Image analysis based on an optical model of the skin
for detection of early signs of melanoma
Supported by EPSRC research grant No GR/M53035
- Addenbrooke's Hospital, Cambridge
- Norwich Hospital
- Astron Clinica
This research is concerned with the characterisation of
pigmented skin lesions in the context of early diagnosis of malignant
melanoma. The goal is to improve the diagnostic accuracy beyond that
currently attainable through routine clinical examination and through
existing image analysis methods.
A new skin image analysis method developed by our group generates
parametric images characterising skin structure and composition. A critical
novelty of the method is that it exploits physics of image formation and
uses an optical model of the skin to interpret the colours occurring in a
lesion image in terms of lesion histology. Information contained in the
parametric images computed through the use of the model complements
that available in direct images, whether
examined visually or by conventional image analysis. It is thus likely to
help in lesion discrimination.
Quantitative measures derived from the parametric images will
characterise the amount, location and distribution of melanin and blood and
the architectural distortions induced by lesions, including the presence of
melanin in the dermis often associated with malignant melanoma.
Experimental validation on a large image database will establish whether
the data provided by the new image analysis method improves diagnosis of
- To experimentaly validate the potential of the new skin image analysis
method to discriminate between melanoma and other pigmented skin lesions
- To develop algorithms for deriving quantitative parameters
characterising skin lesions from the images produced by the new method.
- To establish the predictive value of these parameters in the detection of
early signs of melanoma.
- To contribute to the development of a prototype system to be used in a
- To investigate the clinical potential of the new skin image analysis
method through experimental validation on a large set of lesion images
representing a range of clinical problems.
See examples of diagnostic images
produced by the new technique.
Ela Claridge's Home Page
Page maintained by Ela Claridge