分享:
分享到微信朋友圈
X
临床研究
酰胺质子转移加权成像在前列腺病变鉴别诊断中的价值
陈丽华 王思齐 张雨蒙 宋清伟 林良杰 王家正 张钦和 刘爱连

Cite this article as CHEN L H, WANG S Q, ZHANG Y M, et al. The value of APTw imaging in the differential diagnosis of prostatic diseases[J]. Chin J Magn Reson Imaging, 2026, 17(8): 98-106.本文引用格式 陈丽华, 王思齐, 张雨蒙, 等. 酰胺质子转移加权成像在前列腺病变鉴别诊断中的价值[J]. 磁共振成像, 2026, 17(8): 98-106. DOI:10.12015/issn.1674-8034.2026.08.010.


[摘要] 目的 探讨酰胺质子转移加权(amide proton transfer weight, APTw)成像在前列腺炎、前列腺增生(benign prostatic hyperplasia, BPH)与前列腺癌(prostate cancer, PCa)鉴别诊断中的价值。材料与方法 回顾性分析本院行3.0 T MRI扫描并经病理证实的前列腺病变患者133例,其中BPH 41例,前列腺炎23例,PCa 69例。两名观察者分别测量所有病灶的APTw值。采用组内相关系数(intra-class correlation coefficient, ICC)检验两名观察者测量APTw值的一致性。记录所有患者的年龄、总前列腺特异性抗原(total prostate specific antigen, tPSA)和游离前列腺特异性抗原(free prostate specific antigen, fPSA)。使用多因素logistic回归筛选PCa可能的独立危险因素。先进行良、恶性病变两组间各参数值的差异性比较,再进行PCa、BPH和前列腺炎三组间的比较。符合正态分布的计量资料采用t检验比较良、恶性两组病变之间的差异,采用单因素方差分析(analysis of variance, ANOVA)比较三组之间定量参数值的差异。符合偏态分布的计量资料以中位数(上下四分位数)表示,采用Mann-Whitney秩和检验比较不同组间差异。采用受试者工作特征(receiver operating characteristic, ROC)曲线评估各有差异参数单独及联合在不同组间的诊断效能。采用DeLong检验比较各参数及联合参数诊断效能的差异。结果 两名观察者测量APTw值的一致性良好(ICC>0.75)。多因素logistic回归分析,年龄、tPSA是筛选PCa可能的独立危险因素(P<0.05)。PCa与前列腺良性病变两组间比较,PCa组年龄、tPSA、APTw值均大于前列腺良性病变组,差异均有统计学意义(P<0.05);PCa、BPH和前列腺炎三组间比较,PCa组年龄大于BPH组,差异有统计学意义(P<0.05),其余组间年龄差异无统计学意义;PCa组tPSA大于前列腺炎组和BPH组,差异均有统计学意义(P<0.05),前列腺炎组与BPH组tPSA值差异无统计学意义(P>0.05);PCa组APTw值大于前列腺炎组和BPH组,差异均有统计学意义(P<0.05),前列腺炎组与BPH组APT值差异无统计学意义(P>0.05)。临床指标(年龄、tPSA)、APTw值、APTw值联合临床指标鉴别前列腺良、恶性病变的ROC曲线下面积(area under the curve, AUC)分别为0.754、0.755、0.838,临床指标和APTw值联合临床指标诊断效能差异有统计学意义(P<0.05)。tPSA、APTw值和APTw值联合tPSA鉴别PCa与前列腺炎的AUC分别为0.680、0.739、0.783,诊断效能差异无统计学意义(P>0.05)。结论 APTw值在鉴别前列腺良、恶性疾病中有很好的临床应用前景,联合临床指标可以提升诊断效能,并且能很好地鉴别PCa与前列腺炎、BPH,而且APTw值联合tPSA值对于鉴别PCa与前列腺炎的诊断有一定的提升,为临床诊断及后续治疗方法选择提供新思路。
[Abstract] Objective To explore the value of amide proton transfer weight (APTw) imaging in the differential diagnosis of prostatitis, prostatic hyperplasia (BPH) and prostate cancer (PCa).Materials and Methods A retrospective analysis was made on 133 patients with prostate diseases who underwent 3.0 T MRI scanning and were confirmed by pathology, including 41 cases of BPH, 23 cases of prostatitis and 69 cases of PCa. APTw values of all lesions were measured by two observers. Intra-class correlation coefficient (ICC) was used to test the consistency of APTw values measured by two observers. The age and prostate specific antigen [including total prostate specific antigen (tPSA) and free prostate specific antigen (fPSA)] of all patients were recorded. Multivariate logistic regression was used to screen possible independent risk factors of PCa. Firstly, the differences of parameters between benign and malignant groups were compared, and then the comparisons among PCa, BPH and prostatitis were made. T-test was used to compare the differences between benign and malignant lesions, and one-way analysis of variance (ANOVA) was used to compare the differences of quantitative parameters among the three groups. The measurement data conforming to the skewed distribution are expressed by the median (interquartile interval), and the differences between different groups are compared by Mann-Whitney rank sum test. The receiver operating characteristic (ROC) curve was used to evaluate the diagnostic efficacy of different parameters in different groups. DeLong test was used to compare the diagnostic efficiency of each parameter and joint parameters.Results The APTw values measured by the two observers are in good agreement (ICC > 0.75). Multivariate logistic regression analysis showed that age and tPSA were possible independent risk factors for screening PCa (P < 0.05). Compared with benign prostate diseases (prostatitis and BPH), the age, tPSA and APTw values of PCa group were all higher than those of benign prostate diseases group, the differences were statistically significant (P < 0.05). Compared with PCa, BPH and prostatitis, PCa group was older than BPH group, and the difference was statistically significant (P < 0.05), but there was no statistical difference among other groups. tPSA in PCa group was higher than that in prostatitis group and BPH group, and the differences were statistically significant (P < 0.05), there was no significant difference in tPSA between prostatitis group and BPH group (P > 0.05). The APTw value in PCa group was higher than that in prostatitis group and BPH group, and the difference was statistically significant (P < 0.05), but there was no statistical difference between prostatitis group and BPH group (P > 0.05). The clinical indicators (age, tPSA), APTw value and the area under ROC curve (AUC) of APTw combined with clinical indicators in differentiating benign and malignant prostate lesions were 0.754, 0.755 and 0.838, respectively, and the diagnostic efficiency of clinical indicators and APTw value combined with clinical indicators was statistically significant (P < 0.05). The AUC of tPSA, APTw and APTw combined with tPSA in differentiating PCa from prostatitis were 0.680, 0.739 and 0.783, respectively, and there was no significant difference in diagnostic efficiency (P > 0.05).Conclusions APT value has a good clinical application prospect in differentiating benign and malignant prostate diseases. Combined with clinical indicators, it can improve the diagnostic efficiency, and can well distinguish PCa from prostatitis and BPH. In addition, APTw value combined with tPSA value can improve the diagnosis of PCa and prostatitis to a certain extent, providing new ideas for clinical diagnosis and subsequent treatment methods.
[关键词] 酰胺质子转移加权成像;前列腺癌;前列腺增生;前列腺炎;磁共振成像
[Keywords] amide proton transfer weight;prostate cancer;benign prostatic hyperplasia;prostatitis;magnetic resonance imaging

