MRI radiomics-based long-term evaluation and identification of imaging biomarkers for growth prediction of distinct nodular lesions in plexiform neurofibromas in NF1

1型神经纤维瘤病（NF1）是一种罕见的常染色体显性遗传肿瘤易感性综合征，由神经纤维蛋白基因缺陷引起。受影响的患者发生周围神经鞘肿瘤的风险很高，其中丛状神经纤维瘤（PNF）和明显结节性病变（DNL）具有特殊的临床意义。PNF可长到很大（约为体重的20%），并可导致患者严重发病。DNL可生长在PNF内或单个病变。DNL的生长引起了向恶性前非典型神经纤维瘤（ANF）转变的关注。与恶性肿瘤不同，ANFs作为恶性神经鞘肿瘤的前体，在先前切除后不会出现局部复发，也不会转移。早期发现这种转化过程对于这些患者的预后至关重要，因此对于NF1患者的风险适应患者护理至关重要。据报道，基于图像的特征有希望指示恶性转化，但很难由人类观察者以一致和客观的方式提取和量化这些成像生物标志物。这一限制可以通过自动标准化提取成像特征来解决，即通过放射组学分析，在NF1患者的纵向图像数据中。然而，先前的放射组学研究仅在小型NF1患者队列中使用单一时间点数据进行横断面研究（N<79）。因此，该项目的目的是利用独特的纵向全身(WB)-MRI数据集，首次采用基于长期放射组学的MRI方法，确定NF1患者PNF中DNL发展和生长预测的成像生物标志物。本项目的纳入标准是根据NIH标准诊断为NF1，并可获得≥2次WB MRI检查。249例患者的WB mri覆盖了18年的观察期（2003 - 2021），可用于长期放射组学研究。WB MRI检查总数约为1.300次（平均每例5.2次）。该项目旨在回答以下问题：(1)NF1患者DNL和PNF的放射组学特征是否不同，是否可以自动分类？(2)不同生长模式（线性与非线性体积生长，高/低生长速率）下DNL和PNF的放射组学特征表达是否不同？(3)放射组学能否作为预后影像学参数预测PNF和DNL的生长？(4)全基因缺陷的NF1患者放射组学特征是否与无主要缺失的NF1患者不同（基因型-表型相关性）？(5)放射组学和遗传信息能否结合起来进一步改善PNF和DNL的发展或生长预测？理想情况下，所研究的放射组学方法将能够识别非侵入性预后成像生物标志物，从而对NF1患者进行个体化风险分层。

Neurofibromatosis type 1 (NF1) is a rare autosomal-dominantly inherited tumor predisposition syndrome caused by defects in the neurofibromin gene. Affected patients have a high risk to develop peripheral nerve sheath tumors, of which plexiform neurofibromas (PNF) and distinct nodular lesions (DNL) are of particular clinical relevance. PNF can grow to large sizes (>20% of body weight) and can cause severe morbidity of patients. DNL can grow inside of PNF or as singular lesions. Growth of DNL raises concern for transformation into a pre-malignant atypical neurofibroma (ANF). Unlike malignant tumors, ANFs as precursors to malignant nerve sheath tumors do not show local recurrence after previous resection nor the ability to metastasize. Early detection of such a transformation process is crucial for the outcome of these patients and therefore essential for risk-adapted patient care of NF1 patients. Image-based features have been reported to be promising to indicate the malignant transformation, but it is difficult to extract and quantify such imaging biomarkers in a consistent and objective manner by human observers. This limitation can be addressed by automated standardized extraction of imaging features, that is, by a radiomics analysis, in longitudinal image data of NF1 patients. However, previous radiomics NF1 studies were only performed as cross-sectional studies using single time-point data in small NF1 patient cohorts (N<79). Therefore, the purpose of the project is to identify imaging biomarkers for the development and growth prediction of DNL in PNF in NF1 patients using the first long-term radiomics-based MRI approach in this rare disease using a unique longitudinal whole-body (WB)-MRI data set. Inclusion criteria for the project are diagnosis of NF1 according to the NIH criteria and availability of ≥ 2 WB MRI examinations. WB MRIs of 249 patients covering an observation period of 18 years (2003 - 2021) are available for the long-term radiomics study. The total number of WB MRI examinations is about 1.300 (corresponding to a mean of 5.2 per patient). The project aims to answer the following questions: (1) Do radiomics signatures of DNL and PNF in NF1 patients differ and can be automatically classified? (2) Do radiomics feature expressions of DNL and PNF with different growth patterns (linear vs. non-linear volume growth, high / low growth rates) differ? (3) Can radiomics predict the growth of PNF and DNL as a prognostic imaging parameter? (4) Do radiomics signatures of NF1 patients with whole gene defects differ from those without major deletions (genotype-phenotype correlation)? (5) Can radiomics and genetic information be combined to further improve PNF and DNL development or growth prediction? Ideally, the investigated radiomics approach will enable the identification of non-invasive prognostic imaging biomarkers towards an individualized risk stratification of NF1 patients.