Page 135 - 南京医科大学自然版
P. 135
第45卷第8期 钱 哲,臧 攀,丁 磊,等. 高危前列腺癌患者穿刺病理预测模型的构建及验证研究[J].
2025年8月 南京医科大学学报(自然科学版),2025,45(8):1186-1193,1210 ·1191 ·
A B
Full(AUC:0.856,95%CI:0.812-0.900) 1.00 High risk
PSAD(AUC:0.783,95%CI:0.728-0.839)
1.0 0.75 Low risk
PI⁃RADS socre(AUC:0.697,95%CI:0.652-0.741)
Age(AUC:0.613,95%CI:0.529-0.697) Threshold 0.50
0.8
Number of lesions(AUC:0.592,95%CI:0.519-0.664) 0.25
Histological score(AUC:0.552,95%CI:0.479-0.625)
Sensitivity 0.6 0 0 100 Patient number 300 400
200
0.4
0.2
0
1.0 0.8 0.6 0.4 0.2 0
Specificity
C D
0 10 20 30 40 50 60 70 80 90 100
40
Points
(%) 30 Age
lmportance 20 Lesion 406080
3 1
10
2
0 4 2
Age Lesions PI⁃RADS PSAD Histological score 13
Histological score PI⁃RADS 4 5
PSAD
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5
Total points
0 10 20 30 40 50 60 70 80 90 100 110 120 130
Risk
0.10 0.50 0.90 0.99
E F 1.00
AUC:0.886,95%CI:0.776-0.995 High risk
1.0 0.75 Low risk
Threshold 0.50
0.8 0.25
Sensitivity 0.6 0 0 20 40 60 80
0.4
0.2 Patient number
0
1.0 0.8 0.6 0.4 0.2 0
Specificity
A:ROC curves of the puncture pathology prediction model constructed by each factor individually as well as integrally in the training set(n=424);
The optimal risk threshold was 0.634,and the corresponding specificity and sensitivity of the model in the training set was 96.2% and 65.6%,respec⁃
tively. B:Distribution of patients above and below the threshold in the training set. C:Comparison of the contribution of each factor in the prediction
model. D:Nomogram constructed by integrating the variables in the model. E:ROC curves of the prediction model in the validation set(n=78). The opti⁃
mal risk threshold was 0.495,and the corresponding specificity and sensitivity of the model in the validation set was 87.5% and 80.0%,respectively. F:
Distribution of patients above and below the threshold in the validation set.
图2 预测模型在训练集和验证集中的表现
Figure 2 Performance of the predictive model in the training and validation sets
断的研究大多是基于对前列腺病理的预测 [21] ,然而 究纳入的因素较为繁琐,如磁共振波谱成像和4K评
这并不能为高危 PCa 患者前列腺穿刺时舍弃 SB 提 分等,相比之下,本研究纳入的因素更加简便,容易
供直观具体的证据。因此,本研究更有针对性地构 获取且整合后的模型有着更好的AUC表现 [22-23] 。
建前列腺TB病理的预测模型,能够为模型筛选后的 本研究为单中心回顾性分析,受研究特征制
患者舍弃SB提供更直观的依据。此外,一些类似研 约,当前结论仍需多中心前瞻性研究予以佐证。值

