Optimal bag-of-features using random salp swarm algorithm for histopathological image analysis

Histopathological image classification is a prominent part of medical image classification. The classification of such images is a challenging task due to the presence of several morphological structures in the tissue images. Recently, bag-of-features method has been used for image classification tasks. However, bag-of-features method uses K-means algorithm to cluster the features, which is a sensitive algorithm towards the initial cluster centres and often traps into the local optima. Therefore, in this work, an efficient bag-of-features histopathological image classification method is presented using a novel variant of salp swarm algorithm termed as random salp swarm algorithm. The efficiency of the proposed variant has been validated against 20 benchmark functions. Further, the performance of the proposed method has been studied on blue histology image dataset and the results are compared with 5 other state-of-the-art meta-heuristic based bag-of-features methods. The experimental results demonstrate that the proposed method surpassed the other considered methods.