Optimized vision transformer encoder with cnn for automatic psoriasis disease detection

Psoriasis is a skin disorder that results in swollen skin cells and red, itchy areas on the skin. 40% of the world's population is currently affected by psoriasis. Nowadays, using skin image analysis technology is the main way for detecting psoriasis. Additionally, a number of academics have identified potential machine learning methods for categorising the psoriasis illness. However, the accuracy and computational efficiency of the model still need to be improved. Thus, in this paper, we present an optimized vision transformer for autonomous psoriasis disease detection. Following pre-processing, feature optimized image is attained using convolutional neural network (CNN) which embeds full image and concatenates to each vision transformer encoder layer. It leads the network to always "retain" the full image at the end of each transformer block output. In parallel, the pre-processed images are cropped into patches and these patches along with its positional encoded information are given as input to the optimized transformer encoder. To enhance the performance of transformer, the hyper-parameters of it are optimized using adaptive rabbit optimization algorithm (AROA). Results of this article confirm that the proposed optimized vision transformer model achieved better classification accuracy of 97.7% and F-Score of 96.5%.