陈丽华 1, 2   王思齐 1   张雨蒙 1   宋清伟 1, 2, 3   林良杰 4   王家正 3   张钦和 1, 2, 3   刘爱连 1, 2, 3*  

1 大连医科大学附属第一医院放射科,大连 116011

2 大连市医学影像人工智能技术创新中心,大连 116011

3 辽宁省超级化磁共振专业技术创新中心,大连 116011

4 飞利浦(中国)投资有限公司北京分公司,北京 100600

通信作者:刘爱连,E-mail:liuailian@dmu.edu.cn

作者贡献声明::刘爱连设计本研究的方案,分析和解释本研究的数据,对稿件重要内容进行了修改;陈丽华起草和撰写稿件,获取、分析和解释本研究的数据,获得大连市生命健康领域指导计划项目基金资助;王思齐、张雨蒙、宋清伟、林良杰、王家正、张钦和获取、分析或解释本研究的数据,对稿件重要内容进行了修改;全体作者都同意发表最后的修改稿,同意对本研究的所有方面负责,确保本研究的准确性和诚信。


基金项目: 大连市生命健康领域指导计划项目 2024ZDJH01PT064
收稿日期:2026-03-10
接受日期:2026-07-10
中图分类号:R445.2  R697.3  R737.25 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2026.08.010
本文引用格式 陈丽华, 王思齐, 张雨蒙, 等. 酰胺质子转移加权成像在前列腺病变鉴别诊断中的价值[J]. 磁共振成像, 2026, 17(8): 98-106. DOI:10.12015/issn.1674-8034.2026.08.010.

0 引言

       前列腺疾病是困扰中老年男性常见病之一,主要包括前列腺增生(benign prostatic hyperplasia, BPH)、前列腺炎和前列腺癌(prostate cancer, PCa)。PCa严重危害着全球老年男性的生活健康,其发病率和死亡率分别位于男性恶性肿瘤的第2位和第5位[1, 2]。随着我国人口老龄化趋势的加剧,以及生活水平和诊断技术的提高,PCa的发病率逐年上升[3]。PCa与BPH、前列腺炎在临床表现、实验室检查中,都存在相似表现,为准确诊断带来一定困难。早期PCa与前列腺炎均好发于外周带,而发生在移行带的PCa常伴发BPH,上述病变具有重叠的MRI征象,尤其是发生在外周带的前列腺炎与PCa影像重叠度高,在临床工作中容易误诊和漏诊,但三者的治疗方式及预后却不同,因此精准诊断非常重要[4, 5, 6]。酰胺质子转移加权(amide proton transfer weight, APTw)成像是一种新型分子成像技术,能无创地检测体内蛋白质及多肽的含量,进而反映细胞增殖、肿瘤侵袭等情况[7]。近年来APTw成像在前列腺的研究逐渐增多,研究主要集中于鉴别PCa与BPH、评估PCa危险度等方面[8, 9, 10],关注前列腺炎与PCa和BPH鉴别的研究少见报道,既往研究中将前列腺炎分入良性病变组进行分析,而本研究将前列腺炎作为单独一组与PCa、BPH进行鉴别诊断,为临床提供更细致的前列腺病变诊断,旨在探讨APTw技术及临床实验室指标对PCa与BPH,尤其是与前列腺炎的鉴别价值。

1 材料与方法

1.1 研究对象

       回顾性分析本院2019年4月至2022年9月行前列腺3.0T MRI检查的患者。纳入标准:(1)临床资料完整,经直肠超声引导下经会阴穿刺前列腺活检术或前列腺切除术后病理证实为PCa、BPH或前列腺炎;(2)治疗前1个月内进行MRI检查,序列完整,包括APTw、T2-SPAIR轴位、扩散加权成像(diffusion-weighted imaging, DWI)序列,图像质量良好。排除标准:(1)MRI检查前接受过相关内分泌或手术治疗;(2)MRI检查前2周内行穿刺活检的患者。记录所有患者的年龄和总前列腺特异性抗原(total prostate specific antigen, tPSA)、游离前列腺特异性抗原(freeprostate specific antigen, fPSA)。

       本研究遵循《赫尔辛基宣言》,经大连医科大学附属第一医院伦理委员会审核批准,批准文号:PJ-KS-KY-2022-277,免除受试者知情同意。

1.2 检查方法

       检查前4 h禁食水,膀胱内少量尿液充盈。采用Philips 3.0 T 磁共振扫描仪(Ingenia CX, Philips Healthcare,Best, Netherlands),32通道体部相控阵线圈,患者采用仰卧位。APTw采用3D快速自旋回波DIXON序列进行扫描,并使用化学位移频率选择方法进行抑脂。扫描参数详见表1

