PhD-student: Kajsa Møllersen

Project description

When a dermatologist diagnoses a skin lesion (benign/malignant) she/he looks at the lesion and the surrounding skin through a dermoscope (magnifying lens surrounded by polarized light sources). The data of this project are digital images of skin lesions taken through a dermoscope. These images are then to be analyzed by a computer programme that will give a diagnosis.

The project is to construct this programme. The programme will not be made first and foremost for dermatologists, but for general practitioners, because it is the GP that the patient meets first.

The program will consist of four main parts:

1. Segmentation: Segmenting the pixels of the image into two groups; lesion and skin
2. Feature extraction: Detecting the features assumed to discriminate benign from malignant lesions
3. Feature selection: Selecting the features that in fact discriminate between benign and malignant lesions
4. Classification: Choosing the combination of features that gives the best result, i.e. that detects all malignant lesions while sending as few benign lesions as possible to further examination

The final goal for this project is a cheap and easy-to-use tool that can assist GPs (and dermatologists) in diagnosing skin lesions. The program is not meant to replace the GP, but in collaboration with the GP to increase the number of malignant lesions detected at an early stage, and decrease the number of benign lesions excised and sent to pathological examination. 

Contact information

Kajsa Møllersen, telephone +47 97 78 39 40 and email kajsa.mollersen@telemed.no