Medical images can be a valuable resource for reliable information to supportmedical diagnosis. However, the large volume of medical images makes itchallenging to retrieve relevant information given a particular scenario. Tosolve this challenge, content-based image retrieval (CBIR) attempts tocharacterize images (or image regions) with invariant content information inorder to facilitate image search. This work presents a feature extractiontechnique for medical images using stacked autoencoders, which encode images tobinary vectors. The technique is applied to the IRMA dataset, a collection of14,410 x-ray images in order to demonstrate the ability of autoencoders toretrieve similar x-rays given test queries. Using IRMA dataset as a benchmark,it was found that stacked autoencoders gave excellent results with a retrievalerror of 376 for 1,733 test images with a compression of 74.61%.
展开▼