Модель мультимодальной радиомики (18F-FAPI PET/CT + CMR) для прогноза AL сердечного амилоидоза (AL-CA)

Light-chain cardiac amyloidosis (AL-CA) is an important cardiac manifestation of systemic light-chain amyloidosis and may lead to progressive myocardial involvement, impaired cardiac function, and poor prognosis. The extent of cardiac involvement and treatment response are important determinants of clinical outcomes in patients with AL-CA. Currently, risk assessment and treatment monitoring mainly rely on cardiac biomarkers, disease staging, and conventional cardiac imaging. However, conventional clinical biomarkers and quantitative imaging parameters may not fully characterize the biological activity, tissue remodeling, and longitudinal changes occurring within the myocardium. Therefore, the development of a non-invasive multimodal imaging approach capable of simultaneously characterizing myocardial biological activity, tissue properties, structural remodeling, and functional changes may improve disease monitoring and prognostic assessment in patients with AL-CA.

This prospective observational study will investigate the value of multimodal imaging integrating [18F]FAPI PET/CT and three-dimensional cardiac magnetic resonance (3D CMR), together with radiomics analysis, for the non-invasive assessment, longitudinal monitoring, and prognostic evaluation of AL-CA. [18F]FAPI PET/CT will provide information related to myocardial fibroblast activation and disease-associated biological activity, whereas 3D CMR will provide complementary information regarding cardiac morphology, function, myocardial tissue characteristics, and fibrosis. By integrating the complementary information provided by these imaging modalities and extracting high-dimensional imaging features that may not be identified by conventional visual assessment, multimodal radiomics may provide a more comprehensive characterization of myocardial involvement and its longitudinal changes in patients with AL-CA.

[18F]FAPI PET/CT imaging will be performed using a Siemens Biograph mCT PET/CT system. Myocardial imaging will be performed approximately 60 minutes after intravenous administration of the [18F]FAPI tracer. Participants will undergo image acquisition in the supine position. PET images will be reconstructed using the TrueX+TOF Ultral HD iterative reconstruction method and post-processed on a Siemens Syngo multimodality workstation. Cedars QPS/QGS software will be used for myocardial image analysis, generating left ventricular vertical long-axis, horizontal long-axis, and short-axis images, as well as short-axis polar maps for the assessment of myocardial tracer uptake characteristics and distribution patterns.

For PET/CT image analysis, volumetric regions of interest will be delineated throughout the left ventricular myocardium from the base to the apex on fused PET/CT images for quantitative and semi-quantitative assessment of myocardial [18F]FAPI uptake. The primary imaging parameters will include the maximum standardized uptake value (SUVmax), mean standardized uptake value (SUVmean), and standardized uptake value ratio (SUVR). SUVR will be defined as the ratio of the SUVmean within the myocardial volume of interest to the SUVmean within a reference volume of interest placed in the descending aorta. Myocardial FAPI uptake volume and uptake distribution patterns will also be assessed. To further quantify the cardiac fibroblast activation protein burden, cardiac fibroblast activation protein volume (CFV) and total cardiac FAP (TCF) will be analyzed. CFV will be calculated based on the volume of myocardial voxels meeting the predefined SUV threshold, and TCF will be derived from myocardial uptake intensity and FAP uptake volume. In the absence of visually apparent abnormal uptake, a standardized region of interest will be used for semi-quantitative assessment.

3D CMR imaging will be performed using a Philips 3T magnetic resonance system equipped with a dedicated cardiac coil. Participants will undergo imaging in the supine position. Balanced steady-state free precession (bSSFP) sequences will be used to acquire left ventricular long-axis and contiguous short-axis images for the assessment of cardiac morphology and function. High-resolution three-dimensional late gadolinium enhancement (LGE) imaging will be performed following gadolinium contrast administration to characterize myocardial tissue abnormalities and fibrosis-related changes. The 3D LGE acquisition will incorporate image navigation (iNAV), compressed sensing (CS), and Dixon water-fat separation techniques to improve spatial resolution and image quality and to reduce the effects of respiratory and other motion-related artifacts.

