A Novel Algorithm for Brain Slice Retrieval from Brain MR Image Databases

The paper presents a brain slice retrieval scheme in magnetic resonance (MR) brain volumes acquired from biomedical image sensors. The proposed scheme inputs a 2D brain MR image slice and outputs visually similar slices present in the brain MR volume database. It could be of immense help to the medical practitioners. It might prove to be a useful diagnostic aid and serve as a teaching aid for students in the medical field. Brain MRI artefacts like noise and scanner inhomo-geneities are serious impediments to accurate retrieval. The challenge then is to device an algorithm which is fast, accurate and is robust to variability in tissue intensity values caused by imaging artefacts. To this end, our scheme tries to overcome these hurdles and devise an algorithm which is shown to be robust to inter- and intra-slice intensity variations and geometrical transformations. The edge information of the lesions is extracted via combining multi-resolution methods and spatial methods. The multi-resolution property of wavelets is inherently robust to the present bias fields. The local structure information is then utilised for successful retrieval of similar slices from the brain volumes. The feature point representation of the edge information is shown to be better in comparison to the existing algorithms.