表1  MRI扫描序列及参数
Tab. 1  MRI scanning sequence and parameters

1.3 图像分析与数据测量

       扫描后数据自动传输到Philips Intelli Space Portal(ISP)工作站,将APTw伪彩图与DWI图融合。病灶位置参照病理结果,结合T2WI、DWI及表观扩散系数(apparent diffusion coefficient, ADC)图像确认病灶位置,多发病灶选取最大病灶。由两名分别拥有8年(观察者1,主治医师)和2年(观察者2,住院医师)前列腺MRI诊断经验的影像医师采用双盲法在APTw与DWI融合图像上于病灶最大层面手动勾画感兴趣区(region of interest, ROI)测得APTw值,取两名观察者测量参数平均值作为最终结果,用于后续统计分析。ROI放置标准:(1)局灶性病灶,根据病灶大小手动勾画适当不规则形ROI,ROI面积约为病灶大小的2/3;(2)ROI置于病灶实性部分,避开包膜、尿道及出血、囊变、坏死区;(3)BPH组,弥漫性增生者ROI置于中央腺体区,局灶性增生者ROI置于增生结节(图1图2)。

图1  男,66岁,前列腺炎患者。tPSA=7.20 ng/mL,fPSA=1.49 ng/mL,tPSA/fPSA=0.207。1A:T2WI图,右侧移行带稍低信号,边界不清;1B:DWI图,右侧移行带病灶呈高信号;1C:ADC图,右侧移行带病灶呈低信号;1D:APTw与DWI融合图,APTw值为1.68%;1E:前列腺组织病理图(HE 20×10),腺体周围炎症细胞浸润,间质疏松,间隙增宽。
图2  男,67岁,前列腺癌患者。tPSA=7.11 ng/mL,fPSA=1.22 ng/mL,tPSA/fPSA=0.172。2A:T2WI图,右侧移行带稍低信号,边界清;2B:DWI图,右侧移行带病灶呈高信号;2C:ADC图,右侧移行带病灶呈低信号;2D:APTw与DWI融合图,APTw值为2.10%;2E:前列腺组织病理图(HE 20×10),腺体形态不规则,细胞核增大,核浆比增高,出现大而明显的核仁。tPSA:总前列腺特异性抗原;fPSA:游离前列腺特异性抗原;DWI:扩散加权成像;ADC:表观扩散系数;APTw:酰胺质子转移加权。
Fig. 1  Male, 66 years old, patient with prostatitis. tPSA = 7.20 ng/mL, fPSA = 1.49 ng/mL, tPSA/fPSA = 0.207. 1A: T2WI shows the signal in the right transition zone is slightly low, and the boundary is unclear; 1B: DWI shows that the lesions in the right transitional zone shows high signal intensity; 1C: ADC images shows the lesions in the right transitional zone shows low signal; 1D: Fusion diagram of APTw and DWI, with APTw value of 1.68%; 1E: Histopathological map of prostate (HE 20 × 10), showing infiltration of inflammatory cells around glands, loose stroma and widening gap.
Fig. 2  Male, 67 years old, patient with prostate cancer. tPSA = 7.11 ng/mL, fPSA = 1.22 ng/mL, tPSA/fPSA = 0.172. 2A: T2WI shows the signal in the right transition zone is slightly lower and the boundary is clear; 2B: DWI shows that the lesions in the right transitional zone shows high signal intensity; 2C: ADC image shows the lesions in the right transitional zone showed low signal; 2D: Fusion diagram of APTw and DWI, with APTw value of 2.10%; 2E: Histopathological map of prostate (HE 20 × 10), irregular gland shape, enlarged nucleus, increased nucleoplasm ratio and large and obvious nucleoli. tPSA: total prostate specific antigen; fPSA: free prostate specific antigen; DWI: diffusion-weighted imaging; ADC: apparent diffusion coefficient; APTw: amide proton transfer weight.

1.4 样本量估算

       本研究采用基于受试者工作特征(receiver operating characteristic, ROC)曲线下面积(area under the curve, AUC)的Obuchowski-McClish法进行样本量估算。以APTw值鉴别PCa与前列腺良性病变为主要分析终点,设定无效假设AUC0=0.5(即无诊断能力),期望AUC1=0.775(基于本研究实际获得的APTw鉴别PCa与良性病变的AUC值),双侧α=0.05,检验效能1-β=0.80,PCa组与良性病变组比例为1∶0.93。经计算,PCa组所需样本量最少为25例,良性病变组所需样本量最少为27例,总样本量为52例。本研究实际纳入PCa组69例、良性病变组64例(共133例),满足样本量要求。

1.5 统计学方法

       采用Python 3.11及R 4.3.3软件进行统计分析。连续变量以均数±标准差(x¯±s)表示,组间比较根据数据分布特征选择独立样本t检验(正态分布)或Mann-Whitney U检验(非正态分布),多组间比较采用单因素方差分析(analysis of variance, ANOVA)或Kruskal-Wallis H检验,事后两两比较采用Bonferroni校正。分类变量以频数表示,组间比较采用χ2检验或Fisher精确检验。采用多因素logistic回归分析筛选PCa的独立危险因素,计算调整后OR值及95%置信区间。两名观察者APTw值测量一致性采用ICC评估(ICC>0.75为一致性良好)。绘制ROC曲线评估各指标单独及联合鉴别诊断效能,计算AUC、敏感度、特异度及约登指数,AUC比较采用DeLong检验,敏感度与特异度比较采用McNemar检验。

       为进一步提升PCa与前列腺炎的鉴别诊断能力,本研究引入逻辑回归(logistic regression, LR)、随机森林(random forest, RF)、极端梯度提升(eXtreme Gradient Boosting, XGBoost)、轻量级梯度提升机(light gradient boosting machine, LightGBM)及支持向量机(support vector machine, SVM)五种机器学习模型,采用重复分层5折交叉验证(5×10次重复)进行内部验证。采用SHAP(SHapley Additive exPlanations)值进行模型可解释性分析,评估各特征对PCa预测概率的贡献。采用决策曲线分析(decision curve analysis, DCA)评估不同阈值概率下联合模型的临床净获益。绘制校准曲线并计算Brier评分及Hosmer-Lemeshow检验评估模型校准度。所有检验均以双侧P<0.05为差异有统计学意义。

