School of Computer Science

Projects

Image analysis based on an optical model of the skin for detection of early signs of melanoma

Funded by the EPSRC

This research is concerned with the characterisation of pigmented skin lesions to help with early diagnosis of malignant melanoma, a skin cancer. Our group has developed a novel image analysis method which uses physics-based modelling of optical properties of the skin. The method computes parametric maps characterising skin structure and composition. The images show histological quantities in the skin, such as concentration of pigment melanin, concentration of blood and thickness of collagenous tissue. They also show whether melanin is present in the dermis - such presence is a very sensitive indicator of melanoma. SIAscope is a clinical device based on this research, developed by Astron Clinica and used in dermatology clinics in the UK and beyond.

Further Information


Image interpretation via material specific spectral characterisation models

Funded by the Leverhulme Trust

The goal of the project is to formulate a generic approach to image interpretation based on the spectral characterisation models. This should enable the structure and composition of the materials and tissues to be deduced from their images acquired through a small number of optical filters using a standard digital camera.

Further information


Physics-based image interpretation to aid the detection of early signs of retinopathies

Funded by the EPSRC

This project aims to develop and validate a new physics based image interpretation method for the ocular fundus and to evaluate its potential in detecting early signs of diabetic retinopathies. An understanding of the physical interaction of light with ocular tissue is utilised to formulate a mathematical model capable of predicting colours which correspond to different tissue composition. Colours in digitized clinical images will be interpreted through reference to this model, generating "retinal maps" which show separately the quantities and distribution of blood, retinal pigment and pathological exudates at every image point.

Further information


Model Based Algorithms for Structural and Functional Dual Modality MRI and Optical Imaging

Funded by the EPSRC

In this proposed work, a novel imaging technique will be explored that uses non-harmful application of near infrared light to determine the properties of biological tissue. There are two distinct applications that are of major interest: (1) the imaging of the female breast for detection and characterisation of cancer and (2) small animal imaging for the understanding of basic brain physiology and tumour development and treatment. Under the proposed program a strong collaboration will be established between three international institutions whereby experimental and clinical data will be collected by the collaborators. The data, not readily available in UK, will be then used under this program to develop, explore and identify the best possible numerical models and reconstruction algorithms to provide the most clinically useful images. This will enable the utilisation of the best current instrumentation and computation expertise worldwide to gain knowledge and understanding for the detection and characterisation of disease, while at the same time cementing and bridging a gap of knowledge and expertise in a much needed area of research.


High Density diffuse optical tomography for mapping human brain function

Funded by NIH, USA

This project is in collaboration with our partners at University of Washington in St Louis, USA. The overall goal of this project is to develop optical tomography methods for mapping resting state functional connectivity in order to study the childhood development of brain functions.


NIRFAST

Funded by NIH, USA

This work originates from over a decade of development in near-infrared spec troscopy and imaging at Dartmouth College, USA, and has the goal of formalizing software tool s and systems which allow integration of Near Infrared Spectroscopy (NIRS) into conventional imaging modalities. Hybrid image-guided NIRS systems are being developed with MRI and CT scanners, and commercialization efforts are ongoing to develop commercially viable prototypes to potentially satisfy the imaging demands presented by emerging molecular medicine paradigms. This work will specifically advance the spectral imaging recovery software which has been developed at Dartmouth, called NIRFAST (Near-Infrared, Fluorescence, Absorption and Spectral Tomography). Ultimately, we will develop and provide a standardized platform which will be made available to the research community worldwide to aid in the advancement of the clinical utility.


Automatic detection of blood deprived regions in parametric images of the skin

With Francesca Satta, Florence University

A clinical study using the SIAscope images for diagnosis of malignant melanoma has shown that the presence of blood depravation regions within the lesion is strongly associated with malignancy. This project has developed a computer method for automatic detection of the blood deprived regions. The results of the computer method compared to clinical assessment show very good agreement, with 91% sensitivity and 96% specificity on the set of 95 lesions.


Modelling of edge profiles in pigmented skin lesions

The sharpness of the lesion boundary and the contrast between the lesion and the surrounding skin provide important diagnostic information in the assessment of pigmented skin lesions. This project has developed a new method for computing these parameters by employing an edge model based on a sigmoid function. For each point on the lesion boundary, optimal parameters are found by using an interative least-squares method. One of the parameters returned is the location equivalent to "zero crossing" for each boundary profile. A collection of these points demarkates the lesion boundary.


Symmetry analysis

Whereas most algorithms for symmetry computation are designed to find the best axis of symmetry, in this work we are interested in finding the degree of symmetry for the lesion pattern. Instead of choosing the best symmetry axis, a number of putative symmetry axes are considered and scored. The variability of these scores characterizes the overall symmetry of the lesion. Symmetry scores can be computed for both the lesion shape and for the pattern of pigmentation within the lesion. The latter is computed by removing low-frequency variations underlying the transition from the lesion body to the skin outside.


Completed projects

  • Colour and pigmentation analysis in skin lesions
    with Symon Cotton and Per Hall (Adenbrooke's Hospital, Cambridge)
  • Characterisation of visual features in skin lesions
    with Jon Morris Smith and Per Hall (Adenbrooke's Hospital, Cambridge)
  • Low-level boundary grouping mechanisms for contour completion
    with Alison Todman, School of Computer Science
  • Characterisation of mammographic lesions
    with Josef Richter, School of Computer Science
  • Computational and psychophysical approaches to contour perception
    with School of Psychology and Hamburg University
  • Image registration
    with Mutawarra Hussain, School of Computer Science
  • Medical workstation for neuromedicine
    with Queen Elizabeth Hospital, Birmingham