超越PSA阈值:临床及影像学主导的多模态 风险分层在前列腺癌PSA灰区中的应用
Beyond the PSA Threshold: Clinically and Imaging-Driven Multimodal Risk Stratification in the PSA Gray Zone of Prostate Cancer
摘要: 前列腺特异性抗原(PSA) 4~10 ng/mL的“灰区”人群长期面临诊断特异性不足与过度活检并存的困境。单一血清指标难以反映前列腺癌的生物学异质性,亟需构建整合多维信息的风险分层体系。磁共振成像(MRI)通过提高临床意义前列腺癌(csPCa)的检出率并凭借较高阴性预测值减少不必要活检,重塑了活检前诊断路径。然而,PI-RADS 3病灶及MRI阴性人群仍存在残余风险,提示MRI并未消除不确定性,而是将其转移至再分层阶段。影像组学与深度学习通过定量特征提取与端到端建模,实现由类别分级向连续概率预测的转变,进一步提升灰区人群中csPCa的识别能力。整合影像、临床变量及新型生物标志物的多模态模型在判别性能与风险校准方面均优于单一模态方法。本文系统梳理PSA灰区风险分层证据,提出由“PSA阈值决策”向以影像为核心的连续风险生态系统转型,为精准活检与个体化管理提供理论框架。
Abstract: Patients with prostate-specific antigen (PSA) levels of 4~10 ng/mL—the so-called “gray zone”—face a persistent diagnostic dilemma characterized by low specificity and frequent unnecessary biopsies. A single serum biomarker cannot adequately reflect the biological heterogeneity of prostate cancer (PCa), highlighting the need for integrated risk stratification strategies. Magnetic resonance imaging (MRI) has reshaped the pre-biopsy pathway by improving the detection of clinically significant prostate cancer (csPCa) and reducing avoidable biopsies through its high negative predictive value. However, uncertainty persists in PI-RADS 3 lesions and MRI-negative patients, indicating that MRI redistributes rather than eliminates diagnostic risk. Radiomics and deep learning further refine risk assessment by enabling quantitative feature extraction and end-to-end prediction, shifting evaluation from categorical grading to continuous probability estimation. Multimodal models integrating imaging, clinical variables, and emerging biomarkers consistently demonstrate superior discrimination and risk calibration compared with single-modality approaches. This review synthesizes current evidence on risk stratification in the PSA gray zone and proposes a transition from PSA threshold–based decision-making toward a continuous, imaging-centered risk ecosystem to support precision biopsy and individualized management.
文章引用:邹凤, 何晓静. 超越PSA阈值:临床及影像学主导的多模态 风险分层在前列腺癌PSA灰区中的应用[J]. 临床医学进展, 2026, 16(3): 2305-2313. https://doi.org/10.12677/acm.2026.1631025

1. 引言

前列腺癌(Prostate Cancer, PCa)是全球男性高发恶性肿瘤[1] [2],早期识别有临床意义前列腺癌(Clinically Significant Prostate Cancer, csPCa)与避免过度诊疗之间的平衡始终是诊断核心挑战。该矛盾在血清前列腺特异性抗原(Prostate-specific Antigen, PSA) 4~10 ng/mL的“灰区”人群中尤为突出,即PSA特异性有限、活检阳性预测值(Positive Predictive Value, PPV)低,且csPCa患病率高度异质,导致大量患者接受不必要的活检及承受相关并发症风险[3]-[5]

磁共振成像(Magnetic Resonance Imaging, MRI)的引入重塑了活检前路径。多参数MR (Multiparametric MRI, mpMRI)和双参数MR (Biparametric MRI, bpMRI)通过提高csPCa检出率及较高阴性预测值(Negative Predictive Value, NPV)减少不必要活检,使影像成为活检决策的关键“门控节点”。然而,标准化报告系统(Prostate Imaging Reporting and Data System, PI-RADS) 3病灶及MRI阴性人群仍存在残余风险,这提示MRI并未消除不确定性,而是将决策重心转移至再分层阶段,其风险评估需结合PSA密度(Prostate-Specific Antigen Density, PSAD)等临床变量。