2 结果

2.1 临床资料

       本研究最终入组133例患者,其中PCa组69例,良性病变组64例,其中BPH组41例,前列腺炎组23例。PCa组37例为多发病灶,40例病灶位于移行带;前列腺炎组18例为多发病灶,13例病灶位于移行带。多因素logistic回归分析,年龄、tPSA是筛选PCa可能的独立危险因素(P<0.05),详见表2

2.1.1 PCa和良性病变两组临床资料比较

       PCa组年龄、tPSA高于前列腺良性病变组,差异均有统计学意义(P<0.05),详见表3

表3  PCa与前列腺良性病变组患者的临床资料
Tab. 3  Clinical data of patients with prostate cancer and benign prostate lesions

2.1.2 PCa、BPH和前列腺炎三组间临床资料比较

       PCa组年龄、fPSA大于BPH组,差异有统计学意义(P<0.05),余组间差异无统计学意义。PCa组tPSA大于BPH组、前列腺炎组,差异有统计学意义(P<0.05),BPH组tPSA与前列腺炎组差异无统计学意义(P>0.05),详见表4

表4  PCa、BPH及前列腺炎组临床指标
Tab. 4  Clinical indexes of prostate cancer, prostatic hyperplasia and prostatitis groups
表2  临床资料的多因素logistic回归分析
Tab. 2  Multivariate logistic regression analysis of clinical data

2.2 两名观察者测量APTw值的一致性检验

       两名观察者测量数据的一致性检验良好(ICC>0.75),详见表5。后续分析采用两名观察者测量的平均值。

表5  两名观察者测量PCa、BPH及前列腺炎组APTw值的一致性分析
Tab. 5  Consistency analysis of APT values measured by two observers in PCa, BPH and prostatitis groups

2.3 临床资料及APTw值比较

       PCa组APTw值(2.58%±0.66%)大于前列腺良性病变组(1.89%±0.65%),差异有统计学意义(P<0.001)。PCa组APTw值(2.58%±0.66%)大于前列腺炎组(2.02%±0.74%)和BPH组(1.81%±0.71%),差异均有统计学意义(P<0.001),前列腺炎组与BPH组APTw值差异无统计学意义(P=0.224)。患者临床资料分布、APT值比较及logistic回归分析结果见图3

图3  前列腺病变患者临床资料分布、APTw值比较及logistic 回归森林图。APTw:酰胺质子转移加权;PCa:前列腺癌;BPH前列腺增生;tPSA:总前列腺特异性抗原;fPSA:游离前列腺特异性抗原。
Fig. 3  Clinical data distribution, APTw value comparison and logistic regression forest map of patients with prostate diseases. APTw: amide proton transfer weight; PCa: prostate cancer; BPH: benign prostatic hyperplasia; tPSA: total prostate specific antigen; fPSA: free prostate specific antigen.

2.4 临床指标、APTw值鉴别PCa与前列腺良性病变单独和联合诊断效能

       临床指标、APTw值及APTw联合临床指标鉴别PCa与前列腺良性病变的AUC分别为0.754、0.755及0.838,APTw联合临床指标的诊断效能优于临床指标、APTw值单独诊断效能,差异有统计学意义(P<0.05),详见表6图4

图4  临床指标、APTw 值及APTw 联合临床指标鉴别PCa 的ROC 曲线。APTw:酰胺质子转移加权;PCa:前列腺癌;ROC:受试者工作特征;AUC:曲线下面积;tPSA:总前列腺特异性抗原;fPSA:游离前列腺特异性抗原;BPH:前列腺增生。
Fig. 4  ROC curve of clinical indicators, APTw value and APTw combined with clinical indicators in differentiating PCa from benign prostate lesions. ROC: receiver operating characteristic; APTw: amide proton transfer weight; PCa: prostate cancer; AUC: area under the curve; tPSA: total prostate specific antigen; fPSA: free prostate specific antigen; BPH: benign prostatic hyperplasia.
表6  临床指标、APTw值诊断效能及诊断效能比较
Tab. 6  Comparison of clinical indexes, APTw value and diagnostic efficiency

2.5 tPSA、APTw值鉴别PCa与前列腺炎诊断效能

       tPSA、APTw值及tPSA联合APTw鉴别PCa与前列腺炎的AUC分别为0.680、0.739及0.783,tPSA联合APTw的诊断效能优于tPSA、APTw单独诊断效能,但各组间诊断效能差异无统计学意义(P>0.05),详见表7。Mcnemar检验显示APTw值鉴别PCa与前列腺炎的敏感度大于联合参数(tPSA联合APTw),差异有统计学意义(P<0.001),APTw值鉴别PCa与前列腺炎的特异度小于联合参数,差异有统计学意义(P=0.004);tPSA值及联合参数鉴别PCa与前列腺炎的敏感度及特异度差异无统计学意义(P>0.05)

表7  tPSA、APTw值鉴别PCa与前列腺炎诊断效能及诊断效能比较
Tab. 7  Comparison of diagnostic efficiency and diagnostic efficiency between PCa and prostatitis by tPSA and APTw values

2.6 tPSA、APTw值鉴别PCa与BPH诊断效能

       tPSA、APTw值及tPSA联合APTw鉴别PCa与BPH的AUC分别为0.774、0.796及0.838,tPSA联合APTw的诊断效能优于tPSA、APTw单独诊断效能,但各组间诊断效能差异无统计学意义(P>0.05),见表8。Mcnemar检验显示单一参数与联合参数鉴别PCa与BPH的敏感度及特异度差异均无统计学意义(P>0.05)。

表8  tPSA、APTw值鉴别PCa与BPH诊断效能及诊断效能比较
Tab. 8  tPSA, APTw value differentiating PCa and BPH diagnostic efficiency and diagnostic efficiency comparison

2.7 机器学习模型分析

       PCa与前列腺炎对比中,各机器学习模型5×10交叉验证AUC显示最佳模型为logistic回归,Brier评分=0.164 1,Hosmer-Lemeshow检验P=0.714,提示模型校准良好。PCa与BPH对比中,各机器学习模型5×10交叉验证AUC显示最佳模型为logistic回归,Brier评分=0.155 3,Hosmer-Lemeshow检验P=0.444,提示模型校准良好。PCa与良性病变对比中,各机器学习模型5×10交叉验证AUC显示最佳模型为logistic回归,Brier评分=0.167 9,Hosmer-Lemeshow检验P=0.686,提示模型校准良好。详见图5