Motion-tracking and image-reconstruction techniques will be applied during 3D CMR data acquisition and reconstruction to address motion-related effects. High-resolution 3D LGE images will be reconstructed using compressed sensing and iterative reconstruction approaches to generate three-dimensional water-fat-separated myocardial delayed enhancement images. CVI post-processing software will be used to analyze CMR data and derive parameters reflecting cardiac morphology, function, and myocardial tissue characteristics, including ventricular morphological and functional parameters, LGE burden, native T1, extracellular volume (ECV), and other relevant myocardial tissue parameters. These parameters will be used to characterize myocardial structural remodeling, tissue alterations, and fibrosis and will subsequently be incorporated as conventional imaging variables in the multimodal analysis.

Multimodal image fusion will be performed using the Siemens Ture D multimodality post-processing platform. The purpose of image fusion is to achieve spatial registration, alignment, and integration of [18F]FAPI PET/CT and 3D CMR datasets, thereby combining information on myocardial biological activity and tracer uptake obtained from PET/CT with structural, functional, and tissue characterization obtained from CMR. Differences in spatial resolution, patient positioning, and cardiac motion between imaging modalities will be addressed through image registration. Manual adjustment based on left ventricular myocardial regions of interest will be performed when necessary. The fused datasets will be used to investigate the relationships among myocardial FAPI uptake, tissue characteristics, fibrosis, and structural remodeling.

Radiomics analysis will be performed using Siemens FeAture Explorer (FAE, version 0.5.13). High-dimensional radiomics features will be extracted from PET/CT, 3D CMR, and fused multimodal imaging datasets following standardized image preprocessing and region-of-interest segmentation. Extracted features will include first-order statistical features, morphological features, texture features, gray-level co-occurrence matrix (GLCM) features, gray-level run-length matrix (GLRLM) features, wavelet features, and other relevant image descriptors. Radiomics analysis will further quantify myocardial signal intensity, spatial heterogeneity, texture distribution, and tissue structural characteristics, thereby complementing the information provided by conventional quantitative imaging parameters.

Radiomics features will be integrated with quantitative parameters obtained from PET/CT and CMR, including myocardial FAPI uptake volume and distribution patterns, SUVmax, SUVmean, SUVR, CFV, TCF, LGE burden, native T1, and ECV. Relevant clinical, laboratory, and pathological variables will also be incorporated as appropriate for the planned analyses. The integration of multimodal radiomics features, conventional imaging parameters, and clinical variables will enable the construction of a comprehensive dataset representing different biological and structural dimensions of myocardial involvement in AL-CA.

LASSO-based methods will be used for feature selection to reduce redundancy in high-dimensional radiomics data and identify key variables associated with prognosis. Selected multimodal radiomics features, conventional quantitative imaging parameters, and clinical variables will subsequently be incorporated into predictive models. Machine learning approaches will be used for model development according to the characteristics of the available data and study objectives, and Cox proportional hazards regression models will be used for survival and prognostic analyses. These models will be used to evaluate the incremental predictive value of multimodal imaging features beyond conventional clinical indicators and standard imaging parameters.

Longitudinal multimodal imaging data obtained during follow-up will be analyzed to evaluate dynamic changes in myocardial FAPI uptake, CMR-derived tissue characteristics, and radiomics features. Changes in imaging characteristics between baseline and follow-up examinations will be analyzed in relation to clinical response, disease progression, and survival outcomes. This longitudinal analysis will explore the potential value of multimodal imaging for dynamic monitoring of disease activity and myocardial changes in patients with AL-CA.

The predictive performance of the multimodal radiomics models will also be compared with conventional clinical and imaging-based assessment methods. Comparator approaches will include traditional clinical biomarkers, disease staging, and quantitative imaging parameters derived from PET/CT or CMR alone. By evaluating and comparing the predictive performance of different approaches, this study will investigate whether the integration of [18F]FAPI PET/CT, 3D CMR, multimodal image fusion, and radiomics analysis can provide additional prognostic information.

The overall objective of this study is to develop and evaluate a non-invasive imaging-based approach integrating [18F]FAPI PET/CT and 3D CMR multimodal imaging. The study will explore the potential value of radiomics and integrated predictive models for assessing cardiac involvement, monitoring longitudinal disease changes, and evaluating prognosis in patients with AL-CA. The findings may provide more comprehensive and individualized imaging information for patients with AL-CA and offer additional imaging evidence for treatment response monitoring and clinical risk stratification.