Radiomics and machine learning assessment of multiparametric FDG-PET/MRI for evaluation of prediction of early treatment response to immune checkpoint therapy in patients with Non-Small Cell Lung Cancer (NSCLC).

通过免疫检查点抑制疗法(ICI-TX)治疗癌症的新范式显示，实体肿瘤患者的总体生存期和无进展生存期都有显著改善。在具有PD1疗效的肿瘤中，包括肺癌、头颈部、胃癌和膀胱癌，总生存率都有了巨大的、前所未见的增长。然而，只有大约20%的患者被证明有反应，而80%的患者未能受益。虽然与传统治疗相比，癌症治疗已经提升到了更高的成功水平，但治疗监测的成像仍然局限于基本的传统方法，未能评估早期治疗反应，使有效和无反应的人能够及早和充分地区分，导致急需的治疗改变的延迟以及无效和不必要的治疗成本。在过去的几年里，一些研究证明了对接受化疗/放射治疗的肺癌患者的形态和代谢变化进行联合评估的好处。研究结果强调了PET/CT参数在预测无进展存活率、总存活率以及识别有治疗失败风险的患者方面的预测能力，从而使早期治疗调整成为可能。最近对18F-FDG PET/CT早期预测晚期黑色素瘤患者免疫检查点治疗反应的研究强调了代谢和形态参数联合评估ICI-TX的预测能力。虽然PET/CT能够对代谢和形态参数进行综合分析，但同时的多参数PET/MR成像将基于血流灌注和功能MR参数的潜在成像生物标志物的评估提升到了更高的水平。同步PET/MRI已被证明为基于图像的肿瘤生物学分析提供了一个强大的多参数成像平台，潜在地识别肿瘤的异质性和表型，以及在疾病进展过程中演变的治疗耐药性。这些参数包括肿瘤大小或体积的简单评估、其几何形状、肿瘤纹理分析、肿瘤细胞密度和血管形成、组织氧合、某些代谢物的检测和肿瘤代谢的定量评估。这种综合考虑肿瘤表型的大量定量图像特征的方法通常被称为放射组学。由于放射组学分析的本质，在收集大量数据时，需要机器学习算法来进行计算识别和特征提取。因此，我们研究的目的是评估放射组学和机器学习引导的多参数FDG-PET/MRI在预测早期治疗反应中的作用。

The new paradigm of cancer treatment, by means of immune checkpoint inhibition therapy (ICI-TX), has revealed a significant improvement in overall survival and progression-free survival of patients with solid tumors. Large, unseen-of gains in overall survival have been demonstrated in tumors with proven PD1 efficacy, comprising lung, head and neck, gastric as well as bladder cancer. Nevertheless, only approximately 20 % of patients have been shown to respond, while 80 % of patients fail to benefit. While cancer treatment has been elevated to superior success levels when compared to conventional treatment, imaging for therapy monitoring is still restricted to basic conventional methods, failing to assess early treatment response to enable early and sufficient differentiation between responders and non-responders causing delays in much-needed treatment changes as well as ineffective and unnecessary treatment costs. Over the past few years a number of studies have demonstrated the benefit of the combined assessment of morphological and metabolic changes in patients with lung cancer undergoing chemo-/ radiation therapy. The study results underline the predictive power of PET/CT parameters for prediction of progression-free survival, overall survival as well as identification of patients at risk of treatment failure, enabling early treatment adjustment. Recent investigations on 18F-FDG PET/CT for early prediction of response to immune checkpoint therapy in patients with advanced melanoma have underlined the predictive power of the combined assessment of metabolic and morphologic parameters for ICI-TX assessment. While PET/CT enables the combined analysis of metabolic and morphologic parameters, simultaneous multiparametric PET/MR imaging uplifts the assessment of potential imaging biomarkers to a multitude based on the exploration of perfusion and functional MR parameters. Simultaneous PET/MRI has been shown to provide a powerful multiparametric imaging platform for image-based profiling of tumor biology, potentially identifying tumor heterogeneity and phenotypes as well as evolving therapy resistance in disease progression. These parameters range from simple assessment of tumor size or volume, its geometric shape, tumor texture analysis, tumor cellularity and vascularization to tissue oxygenation, detection of certain metabolites and quantitative assessment of tumor metabolism. This kind of comprehensive consideration of a large number of quantitative image features for tumor phenotyping is often referred to as radiomics. Bound to the nature of radiomics analysis in collecting an extensive volume of data machine learning algorithms are needed for computational identification and feature extraction. Hence, the aim of our study is to evaluate radiomics and machine learning-guided assessment of multiparametric FDG-PET/MRI for prediction of early treatment respons