A HYBRID APPROACH TO SKIN CANCER PREDICTION: INTEGRATING CNN AND HISTOGRAM EQUALIZATION

*Vijai Anand Ramar, Priyadarshini Radhakrishnan, Yashwant Kumar Kolli, Karthik Kushala, Venkataramesh Induru and Thanjaivadivel M.

Skin cancer, and especially melanoma, is the most hazardous and deadly type of cancer requiring early diagnosis to be treated well. This work suggests a hybrid system that incorporates Convolutional Neural Networks together with histogram equalization, image augmentation, and Principal Component Analysis for melanoma detection. Conventional approaches for detecting skin cancer are subject to human error and lengthy diagnostic periods. Our framework enhances image quality by histogram equalization and increases feature diversity in the dataset via image augmentation, enabling the CNN model to learn better and generalize effectively. PCA is applied for feature extraction that minimizes image feature dimensionality while maintaining necessary features for effective classification. The model was assessed using conventional performance measures, which are accuracy (98.65%), precision (92.75%), recall (91.45%), F1-score (94.50%), Kappa (82.36%), and Jaccard index (81.08%), and all of these show its superiority in separating malignant from benign lesions. The experiment results show that the hybrid scheme improves detection precision and is an effective tool for automatic skin cancer diagnosis in the clinic. This framework represents an important step forward in the construction of more accurate, efficient, and scalable melanoma detection frameworks.