图5  机器学习模型交叉验证、校准曲线分析。5A~5C:对比5 种经典机器学习算法在区分PCa 和各良性病变组时的诊断能力,采用5×10 折交叉验证的AUC 值作为核心评价指标;5D~5F:对应场景下最优模型(logistic 回归)的校准曲线,评估模型的预测概率与实际临床结果的一致性。PCa:前列腺癌;AUC:曲线下面积;LR:逻辑回归;RF:随机森林;XGBoost:极端梯度提升;LightGBM:轻量级梯度提升机;SVM:支持向量机;BPH:前列腺增生。
Fig. 5  Cross-validation and calibration curve analysis of learning model. 5A-5C Compare five classic machine learning algorithms, and use the AUC value of 5×10 fold cross-validation as the core evaluation index when distinguishing PCa from different control groups. 5D-5F Correspond to the calibration curve of the optimal model (logistic regression) in the scene, and evaluate the consistency between the prediction probability of the model and the actual clinical results. PCa: prostate cancer; AUC: area under the curve; LR: logistic regression; RF: random forest; XGBoost: eXtreme Gradient Boosting; LightGBM: light gradient boosting machine; SVM: support vector machine.

2.8 SHAP可解释性分析

       SHAP分析显示,APTw值在PCa与良性病变及PCa与前列腺炎鉴别中均为最重要的预测特征,其SHAP值绝对值均值最高,表明APTw对PCa预测概率的贡献最大。年龄和tPSA同样具有重要贡献,而fPSA的贡献相对较小。在PCa与前列腺炎鉴别中,APTw值的SHAP重要性尤为突出,进一步证实APTw在鉴别PCa与前列腺炎中的关键价值。详见图6

图6  SHAP可解释性分析。6A:鉴别PCa和良性病变时,APTw值是对模型预测影响最大的核心特征,其次是tPSA,年龄为第三重要特征,fPSA的影响最小;6B:鉴别PCa和前列腺炎时,APTw值依然是影响最大的核心特征,但tPSA的重要性大幅下降至最末位,年龄和fPSA 的重要性显著上升,成为第二、第三重要的特征。PCa:前列腺癌;APTw:酰胺质子转移加权;tPSA:总前列腺特异性抗原;fPSA:游离前列腺特异性抗原。
Fig. 6  SHAP interpretable analysis. 6A: When differentiating PCa from benign lesions, APTw value is the core feature that has the greatest influence on model prediction, followed by tPSA, age is the third most important feature, and fPSA has the least influence. 6B APTw value is still the most influential core feature in differentiating PCa from prostatitis, but the importance of tPSA drops to the last, and the importance of age and fPSA increases significantly, becoming the second and third important features. PCa: prostate cancer; APTw: amide proton transfer weight; tPSA: total prostate specific antigen; fPSA: free prostate specific antigen.

2.9 DCA

       DCA显示,在较宽的阈值概率范围内,联合模型的净获益均高于“全部干预”和“均不干预”策略,表明联合模型具有临床应用价值。在PCa与前列腺炎鉴别中,联合模型在阈值概率0.2~0.8范围内净获益优于单一参数模型(图7)。

图7  决策曲线分析。7A:在PCa与良性病变鉴别中,联合模型在阈值概率0.2~0.8 范围内,联合模型的净获益高于单一参数模型;7B:在PCa 与前列腺炎鉴别中,联合模型在阈值概率0.2~0.8 范围内,联合模型的净获益优于单一参数模型;7C:在PCa 与BPH 鉴别中,联合模型在阈值概率0.2~0.8 范围内,联合模型的净获益高于单一参数模型。PCa:前列腺癌;BPH:前列腺增生。
Fig. 7  Decision curve analysis. 7A: In the differential diagnosis of PCa and benign lesions, the joint model has a threshold probability of 0.2 to 0.8, and the net benefit of the joint model is higher than that of the single parameter model. 7B: In the identification of PCa and prostatitis, the joint model has a threshold probability of 0.2 to 0.8, and the net benefit of the joint model is better than that of the single parameter model. 7C: In the identification of PCa and BPH, the joint model has a threshold probability of 0.2 to 0.8, and the net benefit of the joint model is higher than that of the single parameter model. PCa: prostate cancer; BPH: benign prostatic hyperplasia.

2.10 临床影响与鉴别诊断分析

       为进一步评估APTw值的临床应用价值,本研究绘制了SHAP特征重要性排序图、tPSA与APTw的联合分布散点图、精确率-召回率曲线、APTw核密度分布图及联合模型混淆矩阵。SHAP特征重要性分析显示,在PCa与前列腺炎鉴别中,APTw值的平均|SHAP|值最高,为最重要的鉴别特征;在PCa与良性病变鉴别中,APTw值同样具有重要贡献。tPSA与APTw联合分布散点图显示,PCa组在tPSA和APTw值上均呈现偏高趋势,且两组间存在一定的分布差异区域,提示联合使用可提高鉴别能力。精确率-召回率(PR)曲线分析显示,在PCa与前列腺炎鉴别中,tPSA+APTw联合模型的平均精确率(AP)优于单一参数,表明在不平衡数据中联合模型具有更好的正类识别能力。APTw核密度分布图直观展示了PCa与前列腺炎两组APTw值的分布重叠与差异区域。联合模型混淆矩阵显示,基于年龄、tPSA和APTw的logistic回归模型在最佳阈值下对PCa与良性病变的整体分类准确率较高。详见图8

