Deep Learning for Synthetic Postcontrast T1-Weighted MRI: A Systematic Review With Targeted Meta-Analysis of Brain Tumor Studies.

Abstract

<b>BACKGROUND</b>. Gadolinium-based contrast agents remain essential for MRI but carry risks. Deep learning (DL) methods have emerged as a potential approach for synthesizing postcontrast T1-weighted images from precontrast sequences alone. <b>OBJECTIVE</b>. The objective of the present study was to systematically review DL-based synthesis of postcontrast T1-weighted MRI, characterize model architectures and evaluation practices across subspecialties, and perform targeted meta-analysis where sufficient literature existed. <b>EVIDENCE ACQUISITION</b>. A systematic search of PubMed, Embase, Cochrane Central, Scopus, and Web of Science (through January 16, 2025) identified peer-reviewed studies using DL to synthesize postcontrast T1-weighted MRI from precontrast sequences in adults. Two reviewers independently screened studies, extracting data on subspecialty, architecture, quantitative metrics, pathology-specific evaluation, and reader studies. Risk of bias was assessed using a modified version of QUADAS-2. Random-effects meta-analysis was performed for brain tumor studies. <b>EVIDENCE SYNTHESIS</b>. Of 268 records identified after deduplication, 41 met inclusion criteria. Most studies focused on neuroimaging (<i>n</i> = 24; 59%), followed by breast imaging (<i>n</i> = 7; 17%) and body imaging (<i>n</i> = 6; 15%). Generative adversarial networks (<i>n</i> = 20; 45%) and convolutional neural networks (<i>n</i> = 19; 43%) predominated. The structural similarity index measure (SSIM, <i>n</i> = 31; 76%) and peak SNR (PSNR, <i>n</i> = 28; 68%) were the most common metrics. Fifty-one percent (<i>n</i> = 21) of studies performed pathology-specific evaluation, which showed substantially lower SSIM and PSNR compared with whole-image metrics. Thirty-seven percent (<i>n</i> = 15) included reader studies, 29% (<i>n</i> = 12) released code, and 61% (<i>n</i> = 25) used single-institution data. Meta-analysis of 15 brain tumor studies (30 models) yielded pooled SSIM of 0.92 (95% CI, 0.90-0.93) and pooled PSNR of 30.6 dB (95% CI, 28.6-32.6). Given extreme heterogeneity (<i>I</i>² > 99%), pooled estimates should be interpreted as descriptive. <b>CONCLUSION</b>. DL-based postcontrast MRI synthesis shows technical feasibility across subspecialties but suffers from substantial heterogeneity in study design, inconsistent quantitative metric computation, and limited clinical validation. <b>CLINICAL IMPACT</b>. Limited rates of reader studies and external validation represent key barriers to clinical translation of DL-based postcontrast MRI synthesis. Standardized evaluation workflows incorporating whole-image metrics, pathology-specific assessment, and reader studies are essential before these techniques can be translated into clinical practice. <b>TRIAL REGISTRATION</b>. PROSPERO (identifier CRD42025639008).