UNASSIGNED: A cohort of 159 MM patients was included, comprising 71 HRC and 88 non-HRC cases. Regions of interest within the most significant tumor lesions on T2WI-FS images were manually delineated, yielding 1688 features. Fourteen radiomic features were selected using 10-fold cross-validation, employing methods such as variance thresholds, Student\'s t-test, redundancy analysis, and least absolute shrinkage and selection operator (LASSO). Logistic regression was utilized to develop three prediction models: a clinical model (model 1), a T2WI-FS radiomic model (model 2), and a combined clinical-radiomic model (model 3). Receiver operating characteristic (ROC) curves evaluated and compared the diagnostic performance of these models. Kaplan-Meier survival analysis and log-rank tests assessed the prognostic value of the radiomic nomograms.
UNASSIGNED: Models 2 and 3 demonstrated significantly greater diagnostic efficacy compared to model 1 (p < 0.05). The areas under the ROC curve for models 1, 2, and 3 were as follows: training set-0.650, 0.832, and 0.846; validation set-0.702, 0.730, and 0.757, respectively. Kaplan-Meier survival analysis indicated comparable prognostic values between the radiomic nomogram and MM cytogenetic status, with log-rank test results (p < 0.05) and concordance indices of 0.651 and 0.659, respectively; z-score test results were not statistically significant (p = 0.153). Additionally, Kaplan-Meier analysis revealed that patients in the non-HRC group, low-RS group, and aged ≤ 60 years exhibited the longest overall survival, while those in the HRC group, high-RS group, and aged > 60 years demonstrated the shortest overall survival (p = 0.004, Log-rank test).
UNASSIGNED: Radiomic nomograms are capable of predicting the HRC status in MM. The cytogenetic status, radiomics model Rad score, and age collectively influence the overall survival of MM patients. These factors potentially contribute to pre-treatment clinical decision-making and prognostic assessment.
■纳入了159名MM患者的队列,包括71例HRC和88例非HRC病例。手动描绘T2WI-FS图像上最重要的肿瘤病变内的感兴趣区域,产生1688个特征。使用10倍交叉验证选择了14个放射学特征,采用方差阈值等方法,学生t检验,冗余分析,和最小绝对收缩和选择运算符(LASSO)。利用Logistic回归建立了三种预测模型:临床模型(模型1),T2WI-FS放射学模型(模型2),和联合临床-放射学模型(模型3)。接收器工作特性(ROC)曲线评估并比较了这些模型的诊断性能。Kaplan-Meier生存分析和对数秩检验评估了放射学列线图的预后价值。
■与模型1相比,模型2和3显示出明显更大的诊断功效(p<0.05)。模型1、2和3的ROC曲线下面积如下:训练集-0.650、0.832和0.846;验证集-0.702、0.730和0.757。Kaplan-Meier生存分析显示放射学列线图和MM细胞遗传学状态之间具有可比性的预后价值,对数秩检验结果(p<0.05),一致性指数分别为0.651和0.659;z评分检验结果无统计学意义(p=0.153)。此外,Kaplan-Meier分析显示,非HRC组的患者,低RS组,年龄≤60岁,总生存期最长,而HRC组的人,高RS组,年龄>60岁的患者总生存期最短(p=0.004,Log-rank检验)。
■放射组学列线图能够预测MM中的HRC状态。细胞遗传学状态,放射学模型Rad评分,和年龄共同影响MM患者的总体生存率。这些因素可能有助于治疗前临床决策和预后评估。