图8  临床影响与鉴别诊断分析。8A:SHAP 特征重要性柱状图(PCa vs. 前列腺良性病变),APTw 的|SHAP|均值最高,tPSA的|SHAP|均值最低,在该场景下对模型预测的影响最小,单指标鉴别能力弱于其他3 个指标。8B:SHAP 特征重要性柱状图(PCa vs. 前列腺炎),APTw 依然是|SHAP|均值最高的核心特征,tPSA 的重要性大幅上升,成为第二重要的特征,而fPSA 的重要性下降至最末位,所有特征的|SHAP|均值整体高于图8A场景,这说明各指标在鉴别PCa 和前列腺炎时,对模型预测的平均影响更大。8C:tPSA与APTw的散点分布图,PCa 患者的散点整体集中在tPSA和APTw的高值区域,BPH和前列腺炎患者的散点集中在低值区域,有明确的分布边界;当tPSA>10 ng/mL 且APTw>2% 时,散点几乎全部为PCa 患者,说明两个指标联合使用时,可大幅提升PCa诊断的特异性。8D:精确率-召回率曲线,tPSA+APTw联合模型(AP=0.924)>APTw单指标模型(AP=0.896)>tPSA单指标模型(AP=0.884),联合模型的整体诊断效能最优;在相同召回率(相同漏诊率)下,联合模型的精确率始终高于两个单指标模型,说明联合模型在保证PCa检出率的同时,能大幅降低误诊率;APTw单指标模型的AP值高于tPSA单指标模型,再次验证了APTw的单指标诊断价值显著优于tPSA。8E:APTw值的密度分布曲线(PCa vs. 前列腺炎),PCa患者的APTw值分布整体向右偏移,峰值集中在2.5%~3.5%区间;前列腺炎患者的APTw值分布整体向左偏移,峰值集中在1.5%~2.5%区间,两条曲线重叠区域较小,说明APTw可有效区分两类人群。8F:模型预测的混淆矩阵图,模型的整体诊断准确率为(47+55)/133≈76.7%,整体诊断性能符合临床筛查的需求。PCa:前列腺癌;APTw:酰胺质子转移加权;tPSA:总前列腺特异性抗原;fPSA:游离前列腺特异性抗原;BPH:前列腺增生。
Fig. 8  Clinical influence and differential diagnosis analysis. 8A: Histogram of Shap feature importance (PCa vs. benign prostate lesions), the average value of |SHAP| of APTw is the highest, and the average value of tPSA is the highest. 8B: Histogram of the importance of Shap features (PCa vs. prostatitis), APTw is still the core feature with the highest average value of |SHAP|, and the importance of tPSA has greatly increased to become the second important feature, while the importance of fPSA has dropped to the last place; the average value of |SHAP| of all features is higher than that of scene Figure 8A, which shows that each index has a greater influence on the average prediction of the model when distinguishing PCa from prostatitis. 8C: Scatter distribution map of tPSA and APTw. The scatter of PCa patients is concentrated in the high-value area of TPSA and APTw, while the scatter of BPH and prostatitis patients is concentrated in the low-value area with clear distribution boundaries; when tPSA > 10 ng/mL and APTw > 2%, almost all the scattered points are PCa patients, which shows that the specificity of PCa diagnosis can be greatly improved when the two indexes are used together. 8D: Accuracy-recall curve, tPSA and APTw combined model (AP = 0.924) > APTw single-index model (AP = 0.896) > tPSA single-index model (AP = 0.884), the overall diagnostic efficiency of the combined model is the best; under the same recall rate (the same missed diagnosis rate), the accuracy of the joint model is always higher than that of the two single-index models, which shows that the joint model can greatly reduce the misdiagnosis rate while ensuring the detection rate of PCa. The AP value of APTw single index model is higher than that of tPSA single index model, which once again verifies that the diagnostic value of APTw single index is significantly better than that of TPSA. 8E: Density distribution curve of APTw value (PCa vs. prostatitis), the APTw value distribution of PCa patients shifted to the right as a whole, and the peak value concentrated in the range of 2.5% to 3.5%; the distribution of APTw value in patients with prostatitis shifted to the left as a whole, with the peak value concentrated in the range of 1.5% to 2.5%, and the overlapping area of the two curves was small, indicating that APTw can effectively distinguish the two groups of people. 8F: The confusion matrix diagram predicted by the model shows that the overall diagnostic accuracy of the model is (47 + 55) / 133 ≈ 76.7%, and the overall diagnostic performance meets the needs of clinical screening. PCa: prostate cancer; APTw: amide proton transfer weight; tPSA: total antigen; fPSA: free prostate specific antigen; BPH: benign prostatic hyperplasia.

3 讨论

       本研究使用APTw技术及临床实验室指标对PCa与BPH,尤其是与前列腺炎进行了鉴别诊断,结果显示APTw值能很好地鉴别前列腺良、恶性疾病,联合临床指标可以提升诊断效能;并且本研究对前列腺疾病进行更加细致的分组,将前列腺炎单独分为一组进行分析,结果显示APTw值能很好地鉴别PCa与前列腺炎、BPH,而且APTw值联合tPSA值对于鉴别PCa与前列腺炎的诊断效能有一定的提升,为临床诊断及后续治疗方法选择提供新思路。

3.1 APTw的原理及临床应用

       APTw是一种新型无创的化学交换饱和转移成像技术,WARD等[11]和REESINK等[12]证明了可交换质子向水的饱和转移过程可以提高分子检测的敏感度。APTw技术则通过探测自由水中氢质子饱和前后的信号变化来间接获得信号值,通过水信号的衰减量来反映蛋白质、多肽的含量及环境变化,从而反映体内细胞增殖情况、肿瘤侵袭性变化等[13, 14, 15]。APTw已经应用于许多疾病的研究中[16, 17],包括前列腺疾病[18, 19]

       相较于常规MRI及DWI、对比增强磁共振成像(contrast-enhanced MRI, DCE-MRI)等功能成像,主要依靠形态、弥散及灌注特征进行宏观诊断,APTw基于分子层面的化学交换饱和转移效应,可无创反映组织内蛋白及质子交换的微观代谢特征,能够从分子代谢维度提供常规MRI 无法获取的量化信息。尤其对于前列腺炎、早期PCa、BPH这类影像表现重叠、容易混淆的病灶,APTw可提供额外定量参数,辅助提升细微病变的鉴别效能。

