Optimization of the Convolutional Neural Networks for Automatic Detection of Skin Cancer

Abstract

<jats:title>Abstract</jats:title><jats:p>Convolutional neural networks (CNNs) are a branch of deep learning which have been turned into one of the popular methods in different applications, especially medical imaging. One of the significant applications in this category is to help specialists make an early detection of skin cancer in dermoscopy and can reduce mortality rate. However, there are a lot of reasons that affect system diagnosis accuracy. In recent years, the utilization of computer-aided technology for this purpose has been turned into an interesting category for scientists. In this research, a meta-heuristic optimized CNN classifier is applied for pre-trained network models for visual datasets with the purpose of classifying skin cancer images. However there are different methods about optimizing the learning step of neural networks, and there are few studies about the deep learning based neural networks and their applications. In the present work, a new approach based on whale optimization algorithm is utilized for optimizing the weight and biases in the CNN models. The new method is then compared with 10 popular classifiers on two skin cancer datasets including DermIS Digital Database Dermquest Database. Experimental results show that the use of this optimized method performs with better accuracy than other classification methods.</jats:p>

Comparison of machine learning strategies for infrared thermography of skin cancer

“…For traditional ML, the suggested ensemble showed an ACC and SP of 89.69% and 94.44%., close to [25] and [26]. Still, sensitivity is a metric to improve, as it is the most important indicator of the capability of an AI system to detect those ill. For DL (ACC=96.91%, SN=94.12%, SP=98.41%), Hosny et al presented a more balanced distribution of metrics (ACC=97.7%, SN=97.34%, SP=97.34%) using a transfer learning approach, while the optimization strategy of Zhang et al [29] yield them higher sensitivity (95%) with a lower accuracy and specificity (91% and 92%) (Table 9). It is important to refer most of these studies possessed an elevated number of samples when compared to the ones included here.…”

“…Despite its inherent complexity, new optimization ideas and applications are constantly being studied. Hosny et al [28] applied theory of transfer learning to a convolutional neural network and replaced the classification layer with a softmax layer (9350 images, ACC=97.7%, SN=97.34%, SP=97.34%), while Zhang et al [29] developed a meta-heuristic optimization algorithm to deal with biases and distribution of weights during training of a Convolutional Neural Network (CNN) for diagnosis of melanocytic lesions (>20000 images, ACC=91%, SN=95%, SP=92). It is worth mentioning that the vast majority of authors chooses to test its AI strategies with dermoscopy images available from public databases, e.g., PH2, ISBI, DermIS, Dermquest, to ease the comparison of achieved results with those of other authors.…”

Optimized Training for Convolutional Neural Network Using Enhanced Grey Wolf Optimization Algorithm

“…In [11], to improve the recognition accuracy by optimizing the results of the solution vectors on CNN, the training process was supported by the PSO algorithm. In [17], an approach based on whale optimization algorithm (WOA) is utilized for optimizing the weight and biases in the CNN model. In [38], an integrated CNN with a genetic algorithm was proposed to learn the structure of deep neural networks.…”

Classification Models for Skin Tumor Detection Using Texture Analysis in Medical Images

“…The diagnosis of skin lesions was studied by Zhang [ 7 ]. The analysis considered Convolutional Neural Networks (CNN) for automatic detection of skin cancer, comparing this with other research methods.…”