在此背景下,影像组学与人工智能(Artificial Intelligence, AI)正成为细化PCa风险分层的重要工具。影像组学通过量化图像特征评估肿瘤异质性[6] [7],深度学习(Deep learning, DL)则可实现端到端的自动化风险预测[8] [9]。基于DL的自动检测系统在某些场景下可达到或超过专家水平,且显著提升了诊断一致性及规模化应用潜力[10] [11]。因此,PSA灰区的核心问题已从“是否活检”转向“如何整合多模态信息进行个体化决策”。影像学在这一过程中逐渐由辅助角色转变为决策链条中的核心协调者。它不仅决定谁进入活检路径,更在多模态整合框架中承担风险再分层与路径优化的枢纽功能。

本综述旨在系统梳理PSA灰区人群中的风险分层证据,并以放射学视角探讨MRI及其延伸技术如何与临床指标、生物标志物及AI模型协同,优化活检决策。通过将PSA灰区重构为一个动态的风险决策场域,我们希望阐明影像学在精准PCa管理中的核心作用,并为构建多模态整合路径提供理论依据。

2. PSA灰区的异质性基础

PSA作为PCa筛查的基石,其“器官特异性而非肿瘤特异性”的生物学本质,构成了PSA灰区(4~10 ng/mL)诊断困境的核心[12]。在此区间内,其水平升高可由PCa、良性增生或炎症等多种病理状态引起,导致良恶性病变的血清学表现高度重叠,诊断特异性显著降低。为应对这一局限,临床实践中采用PSAD、游离/总PSA比值等衍生指标进行初步校正,以提升风险鉴别能力[13]。更深层次的异质性则源于肿瘤本身,其侵袭性、分子特征(如基因融合状态)及代谢谱在个体间差异显著[14]-[17],而前列腺体积、炎症状态等临床因素进一步增加了PSA解读的复杂性与背景噪声[18] [19]。因此,PSA灰区本质上是前列腺疾病多层次异质性的集中体现。依赖单一血清标志物进行决策必然效率低下,这从根本上呼唤着能够整合多维信息、特别是具有高空间分辨率的影像学技术来主导风险分层,以实现精准诊断。为了补充PSA单一指标的局限性,近年来多种PSA衍生物与液体分子标志物被提出,并显示出在灰区中对高风险前列腺癌的再分层潜力。

3. PSA衍生物与液体分子标志物在灰区中的再分层价值

PSA灰区作为临床诊断上的不确定区域,既反映了生物学的复杂性,也暴露了单一PSA测量指标的局限性。在这一灰区内,基于血清和尿液的PSA衍生指标及分子标志物为风险再分层提供了补充生物学信息。

血清4Kscore是一种用于评估PCa风险的综合指标,通过多蛋白联合分析可提升侵袭性癌的预测准确性,其组成主要包括血清四种激肽释放酶蛋白水平(总PSA、游离PSA、完整PSA、人激肽释放酶2)。研究表明,与单纯PSA衡量相比,4Kscore在PSA灰区中对高级别PCa的预测能力更优,从而减少不必要活检次数[20]

尿液标志物同样展示出临床实用性。PCA3是一种PCa特异性长链非编码RNA,其尿液表达与肿瘤存在密切关联,能提高特异性判断,但对高等级癌的精确度一般[21]。相比之下,SelectMDx通过检测HOXC6和DLX1 mRNA表达,聚焦高等级肿瘤风险,其对csPCa的预测AUC显著高于PCA3 [22]。ExoDx Prostate IntelliScore (EPI)利用尿液外泌体中的RNA表达组合预测高级别PCa,在无需直肠指检诱导的情况下达到了较高的阴性预测值,可作为避免初次活检的筛查工具[23]

