术前准确、灵敏地区分良恶性甲状腺结节的诊断刻不容缓。然而现有的临床方法并不能令人满意地解决这个问题。本研究的目的是为东部人群建立一种简单、经济的术前诊断方法。我们的回顾性研究包括 86 例甲状腺乳头状癌患者和 29 例良性病例。使用ITK-SNAP软件绘制感兴趣区域（ROI）的轮廓，并使用Ultrosomics提取放射组学特征。采用全转录组测序和生物信息学分析来鉴定甲状腺结节诊断的候选基因。 RT-qPCR用于评估候选基因的表达水平。基于METLAB 2022平台和LibSVM 3.2语言包建立SVM诊断模型。首先建立放射组学模型。准确度为 73.0%，灵敏度为 86.1%，特异度为 17.6%，PPV 为 81.6%，NPV 为 23.1%。然后，最终筛选出CLDN10、HMGA2和LAMB3进行模型构建。在我们的队列和 TCGA 队列中，所有三个基因在甲状腺乳头状癌和正常组织之间都显示出显着的表达差异。分子模型是根据这些遗传数据和部分临床信息建立的。准确度为85.9%，敏感性为86.1%，特异性为84.6%，PPV 为96.9%，NPV 为52.4%。考虑到上述两种模型效果不是很好，我们对两种模型进行整合和优化，构建了最终的诊断模型（C-甲状腺模型）。在训练集中，准确率为96.7%，敏感性为100%，特异性为93.8%，PPV为93.3%，NPV为100%。在验证集中，准确度为97.6%，灵敏度保持100%，特异度为84.6%，PPV为97.3%，NPV为100%。通过简单、经济的方式成功建立了针对东部人群的诊断组合。仅使用四个基因和临床数据的方法。版权所有 © 2024 Wang、Dai、Lin、Zhang、Wen、Chen、Li、You、Qu、Ni 和 Cai。

A preoperative diagnosis to distinguish malignant from benign thyroid nodules accurately and sensitively is urgently important. However, existing clinical methods cannot solve this problem satisfactorily. The aim of this study is to establish a simple, economic approach for preoperative diagnosis in eastern population.Our retrospective study included 86 patients with papillary thyroid cancer and 29 benign cases. The ITK-SNAP software was used to draw the outline of the area of interest (ROI), and Ultrosomics was used to extract radiomic features. Whole-transcriptome sequencing and bioinformatic analysis were used to identify candidate genes for thyroid nodule diagnosis. RT-qPCR was used to evaluate the expression levels of candidate genes. SVM diagnostic model was established based on the METLAB 2022 platform and LibSVM 3.2 language package.The radiomic model was first established. The accuracy is 73.0%, the sensitivity is 86.1%, the specificity is 17.6%, the PPV is 81.6%, and the NPV is 23.1%. Then, CLDN10, HMGA2, and LAMB3 were finally screened for model building. All three genes showed significant differential expressions between papillary thyroid cancer and normal tissue both in our cohort and TCGA cohort. The molecular model was established based on these genetic data and partial clinical information. The accuracy is 85.9%, the sensitivity is 86.1%, the specificity is 84.6%, the PPV is 96.9%, and the NPV is 52.4%. Considering that the above two models are not very effective, We integrated and optimized the two models to construct the final diagnostic model (C-thyroid model). In the training set, the accuracy is 96.7%, the sensitivity is 100%, the specificity is 93.8%, the PPV is 93.3%, and the NPV is 100%. In the validation set, the accuracy is 97.6%, the sensitivity remains 100%, the specificity is 84.6%, the PPV is 97.3%, and the NPV is 100%.A diagnostic panel is successfully established for eastern population through a simple, economic approach using only four genes and clinical data.Copyright © 2024 Wang, Dai, Lin, Zhang, Wen, Chen, Li, You, Qu, Ni and Cai.