3.2 APTw成像对前列腺良、恶性两组病变的鉴别诊断价值

       JIA等[20]的研究首次评价了APTw检测PCa的可行性和能力,研究中发现 Ca组织的APTw信号强度显著高于外周良性区域。本研究中,PCa组APTw值大于良性病变组,与JIA等研究结果相仿。GUO等[21]研究了APTw对于前列腺移行带癌与增生间的鉴别能力,结果显示移行带PCa与基质型BPH、腺体型BPH之间的APTw信号强度有明显差异,而基质型BPH与腺体型BPH之间的APTw信号强度无明显差异,提示不同类型的BPH间APTw信号强度无本质差异。本研究中,PCa组APTw值大于BPH组,与GUO等研究结果相仿。分析其原因,首先PCa细胞增殖异常活跃、细胞代谢速率显著增高,细胞内增殖、侵袭相关功能性蛋白合成与表达水平明显上调,组织内可交换质子及大分子蛋白含量增多,进而引起APTw信号明显升高;其次,PCa组织微血管密度显著高于良性增生及炎症组织,肿瘤新生血管丰富、血供灌注旺盛,血液中游离蛋白及小分子物质富集,可进一步抬高局部APTw信号水平。而BPH及前列腺炎组织细胞增殖相对平缓、代谢活性较低,血管增生及蛋白表达程度远不及肿瘤组织,因此APTw信号整体偏低,与PCa形成明显参数差异。

3.3 APTw成像对PCa、BPH与前列腺炎三组病变的鉴别诊断价值

       不同于既往研究,本研究对病例进行了更加细致的分组,良性病变包括BPH、前列腺炎,分别与PCa进行了比较。研究结果显示,PCa组APTw值大于BPH组、前列腺炎组,差异均有统计学意义,BPH组APTw值小于前列腺炎组,差异无统计学意义。分析其原因可能为,前列腺炎由于炎性肉芽组织及纤维结缔组织增生会引起微血管增多,但其微血管数量、结构与BPH组织相近、蛋白含量相似,而远低于PCa组织,所以引起APTw信号强度改变低于PCa、而与BPH差异不大。本研究分组补充了既往研究多将前列腺炎归入良性组而无法细化区分的不足,也从分子代谢层面解释了临床中前列腺炎与PCa、BPH影像易混淆的内在机制。

3.4 APTw成像联合临床指标对前列腺病变的诊断价值

       另外,本研究将临床指标也纳入研究,结果显示PCa组年龄、tPSA高于前列腺良性病变组(BPH+前列腺炎症),APTw值联合临床指标的诊断效能优于临床指标、APTw值单独的诊断效能。进一步分析显示,tPSA、APTw值能很好地鉴别PCa与前列腺炎、BPH,tPSA联合APTw的诊断效能优于tPSA、APTw单独的诊断效能。Mcnemar检验显示APTw值鉴别PCa与前列腺炎的敏感度大于联合参数(tPSA联合APTw)。

3.5 局限性

       本研究存在以下局限性:(1)本研究样本量较小,未来有待加大样本量对结果进行进一步验证;(2)APTw与DWI序列的扫描层厚不同,ROI放置时可能未能完全匹配,并且ROI放置在病灶的最大层面,不能完全反映病灶的整体情况,未来将进行多层面测量;(3)APTw序列设备依赖性高,尚未普及,临床中无法替代DWI、CE-MRI,只能作为补充性分子影像手段。

4 结论

       综上所述,APTw技术作为一种新型的MRI技术,可以在体、无创地评价前列腺组织生理状态的变化,APTw在前列腺良、恶性疾病的鉴别诊断方面具有较好的应用潜能,联合临床指标后对诊断效能有所提升,为临床前列腺良、恶性疾病的鉴别诊断提供了新思路。同时,APTw能很好地鉴别PCa与前列腺炎、BPH,与tPSA相比APTw在鉴别PCa与前列腺炎方面有较高的诊断效能,并且其敏感度高,为临床诊断前列腺炎并指导后续治疗方法提供了新指标。

