Skin Lesion Classification Using Hybrid Deep Neural Networks

Mahbod, Amirreza; Schaefer, Gerald; Wang, Chunliang; Ecker, Rupert; Ellinge, Isabella

doi:10.1109/icassp.2019.8683352

Cited by 210 publications

(128 citation statements)

References 21 publications

Supporting

Mentioning

127

Contrasting

Unclassified

Order By: Relevance

“…The main drawback of conventional approaches is a lack of generalisation capability due to high variations in dermoscopic images, different artefacts and insufficient training data. Variations in dermoscopic images are due to different zooming configurations, lighting conditions, instruments or operators, while common artifacts in dermoscopic images include not just skin hair and bubbles but also, among others, dark corners/borders, light reflections or shadows, skin lines, ruler or calibration chart artefacts or ink markings, which can lead to failures of segmentation algorithms, changes in extracted image features and consequently a negative effect on classification accuracy [32,16].…”

Section: Introductionmentioning

confidence: 99%

“…The pre-trained models used in both approaches for skin lesion classification varied in different studies and include AlexNet [39,32,41], VGG16 [32,27,43], VGG19 [32], GoogleNet [44,45,43], ResNet-50 [43,46,47], ResNet-101 [48], ResNet-152 [49,50] Inception-v3 [42,51], Inception-v4 [48,49], variations of DenseNets [31,49], SeNets [31,50] and PolyNets [31]. Moreover, ensembles of fine-tuned deep networks [48,46] and fusing outputs of classical and deep models [41,40] were utilised to boost classification performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fusing fine-tuned deep features for skin lesion classification

Mahbod

Schaefer

Ellinger

et al. 2019

Computerized Medical Imaging and Graphics

Self Cite

197

120

View full text Add to dashboard Cite

Malignant melanoma is one of the most aggressive forms of skin cancer. Early detection is important as it significantly improves survival rates. Consequently, accurate discrimination of malignant skin lesions from benign lesions such as seborrheic keratoses or benign nevi is crucial, while accurate computerised classification of skin lesion images is of great interest to support diagnosis. In this paper, we propose a fully automatic computerised method to classify skin lesions from dermoscopic images. Our approach ensembles deep features from several well-established convolutional neural networks (CNNs) at different abstraction levels in combination with a support vector machine classifier to distinguish malignant melanomas from benign lesions. Importantly, the CNNs are pretrained on a common natural image database and then fine-tuned on a limited set of dermoscopic skin lesion images. Finally, prediction probability classification vectors obtained from different models with different training settings are fused to provide improved classification performance. Evaluated on the 600 test images of the ISIC 2017 skin lesion classification challenge, the proposed algorithm yields an area under receiver operating characteristic curve (AUC) of

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Fusing fine-tuned deep features for skin lesion classification

Mahbod

Schaefer

Ellinger

et al. 2019

Computerized Medical Imaging and Graphics

Self Cite

197

120

View full text Add to dashboard Cite

show abstract

“…After a huge success of deep learning, deep learning models are also used for the classification of skin lesions. Using pre-trained deep learning models AlexNet [13], VGG16 [14] and ResNet-18 [15]; a fully automated classification scheme was proposed by Mahbod et al [16]. In the proposed classification scheme, features were generated using AlexNet, VGG16 and ResNet-18 and passed to SVM classifier for final classification.…”

Section: Literature Reviewmentioning

confidence: 99%

Mobile-based Skin Lesions Classification Using Convolution Neural Network

Hameed¹,

Shabut²,

Hameed³

et al. 2020

AETiC

View full text Add to dashboard Cite

This research work is aimed at investing skin lesions classification problem using Convolution Neural Network (CNN) using cloud-server architecture. Using the cloud services and CNN, a real-time mobile-enabled skin lesions classification expert system “i-Rash” is proposed and developed. i-Rash aimed at early diagnosis of acne, eczema and psoriasis at remote locations. The classification model used in the “i-Rash” is developed using the CNN model “SqueezeNet”. The transfer learning approach is used for training the classification model and model is trained and tested on 1856 images. The benefit of using SqueezeNet results in a limited size of the trained model i.e. only 3 MB. For classifying new image, cloud-based architecture is used, and the trained model is deployed on a server. A new image is classified in fractions of seconds with overall accuracy, sensitivity and specificity of 97.21%, 94.42% and 98.14% respectively. i-Rash can serve in initial classification of skin lesions, hence, can play a very important role early classification of skin lesions for people living in remote areas.

show abstract

“…The evaluations result of the presented method are compared with traditional low‐level features are showed improved performance. Mahbod, Ecker, and Ellinger () introduced a new fully automated method for skin lesion classification using pretrained CNN features. The extracted features are classified by ensemble learning and showed improved performance on international skin imaging collaboration (ISIC) 2017 dataset.…”

Section: Related Workmentioning

confidence: 99%

Construction of saliency map and hybrid set of features for efficient segmentation and classification of skin lesion

Khan

Akram

Sharif

et al. 2019

Microscopy Res & Technique

View full text Add to dashboard Cite

Skin cancer is being a most deadly type of cancers which have grown extensively worldwide from the last decade. For an accurate detection and classification of melanoma, several measures should be considered which include, contrast stretching, irregularity measurement, selection of most optimal features, and so forth. A poor contrast of lesion affects the segmentation accuracy and also increases classification error. To overcome this problem, an efficient model for accurate border detection and classification is presented. The proposed model improves the segmentation accuracy in its preprocessing phase, utilizing contrast enhancement of lesion area compared to the background. The enhanced 2D blue channel is selected for the construction of saliency map, at the end of which threshold function produces the binary image. In addition, particle swarm optimization (PSO) based segmentation is also utilized for accurate border detection and refinement. Few selected features including shape, texture, local, and global are also extracted which are later selected based on genetic algorithm with an advantage of identifying the fittest chromosome. Finally, optimized features are later fed into the support vector machine (SVM) for classification. Comprehensive experiments have been carried out on three datasets named as PH2, ISBI2016, and ISIC (i.e., ISIC MSK‐1, ISIC MSK‐2, and ISIC UDA). The improved accuracy of 97.9, 99.1, 98.4, and 93.8%, respectively obtained for each dataset. The SVM outperforms on the selected dataset in terms of sensitivity, precision rate, accuracy, and FNR. Furthermore, the selection method outperforms and successfully removed the redundant features.

show abstract

Skin Lesion Classification Using Hybrid Deep Neural Networks

Cited by 210 publications

References 21 publications

Fusing fine-tuned deep features for skin lesion classification

Fusing fine-tuned deep features for skin lesion classification

Mobile-based Skin Lesions Classification Using Convolution Neural Network

Construction of saliency map and hybrid set of features for efficient segmentation and classification of skin lesion

Contact Info

Product

Resources

About