Multi-Class Skin Problem Classification Using Deep Generative Adversarial Network (DGAN)

Abstract

The lack of annotated datasets makes the automatic detection of skin problems very difficult, which is also the case for most other medical applications. The outstanding results achieved by deep learning techniques in developing such applications have improved the diagnostic accuracy. Nevertheless, the performance of these models is heavily dependent on the volume of labelled data used for training, which is unfortunately not available. To address this problem, traditional data augmentation is usually adopted. Recently, the emergence of a generative adversarial network (GAN) seems a more plausible solution, where synthetic images are generated. In this work, we have developed a deep generative adversarial network (DGAN) multi-class classifier, which can generate skin problem images by learning the true data distribution from the available images. Unlike the usual two-class classifier, we have developed a multi-class solution, and to address the class-imbalanced dataset, we have taken images from different datasets available online. One main challenge faced during our development is mainly to improve the stability of the DGAN model during the training phase. To analyse the performance of GAN, we have developed two CNN models in parallel based on the architecture of ResNet50 and VGG16 by augmenting the training datasets using the traditional rotation, flipping, and scaling methods. We have used both labelled and unlabelled data for testing to test the models. DGAN has outperformed the conventional data augmentation by achieving a performance of 91.1% for the unlabelled dataset and 92.3% for the labelled dataset. On the contrary, CNN models with data augmentation have achieved a performance of up to 70.8% for the unlabelled dataset. The outcome of our DGAN confirms the ability of the model to learn from unlabelled datasets and yet produce a good diagnosis result.

An Ensemble of Transfer Learning Models for the Prediction of Skin Lesions with Conditional Generative Adversarial Networks

“…However, data augmentation might be one of two diferent types: picture transformations using the conventional methods or the development of synthetic images using the capabilities of GAN architecture. Several of the earlier researchers successfully implemented the GAN algorithm as a means of image augmentation, and they obtained performance improvements as a result [29,43]. However, the synthetic images that were produced by our CGAN generator and displayed in Figure 4 did not appear to be efective enough to contribute to improved classifcation performance even though a prior work successfully applied the CGAN to produce synthetic skin lesion images using the HAM10000 dataset.…”

“…In the majority of instances, a lack of data or an imbalance of data between classes included in the dataset is the fundamental cause of poor performance. A recent study [29] created a deep generative adversarial network (DGAN) multi-class classifer capable of generating images of skin disorders by learning the distribution of authentic data from publicly available datasets. To handle the class-imbalanced dataset, they used images from several Internet databases.…”

Multi-Scale Attention and Structural Relation Graph for Local Feature Matching

“…Some methods [14], [27], [28], [29] utilize convolution neural networks to learn sparse and local descriptors. Some methods [30], [31], [32], [33] use CNN to make contributions in the field of identification. In particular, the different approaches will each focus on their own concerns.…”

Deep Learning-based Automatic System for Diagnosis and Classification of Skin Dermatoses

“…Generative Adversarial Networks (GANs) are a kind of deep learning technique that may create new data samples that are statistically very similar (Figure 11). To combat the issue of having insufficient data, GAN is used to create high-resolution pictures [11]- [14]. In GAN, the generator and discriminator are two separate models.…”