这些分子标志物的共同趋势是由单一阈值判断转向连续概率化风险输出,并强调对高分级前列腺癌的识别能力。在临床实践中,它们可以作为先于或并行于MRI的风险预筛查层,帮助临床在灰区内更合理地选择需MRI检查的人群,并减少不必要的侵入性活检。

4. MRI在灰区中的作用

4.1. MpMRI与BpMRI的选择及临床价值

PSA灰区的本质是风险分布的高度异质性,MRI的引入则标志着这一不确定性开始被结构化管理。MRI并未消除灰区,而是将原本基于单一血清指标的“概率性决策”,转化为以影像为中枢的“分层式决策路径”。在灰区人群中,MRI不再是辅助检查,而成为决定“谁需要活检、何时活检以及如何活检”的核心门控节点。

MpMRI已被确立为活检前风险评估的关键工具。PROMIS研究首次在前瞻性、多中心框架下系统验证了其在活检路径中的诊断效能,使用mpMRI对男性进行筛查可能使27%的患者避免初次活检,并减少5%的临床无意义癌症的诊断[24]。这一发现的临床意义在于,MRI的核心价值并非提高阳性预测能力,而在于其较高的阴性预测能力,使部分患者能够安全延缓或避免活检。随后PRECISION研究进一步证明,MRI引导靶向活检较传统系统活检显著提高csPCa检出率,同时减少低风险癌诊断[25]。这一研究从干预层面验证了MRI在重塑诊断路径中的实质性作用。更重要的是,多项meta-analysis证实,mpMRI在活检前路径中的高NPV具有稳定表现,而其PPV则受患病率与研究人群差异影响较大[26]。因此,在PSA灰区中,MRI的决策价值更多体现在“排除高风险”的能力,而非单纯“确认存在癌症”。

在此背景下,bpMRI逐渐受到关注。bpMRI通常省略动态对比增强序列(dynamic contrast-enhanced, DCE),仅保留T2加权成像(T2-weighted imaging, T2WI)、弥散加权成像(diffusion-weighted imaging, DWI),在保持主要诊断信息的同时显著缩短扫描时间、降低成本并提高可及性。多项研究显示,在活检前人群中,bpMRI对csPCa的诊断性能与mpMRI高度相似[27]。对于PSA灰区这一临床决策成本敏感人群而言,bpMRI的现实优势尤为重要:更短的检查时间、更低的费用以及更高的可推广性,使其在初筛与风险再分层场景中具有潜在结构性优势。因此,在灰区管理中,问题不再是“是否进行MRI”,而是“在何种情境下选择何种MRI路径”。

4.2. PI-RADS 3及MRI阴性人群的管理

在既定的影像分层框架下,MRI并未终结不确定性,而是将其收敛于两个最具决策挑战的节点:PI-RADS 3病灶以及MRI阴性但临床风险仍存的患者。前者构成“影像灰区”的核心场景,后者则提示影像阴性并非风险终点。二者共同界定了PSA灰区管理中的“残余风险区”。

PI-RADS 4~5通常对应较高的csPCa概率,其阳性预测值在多数研究中表现稳定。相比之下,PI-RADS 3被定义为“临床意义不明确的可疑病灶”,其挑战不在于定义本身,而在于高度可变的病理对应概率。系统评价显示,其csPCa检出率约为10%~30%,且不同中心差异显著[28]。有研究进一步证实,PI-RADS 3病灶中csPCa的实际风险显著受患者基础临床特征的影响[29]。这表明PI-RADS 3并非一个稳定的诊断类别,而是一个概率分布宽广的中间层级。因此,PI-RADS 3不应被视为决策终点,而应被理解为风险再分层的起点。其临床管理核心在于引入二级变量以完成概率校准。PSAD是目前证据最为一致的补充指标之一。在PI-RADS 3病灶中,PSAD ≥ 0.15 ng/mL/cm3与csPCa风险显著增加相关[30]。与此同时,ADC值亦展现出附加判别能力,低ADC值与更高Gleason评分显著相关[31] [32]

