Evaluasi tingkat akurasi penggunaan metode deep learning dengan algoritma convolutional neural network pada citra dermatoskopi untuk efisiensi diagnosis kutaneus melanoma: Sebuah telaah literatur sistematis

Abstract

Abstract Background: Cutaneous melanoma (KM) is a malignant cancer on the surface of the skin that requires immediate diagnosis. Deep learning methods using artificial intelligence with various algorithms have been applied to analyze dermatoscopy image results with the aim of making diagnosis more efficient. Purpose: To use deep learning methods with convolutional neural network (CNN) algorithms to accurately identify skin melanoma from dermatoscopy images. Method: Systematic literature review study using the Mendeley.com literature search application. The keywords used are deep learning in cutaneous melanoma. The inclusion criteria in this study are articles published in reputable international journals indexed by Scopus and have been cited, published within the last 5 years, discussing the use of deep learning methods with convolutional neural networks (CNN) algorithms in interpreting dermatoscopy image results with suspected cutaneous melanoma, and include clear accuracy values. Results: There are 15 articles that meet the inclusion criteria, the results show that the accuracy of using deep learning with the CNN algorithm reaches 92%, this accuracy exceeds the accuracy used by dermatologists and other machine learning methods. Conclusion: The application of deep learning methods with the CNN algorithm can produce accurate diagnoses in a short time. This gives hope that it can be integrated into smartphones, so that the diagnosis of skin melanoma becomes more efficient. Keywords: Convolutional Neural Network (CNN); Cutaneous Melanoma; Deep Learning; Dermatoscopic Image.