Computer-aided diagnosis of skin lesions
Malignant melanoma is nowadays one of the leading cancer causes among many white-skinned populations. Fortunately the curability of skin cancer is very high, if it is treated surgically: in thin melanomas (< 1mm) the cure rate is between 92% and 98%, whereas in thicker tumors or advanced disease, prognosis gets worse and treatment remains palliative. Consequently, in order to reduce mortality from melanomas, the prevention campaign in the population at risk should also provide the diagnose of melanomas in their early stage (i.e. with a tumor thickness below 1 mm) through the identification of precursors based on advances in skin imaging technology.
Epiluminescence microscopy (ELM, also known as dermatoscopy or dermoscopy) has become an established noninvasive tool to improve the early detection of melanoma and other skin cancers while reducing unnecessary excisions of benign lesions. Through the adoption of different incident light magnification systems with an oil immersion technique, dermoscopy allows the clinician to recognize numerous features (submacroscopic morphologic structures as well as vascular patterns) located in the different compartments of the skin to distinguish benign and malignant skin tumors.
Based on the interpretation of the features inspected by dermoscopy, three different diagnostic models have become more widely accepted by clinicians: i) Pattern analysis, ii) the ABCD-rule of dermatoscopy, based on a semi-quantitative assessment of the following dermoscopic criteria: Asymmetry , Border, Color and Different structures; iii) the ELM 7-point checklist which is a scoring diagnosis analysis, defining only seven standard ELM criteria: Atypical pigment network. Blue-whitish veil, Atypical vascular pattern, Irregular streaks, Irregular pigmentation, Irregular dots/globules, Regression structures. Compared with the clinical diagnosis by the naked eye, there is an improvement of diagnostic accuracy (10% to 30% higher sensitivity) when using dermoscopy in skin tumors. However, due to the complexity of patterns and their interpretation, the results of dermoscopic examination have still limitations especially for beginners and users not trained specifically.
Starting from these considerations, the authors have tackled the problem of defining suitable image processing algorithm implementing the 7-Point Check List. More in details, the research project has been carrying out the development of an automatic system able to validate the identification of 7 parameters on each lesion observed by clinicians. To this aim, after a preliminary study about the image processing techniques for the extraction of the pigmented lesion (from healty skin) and the detection of chromatic features, further algorithms were introduced as suitable means for the detection of morphological structures correlated with the criteria defined in diagnostic method of interest.
- Involved researchers: Giuseppe Di Leo, Consolatina Liguori, Alfredo Paolillo, Antonio Pietrosanto, Paolo Sommella
- Collaborations: University "Federico II" of Naples, Italy
