Skin cancer has been one of the most common forms of cancer around the world. Due to the efficiency and high accuracy of artificial intelligence, more and more hospitals are using it to assist doctors in finding out cancer quickly. In our work, we trained a convolutional neural network with a dataset called ‘Skin Cancer ISIC’ to detect appearances of 9 different kinds of skin cancers. Firstly, we trained a convolutional neural network model with the original data from the dataset. It contains several convolutional and max pooling layers, and its accuracy achieved 45%. Although the classification of 9 species will undoubtedly decrease the accuracy of the model compared to making fewer species, 45% is not an acceptable value for medical judgment. To improve the accuracy of our model, we used data enhancement and retrained our model. During the training of the dataset, we find out that the pictures of the dataset are not in the same standard. The brightness, contrast, size, and shape of the images are different, which increased the difficulty of learning. By rotating pictures, adjust brightness, equalizing histogram, adding random noise, flip and adding random USM, we reach an accuracy of 60%. Moreover, we employed ResNet-50, a pre-training model, as the convolutional neural network model to further improve our accuracy. The final testing result gets an accuracy over 65 %, which is a huge improvement from the beginning.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.