[1]
SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA Cancer J Clin, 2021, 71(3): 209-249. DOI: 10.3322/caac.21660.
[2]
吴琪, 范伯男, 李岩. 2022全球癌症统计报告分析解读: 中国与世界癌症疾病负担与流行趋势[J]. 诊断学理论与实践, 2025, 24(2): 135-145. DOI: 10.16150/j.1671-2870.2025.02.004.
WU Q, FAN B N, LI Y. Analysis and interpretation of the 2022 Global Cancer Statistics Report: cancer burden and epidemiological trends in China and the world[J]. J Diagn Concepts Pract, 2025, 24(2): 135-145. DOI: 10.16150/j.1671-2870.2025.02.004.
[3]
XIA C F, DONG X S, LI H, et al. Cancer statistics in China and United States, 2022: profiles, trends, and determinants[J]. Chin Med J, 2022, 135(5): 584-590. DOI: 10.1097/cm9.0000000000002108.
[4]
方磊, 曹孟鸿, 方慧, 等. T2FLAIR在前列腺炎与前列腺癌鉴别诊断中的初步研究[J]. 中国医学计算机成像杂志, 2022, 28(3): 281-285. DOI: 10.19627/j.cnki.cn31-1700/th.2022.03.013.
FANG L, CAO M H, FANG H, et al. A preliminary study of T2 FLAIR sequence in the differential diagnosis of prostatitis and prostate cancer[J]. Chin Comput Med Imaging, 2022, 28(3): 281-285. DOI: 10.19627/j.cnki.cn31-1700/th.2022.03.013.
[5]
PINSKY P F, PARNES H. Screening for prostate cancer[J]. N Engl J Med, 2023, 388(15): 1405-1414. DOI: 10.1056/nejmcp2209151.
[6]
SCHIEDA N, PURYSKO A S. Prostate cancer imaging[J]. Radiol Clin N Am, 2024, 62(1): xv-xvi. DOI: 10.1016/j.rcl.2023.07.004.
[7]
李茜玮, 陈丽华, 王楠, 等. 初探酰胺质子加权联合mDixon-Quant序列评估前列腺癌骨转移的价值[J]. 中国临床医学影像杂志, 2022, 33(8): 540-545. DOI: 10.12117/jccmi.2022.08.002.
LI X W, CHEN L H, WANG N, et al. Preliminary study on the application of amide proton transfer-weighted combined with mDixon-Quant imaging in differentiating bone metastasis of prostate cancer[J]. J China Clin Med Imaging, 2022, 33(8): 540-545. DOI: 10.12117/jccmi.2022.08.002.
[8]
任雪, 刘爱连, 陈丽华, 等. APT联合mDIXON-Quant成像对前列腺癌和前列腺增生的鉴别诊断价值[J]. 放射学实践, 2023, 38(1): 58-64. DOI: 10.13609/j.cnki.1000-0313.2023.01.011.
REN X, LIU A L, CHEN L H, et al. The value of APT combined with mDIXON-Quant imaging in the differential diagnosis of prostate cancer and benign prostatic hyperplasia[J]. Radiol Pract, 2023, 38(1): 58-64. DOI: 10.13609/j.cnki.1000-0313.2023.01.011.
[9]
胡文君, 刘爱连, 陈丽华, 等. 酰胺质子转移成像联合T2-mapping鉴别前列腺癌与前列腺增生的价值[J]. 放射学实践, 2022, 37(11): 1416-1421. DOI: 10.13609/j.cnki.1000-0313.2022.11.015.
HU W J, LIU A L, CHEN L H, et al. The value of amide proton transfer imaging combined with T2-mapping for differentiating prostate cancer from benign prostatic hyperplasia[J]. Radiol Pract, 2022, 37(11): 1416-1421. DOI: 10.13609/j.cnki.1000-0313.2022.11.015.
[10]
张鹏运, 汤芸行, 姜昊洋, 等. 三维酰胺质子转移成像鉴别前列腺癌伴骨转移与不伴骨转移的可行性研究[J]. 磁共振成像, 2022, 13(12): 100-103, 110. DOI: 10.12015/issn.1674-8034.2022.12.017.
ZHANG P Y, TANG Y X, JIANG H Y, et al. Feasibility study of three-dimensional amide proton transfer imaging in differentiating prostate cancer with and without bone metastasis[J]. Chin J Magn Reson Imaging, 2022, 13(12): 100-103, 110. DOI: 10.12015/issn.1674-8034.2022.12.017.
[11]
WARD K M, ALETRAS A H, BALABAN R S. A new class of contrast agents for MRI based on proton chemical exchange dependent saturation transfer (CEST)[J]. J Magn Reson, 2000, 143(1): 79-87. DOI: 10.1006/jmre.1999.1956.
[12]
REESINK D J, ARTEAGA DE CASTRO C S, VAN DER VELDEN T, et al. Feasibility of clinical studies of chemical exchange saturation transfer magnetic resonance imaging of prostate cancer at 7 T[J/OL]. NMR Biomed, 2023, 36(10): e4958 [2026-03-09]. https://pubmed.ncbi.nlm.nih.gov/37094995/. DOI: 10.1002/nbm.4958.
[13]
CHEN W C, LIU G Q, CHEN J L, et al. Whole-tumor amide proton transfer-weighted imaging histogram analysis to predict pathological extramural venous invasion in rectal adenocarcinoma: a preliminary study[J]. Eur Radiol, 2023, 33(7): 5159-5171. DOI: 10.1007/s00330-023-09418-1.
[14]
WU M H, JIANG T L, GUO M, et al. Amide proton transfer-weighted imagingand derived radiomics in the classification of adult-type diffuse gliomas[J]. Eur Radiol, 2024, 34(5): 2986-2996. DOI: 10.1007/s00330-023-10343-6.
[15]
OBDEIJN I V, WIEGERS E C, ALIC L, et al. Amide proton transfer weighted imaging in pediatric neuro-oncology: initial experience[J/OL]. NMR Biomed, 2024, 37(6): e5122 [2026-03-09]. https://pubmed.ncbi.nlm.nih.gov/38369653/. DOI: 10.1002/nbm.5122.
[16]
KULANTHAIVELU K, JABEEN S, SAINI J, et al. Amide proton transfer imaging for differentiation of tuberculomas from high-grade gliomas: Preliminary experience[J]. Neuroradiol J, 2021, 34(5): 440-448. DOI: 10.1177/19714009211002766.
[17]
WU B L, JIA F, LI X K, et al. Amide proton transfer imaging vs diffusion kurtosis imaging for predicting histological grade of hepatocellular carcinoma[J]. J Hepatocell Carcinoma, 2020, 7: 159-168. DOI: 10.2147/JHC.S272535.
[18]
YIN H J, WANG D D, YAN R F, et al. Comparison of diffusion kurtosis imaging and amide proton transfer imaging in the diagnosis and risk assessment of prostate cancer[J/OL]. Front Oncol, 2021, 11: 640906 [2026-03-09]. https://pubmed.ncbi.nlm.nih.gov/33937041/. DOI: 10.3389/fonc.2021.640906.
[19]
QIN X Y, LV J, ZHANG J M, et al. Amide proton transfer imaging has added value for predicting extraprostatic extension in prostate cancer patients[J/OL]. Front Oncol, 2024, 14: 1327046 [2026-03-09]. https://pubmed.ncbi.nlm.nih.gov/38496759/. DOI: 10.3389/fonc.2024.1327046.
[20]
JIA G, ABAZA R, WILLIAMS J D, et al. Amide proton transfer MR imaging of prostate cancer: a preliminary study[J]. Magnetic Resonance Imaging, 2011, 33(3): 647-654. DOI: 10.1002/jmri.22480.
[21]
GUO Z X, QIN X Y, MU R H, et al. Amide proton transfer could provide more accurate lesion characterization in the transition zone of the prostate[J]. J Magn Reson Imaging, 2022, 56(5): 1311-1319. DOI: 10.1002/jmri.28204.

上一篇 多模态MRI联合临床病理预测乳腺癌新辅助治疗后腋窝淋巴结病理完全缓解
下一篇 增强型动脉自旋标记与三维动脉自旋标记脑血流量测量的一致性研究
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2