Image analysis in the RGB and HS colour planes for a computer-assisted diagnosis of cutaneous pigmented lesions
| Autor: | Gabrina Tragni, Renato Marchesini, Cesare Bartoli, Stefano Tomatis, Bruno Farina, Aldo Bono |
|---|---|
| Rok vydání: | 1998 |
| Předmět: |
Male
Cancer Research Pathology medicine.medical_specialty Skin Neoplasms genetic structures Visual descriptors Color Statistics Nonparametric 030218 nuclear medicine & medical imaging Diagnosis Differential 03 medical and health sciences 0302 clinical medicine medicine Image Processing Computer-Assisted Humans Melanoma Nevus Mathematics Hue Color representation business.industry Image (category theory) Pattern recognition General Medicine Color model Oncology 030220 oncology & carcinogenesis Variegation (histology) RGB color model Benign nevus Female Artificial intelligence business |
| Zdroj: | Tumori. 84(1) |
| ISSN: | 0300-8916 |
| Popis: | Aims and background A study was carried out to evaluate the effectiveness of image analysis performed by the two color representation models when a computer-assisted diagnosis of melanoma is involved. Methods Color images of 40 skin pigmented lesions, which included 12 melanomas, were acquired by a standard color RGB video camera and stored in a PC for off-line processing. Image analysis was performed in the red green and blue color representation model and using hue and saturation color components. To describe shape and color characteristics of each lesion, including area, roundness and color variegation, 16 parameters were derived from red, green, blue, hue and saturation color planes and tested as possible variables useful to differentiate melanomas from benign nevi. Results The test gave a result of significance for six of the 16 derived image descriptors. The general trend of our data was in agreement with clinical observations according to which melanoma is usually darker, more variegated and less round than a benign nevus, whereas lesion dimension of melanomas and benign lesions was not significantly different. Conclusions Our preliminary results suggested that image analysis performed on hue and saturation-derived and red green and blue-derived data could better discriminate melanoma from nevi than separately using the two color representation models. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|