Geometrical Modelling (GM Group)

Research Areas

Developing fundamental algorithms for processing 3D/4D mesh models, especially for human anatomical model in healthcare applications.

The reconstruction of geometrical model from images necessitate the generation of 3D meshes that can accurately capture the original shape of the object without distortion. Furthermore, if these 3D meshes are to be used for computational analysis, there must be a quality check on the individual elements within the mesh. The Geometrical Modelling (GM) team have developed extensive capabilities in the area of 3D geometry reconstruction using different imaging modalities such as MRI and CT. In additional, we are also able to capture and track the spatial and temporal motions embedded in the image dataset to create 4D (3D + time) meshes. These 4D meshes are guaranteed to have superior mesh quality with 1-to-1 point correspondence, and is suitable for advanced computational processing, such as shape analysis, motion analysis and finite element modelling. The inputs to the method are multiple 3D mesh models of the different frames/phases of the cardiac cycle. These models have no initial correspondence in terms of number of vertices/points and mesh connectivity.

Establish shape analytics methodologies for analysing mesh properties of 3D/4D anatomical digital models to extract clinical insights.

Heart failure (HF) imposes major global health care burden on society and suffering for the individual. As such, the early diagnosis and identification of the underlying etiology of HF is of paramount importance; some causes require specific treatment and may be correctable. Clinically, diverse imaging tests such as echocardiography, nuclear scintigraphy and magnetic resonance imaging (MRI) are used for diagnosis and prognostication. MRI is arguably the most accurate and reproducible. The Geometrical Modelling (GM) team have pioneered the concept of using shape and geometry reconstructed from MRI images as a descriptor of local heart functions that does not require a frame of reference in contrast to other measures. Furthermore, the team has also developed a computer-aided diagnostic software “Cardiowerkz” that provides a suite of indices for assessing cardiac health by combining MRI and advanced computation methodology. Currently, the team is working with the National Heart Centre Singapore to run clinical trials using “Cardiowerkz” to evaluate the feasibility of using these indices for the clinical diagnostics of HF patients. If successful, these indices could aid in risk stratification and clinical management of HF patients and open up a new paradigm for diagnosis and monitoring of HF patients.    

Utilising image processing capabilities for extraction of useful domain specific indicators on biological-related image slices.



Digital pathology is a disruptive technology in diagnostic and anatomic pathology that encompasses image acquisition of histological slides, image analysis, storage and display management, and will allow academicians and medical professionals to access large databases of digitized image information. Digital pathology promises to overhaul the daily workflow and activities in diagnostic labs faced with a shortage of qualified pathologists, increasing workloads, and demands for greater efficiency. The digitization of histology slides creates opportunities for quantitative analysis through computer-based image processing for disease detection, diagnosis and prognosis to complement the opinion of the pathologist. Advanced image processing and machine learning algorithms are required to assist in cancer grading and differentiation between cancer subtypes. The Geometrical Modelling (GM) team has extensive experience in the development of computer-aided diagnosis system to address problems in real clinical scenarios. Besides developing algorithms for image analysis and value-adding insight generation, the team also has expertise in using analytics for complex analysis of biological images, such as feature classification for evaluation of phenotypic signatures, allowing the team to address the challenges of image analysis for high throughput screening of histology images.