces and to train the system. As the facial deformations associated with the different expressions are distinct, the morphing parameters associated with each expression are unique. These morphing parameters are used for expression recognition and classification and we demonstrate that the morphing parameters associated with various expressions can be distinctly clustered in the expression space. Expression recognition for novel faces is then performed by measuring the displacement of the key feature points from the neutral face to any novel expressive face and comparing the morphing parameters so obtained with those of the prototypical expressions. The system was found to successfully classify four (Angry, Sad, Happy, Surprise) out of the six prototypical expressions (Angry, Sad, Happy, Surprise, Fear, Disgust) with an average successful recognition rate of 84.88% for supervised learning and 78.5% for unsupervised learning. The proposed approach is less expensive than 3D-based morphing in terms of computational time and resources required and can also be used for pose-invariant expression recognition when a 3D face database is used. Also, unlike the traditional Active Appearance Model-based approach where feature points need to be marked for all faces, our approach requires marking of feature points only on the reference neutral face model.