Artículo de revisión

Artificial intelligence for skin lesion classification and diagnosis in dermatology: A narrative review

Back to article
Types of artificial intelligence [8].
AI typeCharacteristicsApplicationsLimitations
Deep learningUses deep neural networks to solve complex problemsPattern recognition in images, natural language processing, and gamingRequires substantial amounts of data, computationally intensive
Neural networksComputational model inspired by the human brain, composed of interconnected nodesSpeech recognition, pattern recognition, image processingRequires extensive training, complexity in interpretation
Reinforcement learningAn agent learns through interaction with an environment and receives feedback in the form of rewards or penaltiesGaming, robotics, and autonomous decision-makingRequires extended training time, design complexity
Feedforward neural networksInformation flows in one direction, without cycles or loopsPattern recognition in images, natural language processingDoes not handle sequential data well, such as time series
Recurrent neural networksInclude cyclical connections to handle sequential data and retain information over timeNatural language processing, automatic translation, and time seriesMay face issues of vanishing or exploding gradients
Convolutional neural networksSpecifically designed for image processing, with shared weight layers and pooling layersObject recognition in images, and medical diagnosis based on imagesRequires enormous amounts of training data, specialized hardware
Decision treesTree-shaped models that make decisions based on features at each nodeMedical diagnosis, customer classification, decision-makingMay be prone to overfitting, difficulty capturing complex relationships
Random forestsAn ensemble method that combines multiple decision trees to improve accuracyClassification, regression, anomaly detectionComplexity in interpretation, resource-intensive
Support vector machinesClassifies data points by finding the hyperplane that best separates classesText classification, speech recognition, bioinformaticsMay not perform well on large datasets, sensitive to the choice of kernel
Logistic regressionLinear model used for binary classification problemsRisk prediction, disease classification, and marketingLimited to binary classification problems, does not handle nonlinear relationships

Source: prepared by the authors based on the results of the study.