Hybrid Xception InceptionV3 Model for Skin Disease Classification

Abstract

Skin cancer ranks as a highly perilous and widespread illness worldwide, necessitating prompt and precise diagnosis to enable successful treatment. Deep learning (DL), especially convolutional neural networks (CNNs), has exhibited notable efficacy in medical image classification tasks. This research introduces a hybrid DL technique that integrates Xception (XCP) and InceptionV3 (INCP3) models for classifying multiple skin disease categories. The model is trained and validated with data from the HAM10000 and ISIC repositories, comprising a broad spectrum of lesion types. To mitigate dataset imbalance and enhance generalizability, the proposed approach utilizes data augmentation, class balancing, and transfer learning. To avoid overfitting, the Adam optimizer is used along with dropout regularization, early stopping, and learning rate scheduling. The experimental results reveal that the model secures a training accuracy of 98.89% and a validation accuracy of 88.43%, surpassing standalone CNNs. Evaluation metrics like precision, recall, F1-score, ROC-AUC, and confusion matrices validate its capability in correctly classifying diverse skin lesions. These findings underscore the practicality of employing hybrid DL models in computer-assisted skin cancer diagnosis, with future investigations centring on explainable AI techniques and integration into clinical settings.

Ähnliche Arbeiten