A Hybrid Segmentation Method BCOT for Skin Lesion Detection

Skin cancer is a widespread disease that has the potential to be life-threatening, making early detection decisive for effective treatment. The biopsy method is the formal approach employed for diagnosing skin cancer, which is highly uncomfortable and a time-intensive procedure. This research work proposes a comprehensive approach to the detection of skin cancer, integrating advanced digital image processing techniques for easy identification of cancer. A Multi-filter Fusion and Equalization technique (MFE) is proposed in the pre-processing step to enhance the image quality by removing noise and highlighting relevant features in Digital Dermoscopic (DD) images. A boundary contrast-based Otsu threshold method (BCOT) is proposed for segmentation, ensuring precise delineation of lesion boundaries. In the feature extraction phase, the grey-level co-occurrence matrix (GLCM) method is deployed to capture relevant textural information. GLCM captures details about how intensity values are distributed with each other. To classify various types of skin cancers, a Support Vector Machine (SVM) is leveraged in the classification step for differentiation among cancer types. The proposed method isolates the given image into malignant melanoma or benign melanoma. The experimental results are obtained based on the dataset, which is available on the public database comprising 600 images of malignant type and benign type diseases sourced from the International Skin Imaging Collaboration (ISIC) 2020. Sixty percent of the images were allocaied for training the model, while the remaining forty percent were designated for testing purposes. The presented approach attains a precision level of 97.92%. The software used for this study is MATLAB R2023b version