Table 10.

Machine learning and deep learning approaches.

References	Year	Approach	Result
[63]	2020	Two preprocessing methods and an automatic segmentation method based on semi-supervised learning are provided for usage with the offered dermoscopy images.	Deep learning techniques will be used in future research to improve the accuracy of the calculated coefficient values.
[64]	2018	This paper creates a novel approach for diagnosing lesions using deep learning and a localized feature encoding system.	Generate various feature values to work among a high amount of variation of lesions.
[39]	2019	In this article, we investigate the convolution neural network, a deep learning system that uses dermoscopy images to predict small skin changes.	The model’s performance is validated using images of lesions and the area under the curve for a lesion.
[65]	2019	Therefore, the manual analysis will have less room for interpretation and bias. Skin images may be automatically examined for melanoma and differentiated lesions using a deep learning method for working with lesion patterns.	Convolutional neural networks such as Inception-v4, ResNet-152 and DenseNet-161 were used to classify images of melanoma and skin discoloration plasma. To generate lesion segmentation masks, two U-Nets were trained.
[66]	2018	In the HAM10000 collection of dermoscopy images, skin cancer is recognized using the CNN model as an identifier.	RNN and other deep learning algorithms provided the most accurate cancer diagnoses, according to the results.
[67]	2019	An automatic system to improve the performance of classification for the efficient diagnosis of melanoma.	Results are reported for both segmented and non-segmented picture classifications. To expand the scope of the work, probabilistic graphical models can be added to this network.
[68]	2022	This paper offers a new judgement system known as an NN classifier that may more accurately diagnose skin lesions using deep learning methods such as neural networks and feature-based algorithms.	The stage of classification is implemented using SVM. The results obtained by the proposed system cover higher accuracy.
[69]	2019	The ABCD rule, GLCM and HOG algorithms are described as being used for feature extraction. Utilizing geodesic active shape, the lesion is separated to provide access to its features.	We categorized items with a sensitivity of 97.8% and a specificity of 0.94 using SVM classifiers. KNN’s application produced sensitivity and specificity of 86.2% and 85.0%, respectively.
[69]	2019	The majority of research is focused on separating melanoma-causing lesions from those that are not.	Our system of classifying skin tumors was shown to be accurate. Our results imply that doctors might be able to use the suggested technique to identify AM early.