Medical imaging is an important field of research used for the diagnosis and prediction of diseases. Melanoma is considered as one of the hazardous types of cancers and if detected in early stages, it can be cured easily using simple methods. By using clinical examination, it is difficult to predict melanoma at early stages with high accuracy. This paper proposes a novel strategy for the detection of melanoma by skin malignant growth and also proposes a method for early prediction. The proposed system is based on Deep learning algorithm for the prediction of the affected area and type of melanoma using the metrics precision, accuracy, recall and F1 score. The pre-processing methods are utilized for enhancing the image. The Active contour segmentation process differentiates the infected regions from the healthy skin regions. SOM and CNN classifiers are used for the process of classification of melanoma. A randomly chosen sample of 500 images are taken, 350 images are used as the training dataset and 150 images are used as a testing dataset, for which the proposed system showed high efficiency in the detection of melanoma with a greater accuracy of 90%.

医学成像是用于疾病诊断和预测的重要研究领域。黑色素瘤被认为是癌症的危险类型之一，如果在早期发现，可以使用简单的方法轻松治愈。通过临床检查，很难早期准确地预测黑色素瘤。本文提出了一种通过皮肤恶性生长检测黑色素瘤的新策略，并提出了一种早期预测的方法。所提出的系统基于深度学习算法，使用度量精度，准确性，召回率和F1分数来预测受影响区域和黑色素瘤的类型。预处理方法用于增强图像。主动轮廓分割过程可将感染区域与健康皮肤区域区分开。SOM和CNN分类器用于黑色素瘤的分类过程。随机抽取500幅图像作为样本，将350幅图像用作训练数据集，并将150幅图像用作测试数据集，为此，所提出的系统在检测黑素瘤方面显示出较高的效率，准确率高达90％。