Application of Convolutional Neural Networks in the Automatic Detection of Cutaneous Melanoma

Abstract

Early diagnosis of melanoma is crucial for improving survival rates, which has driven the development of deep learning models for its automated detection. This research aims to evaluate the performance of a convolutional neural network (CNN) in classifying dermoscopic images of skin lesions, comparing its accuracy with that of dermatology experts. To achieve this, a CNN was trained using a set of images that were preprocessed to improve the model's generalization ability. The evaluation was carried out using quality metrics such as accuracy, precision, sensitivity, and F1-score. In addition, the ROC curve and confusion matrix were used to analyze the balance between false positives and false negatives in the classification. The results showed that the CNN outperformed dermatologists in terms of specificity and sensitivity, with an area under the curve (AUC) close to 1, indicating high discriminatory power. The confusion matrix revealed that the classification was correct in most cases, minimizing type I and type II errors. In conclusion, the implementation of neural networks in melanoma diagnosis represents a promising tool for medical care. However, opportunities for improvement were identified, such as adjusting decision thresholds and optimizing image preprocessing, which will increase the accuracy of the model in future clinical applications.