The accelerating magnetic resonance imaging (MRI) reconstruction process is a challenging ill-posed inverse problem due to the excessive under-sampling operation in k-space. In this paper, we propose a recurrent Transformer model, namely ReconFormer, for MRI reconstruction, which can iteratively reconstruct high-fidelity magnetic resonance images from highly under-sampled k-space data (e.g., up to 8× acceleration). In particular, the proposed architecture is built upon Recurrent Pyramid Transformer Layers (RPTLs). The core design of the proposed method is Recurrent Scale-wise Attention (RSA), which jointly exploits intrinsic multi-scale information at every architecture unit as well as the dependencies of the deep feature correlation through recurrent states. Moreover, benefiting from its recurrent nature, ReconFormer is lightweight compared to other baselines and only contains 1.1 M trainable parameters. We validate the effectiveness of ReconFormer on multiple datasets with different magnetic resonance sequences and show that it achieves significant improvements over the state-of-the-art methods with better parameter efficiency. The implementation code and pre-trained weights are available at https://github.com/guopengf/ReconFormer.

Purpose: To quantify the accuracy of amide proton transfer-weighted (APTw) MRI for identifying active glioma post-treatment via radiographically guided stereotactic tissue validation. Experimental Design: Twenty-one patients referred for surgery for MRI features concerning for tumor progression versus treatment effect underwent pre-operative APTw imaging. Stereotactic biopsies were taken from regions of interest with varying APTw signal intensities. The relationship between final clinical pathology as well as the histopathology of each of the 64 specimens was analyzed relative to APTw results. Analysis of confirmed recurrent tumor or treatment effect tissue was used to perform receiver-operating-characteristic (ROC) analysis. Results: Eighteen of 21 patients had recurrent tumor, and 3 had treatment effect on clinical pathology. In 12 patients, there were multiple histopathologic assignments confirmed within the same tumor. Of the 64 total specimens, 20 specimens were active glioma, 27 mixed active and quiescent glioma, and 17 quiescent/no identifiable tumor. APTw signal intensity and histopathologic assignment, cellularity, and proliferation index had significant positive correlations (R = 0.651, 0.580, and 0.458, respectively; all P < 0.001). ROC analysis with a 1.79% APTw intensity cutoff differentiated active from non-active tumor (AUC of 0.881) with 85.1% sensitivity and 94.1% specificity. Analysis of clinical pathology showed the mean APTw intensity for each patient had 94.4% sensitivity and 100% positive predictive value for identifying recurrent glioma at this cutoff. Conclusions: APTw imaging hyperintensity may be a marker of active malignant glioma. It is able to distinguish between regions of heterogeneous abnormality on anatomical brain MRI with high sensitivity and specificity.