Relevance feedback for enhancing content based image retrieval and automatic prediction of semantic image features: Application to bone tumor radiographs.

•Developed a hybrid CBIR system that aggregates three levels of information – quantitative image features, semantic features, and user feedback.•Propose an approach to predict semantic features of the query image by exploiting the relevance feedback and the quantitative features.•While validating on bone tumor radiographic images, our system achieved mean average precision (MAP) value ∼0.90 where the initial MAP with baseline CBIR was 0.20.•Encouraging results of hybrid CBIR highlight new directions in radiological image interpretation employing semantic CBIR combined with relevance feedback of visual similarity.•Hybrid CBIR has the potential to aid in differential diagnosis of bone tumors by presenting a number of possible lesion types that match the query image.