Simultaneous Super-Resolution and Cross-Modality Synthesis in Magnetic Resonance Imaging

Multi-modality magnetic resonance imaging (MRI) has enabled significant progress to both clinical diagnosis and medical research. Applications range from differential diagnosis to novel insights into disease mechanisms and phenotypes. However, there exist many practical scenarios where acquiring high-quality multi-modality MRI is restricted, for instance, owing to limited scanning time. This imposes constraints on multi-modality MRI processing tools, e.g., segmentation and registration. Such limitations are not only recurrent in prospective data acquisition but also when dealing with existing databases with either missing or low-quality imaging data. In this work, we explore the problem of synthesizing high-resolution images corresponding to one MRI modality from a low-resolution image of another MRI modality of the same subject. This is achieved by introducing the cross-modality dictionary learning scheme and a patch-based globally redundant model based on sparse representations. We use high-frequency multi-modality image features to train dictionary pairs, which are robust, compact, and correlated in this multimodal feature space. A feature clustering step is integrated into the reconstruction framework speeding up the search involved in the reconstruction process. Images are partitioned into a set of overlapping patches to maintain the consistency between neighboring pixels and increase speed further. Extensive experimental validations on two multi-modality databases of real brain MR images show that the proposed method outperforms state-of-the-art algorithms in two challenging tasks: image super-resolution and simultaneous SR and cross-modality synthesis. Our method was assessed on both healthy subjects and patients suffering from schizophrenia with excellent results.