Hierarchical approach for the classification of multi-class skin lesions based on deep convolutional neural networks

Samia Benyahia (1) , Boudjelal Meftah (1) , Olivier Lézoray (2)

1 University of Mascara - Department of Computer Science
2 Equipe Image - Laboratoire GREYC - UMR6072

Skin lesion is one of the most critical challenges nowadays due to the difficulty of distinguishing a benign lesion from a malignant one. Melanoma represents a malignant melanocytic type of cancer among the most dangerous ones. In contrast, basal cell carcinoma and squamous cell carcinoma represent no malignant melanocytic types of cancer that threaten many human lives. Fortunately, there is some possibility of a cure it if is early detected and well treated. Currently, dermatologists use a hierarchical visual categorization of the lesion or skin biopsy for the diagnostic of skin lesion types. However, computer-aided detection methods can be more accurate, faster, and less expensive than human based techniques. We propose to combine both strategies to develop an efficient skin lesion classification models: the hierarchical organization accredited by dermatologists and a deep learning architecture. In this work, we propose a new hierarchical model for detecting various types of skin lesions based on the combination of several models of convolutional neural networks, where each model is specialized to some types of skin lesion according to the taxonomy. The obtained results highlight the benefits of addressing the classification of different skin lesions with CNNs in such a hierarchically structured way.

Mots clés: Classification; Convolutional Neural Networks; Hierarchical Diagnosis; Skin Lesion Diagnosis

Domaines: Traitement des images [eess.IV]

Samia Benyahia, Boudjelal Meftah, Olivier Lézoray. Hierarchical approach for the classification of multi-class skin lesions based on deep convolutional neural networks. ICPRAI, Jun 2022, Paris, France. ⟨10.1007/978-3-031-09282-4_12⟩. ⟨hal-03688422⟩