Improving reproducibility in MRI with vendor-neutral acquisitions and transparent workflows

髓磷脂是中枢神经系统的重要组成部分。髓鞘隔离轴突具有三重作用：允许快速电传导，保护轴突，并提供结构支持。从临床角度来看，髓磷脂也具有相关性，因为在多种神经系统疾病（如多发性硬化症）中经常观察到脱髓鞘，定量磁共振成像（qMRI）将单位分配给MR图像，使得通过MRI扫描仪量化髓磷脂含量成为可能。定量MRI难以在临床广泛应用的一个原因是qMRI协议缺乏透明度和标准化。大多数qMR技术都是内部开发的，缺乏适当的框架来部署在不同MRI供应商生产的多个扫描仪上。为了克服这个问题，我们建议开发一个供应商中立的平台，该平台使用开源脉冲序列开发（“Spinbench”），与扫描仪实时交互，以及一个开源图像分析工具箱（qMRLab.org），旨在将定量MRI整合在一个框架下。我们的长期目标是提高qMRI地图的准确性和可靠性，无论是在幻影中还是在体内。我们将首先关注基本的qMRI参数（T1），然后慢慢扩展到其他对髓磷脂敏感的qMRI模式。利用这一框架，我们将(i)开发处理qMRI图的标准化管道；（ii）验证我们在模拟人体组织（幻影）的对象中的管道；（iii）对健康受试者和多发性硬化症（MS）患者的髓磷脂含量进行多供应商MRI研究，表明该方案在健康和病理神经组织中是可重复和临床可行的。MS将被用作概念的证明，因为它的特点可以比较不同类型的髓鞘变性，从局灶性炎症到全局神经变性。通过在不同供应商之间复制髓磷脂测量值，我们将使临床医生能够使用标准化的工具进行诊断、随访和预后。此外，这将使从人工智能（AI）算法中消除依赖供应商的偏见成为可能。最后，我们计划将这个平台推广到其他病理和身体部位，最终建立一个定量MRI的“应用程序商店”，该商店将与供应商无关，开放源代码，并可供全球临床医生使用。

Myelin is a key component of the central nervous system. The myelin sheaths insulate axons with a triple effect: allowing fast electrical conduction, protecting the axon, and providing structural support. Myelin is also relevant from a clinical perspective, given that demyelination is often observed in several neurological diseases such as multiple sclerosis Quantitative magnetic resonance imaging (qMRI) assigns units to MR images, making it possible to quantify myelin content across MRI scanners. The one thing that makes a broad clinical adoption of quantitative MRI difficult is the lack of transparency and standardization in qMRI protocols. Most qMR techniques are developed in-house and lack the proper framework to deploy on multiple scanners manufactured by different MRI vendors. To overcome this problem, we propose to develop a vendor-neutral platform that uses open-source pulse sequence development ("Spinbench"), real-time interaction with the scanner, and an open-source image analysis toolbox (qMRLab.org) that aims to bring quantitative MRI under one umbrella. Our long-term goal is to improve the accuracy and reliability of qMRI maps across vendors, in phantoms and in vivo. We will start by focusing on fundamental qMRI parameters (T1), and slowly expand to incorporate other qMRI modalities sensitive to myelin. Using this framework, we will (i) Develop a standardized pipeline for processing qMRI maps; (ii) Validate our pipeline in objects designed to mimic human tissue (phantoms); (iii) Conduct a multi-vendor MRI study of the myelin content in healthy subjects and multiple sclerosis (MS) patients, showing that the protocols are reproducible and clinically feasible in healthy and pathological nervous tissue. MS will be used as a proof of concept because its characteristics enable the comparison of different types of myelin degeneration, from focal inflammation to global neurodegeneration. By reproducing myelin measurements across vendors, we will make it possible for clinicians to use standardized tools for diagnosis, follow-up, and prognosis. Additionally, this will make it possible to remove vendor-dependent bias from artificial intelligence (AI) algorithms. Finally, we plan to generalize this platform to other pathologies and body parts, eventually building an `App store' for quantitative MRI that will be vendor-agnostic, open-source and accessible to clinicians worldwide.