Vries Hanne S, van Praagh Gijs D, Nienhuis Pieter H, Alic Lejla, Slart Riemer H J A
Department of Nuclear Medicine and Molecular Imaging, University Medical Centre Groningen, University of Groningen, 9700 RB Groningen, The Netherlands.
Department of Magnetic Detection & Imaging, Technical Medical Centre, Faculty of Science and Technology, University of Twente, 7522 NH Enschede, The Netherlands.
Diagnostics (Basel). 2025 Feb 4;15(3):367. doi: 10.3390/diagnostics15030367.
: To investigate the feasibility of a machine learning (ML) model based on radiomic features to identify active giant cell arteritis (GCA) in the aorta and differentiate it from atherosclerosis in follow-up [F]FDG-PET/CT images for therapy monitoring. : To train the ML model, 64 [F]FDG-PET scans of 34 patients with proven GCA and 34 control subjects with type 2 diabetes mellitus were retrospectively included. The aorta was delineated into the ascending, arch, descending, and abdominal aorta. From each segment, 95 features were extracted. All segments were randomly split into a training/validation ( = 192; 80%) and test set ( = 46; 20%). In total, 441 ML models were trained, using combinations of seven feature selection methods, seven classifiers, and nine different numbers of features. The performance was assessed by area under the curve (AUC). The best performing ML model was compared to the clinical report of nuclear medicine physicians in 19 follow-up scans (7 active GCA, 12 inactive GCA). For explainability, an occlusion map was created to illustrate the important regions of the aorta for the decision of the ML model. : The ten-feature model with ANOVA as the feature selector and random forest classifier demonstrated the highest performance (AUC = 0.92 ± 0.01). Compared with the clinical report, this model showed a higher PPV (0.83 vs. 0.80), NPV (0.85 vs. 0.79), and accuracy (0.84 vs. 0.79) in the detection of active GCA in follow-up scans. : The current radiomics ML model was able to identify active GCA and differentiate GCA from atherosclerosis in follow-up [F]FDG-PET/CT scans. This demonstrates the potential of the ML model as a monitoring tool in challenging [F]FDG-PET scans of GCA patients.
为了研究基于放射组学特征的机器学习（ML）模型在[F]FDG-PET/CT随访图像中识别主动脉活动性巨细胞动脉炎（GCA）并将其与动脉粥样硬化相鉴别的可行性，以用于治疗监测。为训练ML模型，回顾性纳入了34例经证实的GCA患者的64次[F]FDG-PET扫描以及34例2型糖尿病对照受试者的扫描。将主动脉划分为升主动脉、主动脉弓、降主动脉和腹主动脉。从每个节段提取95个特征。所有节段随机分为训练/验证集（n = 192；80%）和测试集（n = 46；20%）。总共训练了441个ML模型，使用了七种特征选择方法、七种分类器和九种不同数量特征的组合。通过曲线下面积（AUC）评估性能。将表现最佳的ML模型与19次随访扫描（7例活动性GCA，12例非活动性GCA）中核医学医师的临床报告进行比较。为了便于解释，创建了一个遮挡图来说明主动脉中对ML模型决策重要的区域。以方差分析作为特征选择器和随机森林分类器的十特征模型表现出最高性能（AUC = 0.92±0.01）。与临床报告相比，该模型在随访扫描中检测活动性GCA时显示出更高的阳性预测值（0.83对0.80）、阴性预测值（0.85对0.79）和准确性（0.84对0.79）。当前的放射组学ML模型能够在[F]FDG-PET/CT随访扫描中识别活动性GCA并将GCA与动脉粥样硬化相鉴别。这证明了ML模型作为GCA患者具有挑战性的[F]FDG-PET扫描监测工具的潜力。
