Mid Sweden University

In the past three years, deep convolutional neural networks (DCNNs) have achieved promising performance in detecting skin cancer. However, improving the accuracy and efficiency of the automatic detection of melanoma is still urgent due to the visual similarity of benign and malignant dermoscopy. There is also a need for fast and computationally effective systems for mobile applications targeting caregivers and homes. This paper presents the You Only Look Once (Yolo) algorithms, which are based on DCNNs applied to the detection of melanoma. The Yolo algorithms comprise YoloV1, YoloV2, and YoloV3, whose methodology first resets the input image size and then divides the image into several cells. According to the position of the detected object in the cell, the network will try to predict the bounding box of the object and the class confidence score. Our test results indicate that the mean average precision (mAP) of Yolo can exceed 0.82 with a training set of only 200 images, proving that this method has great advantages for detecting melanoma in lightweight system applications.

Deep convolution neural networks (DCNNs) enable effective methods to predict the melanoma classes otherwise found with ultrasonic extraction. However, gathering large datasets in local hospitals in Sweden can take years. Small datasets will result in models with poor accuracy and insufficient generalization ability, which has a great impact on the result. This paper proposes to use a K-Fold cross validation approach based on a DCNN algorithm working on a small sample dataset. The performance of the model is verified via a Vgg16 extracting the features. The experimental results reveal that the model built by the approach proposed in this paper can effectively achieve a better prediction and enhance the accuracy of the model, which proves that K-Fold can achieve better performance on a small skin cancer dataset. 

Deep Convolution Neural Networks (CNN) enable advanced methods to predict the skin cancer classes through the automatic analysis of digital dermoscopic images. However, small datasets' availability often allows the models to be characterized by low prediction accuracy and poor generalization ability, which significantly influences clinical decisions. This paper proposes to use an original ensembling of multiple CNNs as feature extractors able to detect and measure skin lesions atypical criteria according to the well-known diagnostic method 7-Point Check List. The experimental results show that the Artificial Intelligence-based model can suitably manage the classification uncertainty of the single CNNs and finally distinguish melanomas from benignant nevi. Diagnostic performance is promising in terms of sensitivity and specificity towards a decision-supporting system used by a dermatologist with low experience during clinical practice.