Automatic Skin Cancer Detection in Dermoscopy Images Based on Ensemble Lightweight Deep Learning Network

Abstract

The complex detection background and lesion features make the automatic detection of dermoscopy image lesions face many challenges. The previous solutions mainly focus on using larger and more complex models to improve the accuracy of detection, there is a lack of research on significant intraclass differences and inter-class similarity of lesion features. At the same time, the larger model size also brings challenges to further algorithm application; In this paper, we proposed a lightweight skin cancer recognition model with feature discrimination based on fine-grained classification principle. The propose model includes two common feature extraction modules of lesion classification network and a feature discrimination network. Firstly, two sets of training samples (positive and negative sample pairs) are input into the feature extraction module (Lightweight CNN) of the recognition model. Then, two sets of feature vectors output from the feature extraction module are used to train the two classification networks and feature discrimination networks of the recognition model at the same time, and the model fusion strategy is applied to further improve the performance of the model, the proposed recognition method can extract more discriminative lesion features and improve the recognition performance of the model in a small amount of model parameters; In addition, based on the feature extraction module of the proposed recognition model, U-Net architecture, and migration training strategy, we build a lightweight semantic segmentation model of lesion area of dermoscopy image, which can achieve high precision lesion area segmentation end-to-end without complicated image preprocessing operation; The performance of our approach was appraised through widespread experiments comparative and feature visualization analysis, the outcome indicates that the proposed method has better performance than the start-of-the-art deep learning-based approach on the ISBI 2016 skin lesion analysis towards melanoma detection challenge dataset. INDEX TERMS Dermoscopy Images, Skin cancer detection, Lightweight deep learning network, Finegrained feature.

Machine Learning and Deep Learning Methods for Skin Lesion Classification and Diagnosis: A Systematic Review

“…To control the imbalanced datasets, they utilized a loss function based on the Jaccard distance and the focal loss. Wei et al [191] proposed a skin lesion recognition system based on fine-grained classification to discriminate features. A different lightweight CNN was utilized for segmentation and classification.…”

Skin Lesions Classification Using Deep Learning Techniques: Review

“…Another method to extract features based on deep learning [92], which use Encoder-decoder Fully Convolutional Network (FCN) method, is suitable for small volumes of input data and requires only a few parameters, making the method easy to interpret. Also, up to 73% of the reviewed skin lesion classification methods were based on the deep learning algorithm (KNN), where these models based on transfer learning like GoogleNet and Inception-v3 in [72], DenseNet 201 in [73], AlexNet in [79,80] Inception v3, InceptionResNet-v2, and ResNet 152 in [84], fusion MobileNet and DenseNet in [85], DenseNet in [8]. The best accuracy was obtained among them in [86] GoogelNet 99.29%, in [79] AlexNet 98.61%, and the DenseNet area under curve AUC is 98.16%.…”

“…The model achieves a 93.5 percent accuracy that is higher than previous approaches. WEI et al [85] proposed a lightweight skin cancer detection model in image classification using fine-grained classification theory based on MobileNet and DenseNet. In this model, a lesion classification network and a feature discrimination network are made up of feature extraction.…”

Skin Lesion Classification Based on Deep Convolutional Neural Networks Architectures

“…Employing a sophisticated model with a large computation overhead might render challenges to applicability in the real world. Motivated by using a less sophisticated model, researchers in [73]. Put forth a discriminant dermoscopy image lesion recognition model.…”