在此基础上,机器学习与影像组学模型开始介入该灰区场景。整合影像特征与临床变量的多变量模型已被证明可提升csPCa识别能力,并减少不必要活检[33]-[35]。这一趋势标志着PI-RADS体系正从“等级划分”向“概率输出”演进——即由静态分级语言过渡为动态风险估计框架。

5. 工智能在PSA灰区中的应用

5.1. 影像组学

在结构化MRI评估体系(如PI-RADS)确立之后,灰区中的核心问题逐渐从“是否存在病灶”转向“该病灶具有多大侵袭性风险”。然而,传统影像判读依赖视觉特征识别,难以捕捉影像中潜在的微观纹理信息与肿瘤异质性。影像组学通过高通量提取影像定量特征,为灰区风险的进一步细化提供了数据驱动的工具,其理论基础在于:医学影像不仅反映宏观结构改变,还蕴含与肿瘤基因表达、细胞密度及微环境相关的定量信息。Lambin等提出,影像组学可作为连接影像与精准医学之间的桥梁,将图像特征转化为可解释的生物学表征[36]。多项研究证实,基于T2WI、DWI及ADC图像的影像组学特征能够区分良性病灶与csPCa,并预测Gleason分级[37] [38]

对于PSA灰区而言,影像组学的真正价值体现在对“中间风险群体”的再分层能力,尤其是PI-RADS 3病灶。Hectors等在J Magn Reson Imaging中针对PI-RADS 3病灶建立影像组学模型,结果显示其对csPCa的预测能力显著优于PSAD等单变量指标,AUC达到0.76,并显著减少潜在不必要活检[33]。此外,多中心研究进一步表明,将影像组学评分与临床变量(如PSAD、年龄)结合可构建综合预测模型,其性能显著优于单纯PI-RADS评分。这种整合模型在外部验证队列中仍保持稳定表现,提示其具有潜在临床推广价值[39]。值得注意的是,影像组学在MRI阴性但临床风险仍存的人群中也显示出潜在价值。部分研究指出,即便视觉评估未见明确病灶,定量纹理特征仍可识别潜在高风险区域,为“隐匿型”csPCa提供额外线索[40]。然而,影像组学在PSA灰区中的临床应用仍面临挑战,包括ROI勾画方式的差异、特征选择方法的异质性、模型过拟合风险、外部验证不足等。近年来,影像组学质量评分系统(Radiomics Quality Score, RQS)被提出,用于规范研究设计与报告标准[41]

总体而言,影像组学为PSA灰区管理提供了一种“定量显微镜”,使MRI不再仅仅是结构影像,而成为可挖掘肿瘤异质性的高维数据源。在PI-RADS 3及MRI阴性等灰区情境中,影像组学能够进一步压缩不确定性区间,使风险评估由类别判断过渡为连续概率预测。它并非取代PI-RADS,而是在其框架之上叠加定量强化层,推动灰区管理迈向精细化与个体化决策。

5.2. 深度学习

尽管影像组学通过人工设计特征实现了对MRI中潜在信息的定量挖掘,但其仍依赖于ROI勾画、特征筛选及统计建模等步骤,流程复杂且易受人为因素影响。相比之下,DL通过卷积神经网络(CNN)等架构实现端到端特征学习,能够自动从原始影像数据中提取多层级空间表征,从而减少人为干预并提高模型泛化能力。在前列腺领域,DL最早应用于病灶检测与分割任务。随后研究表明,基于mpMRI构建的深度神经网络可直接预测csPCa,并在诊断性能上与经验丰富的放射科医师相当甚至更优。在一项多中心研究中,基于mpMRI的DL模型在csPCa识别中的AUC达到0.85以上,并在外部验证队列中保持稳定性能,显示出良好的泛化能力[42]。与PI-RADS相比,DL模型在中间风险人群中展现出更强的区分能力。研究表明,在PSA < 10 ng/mL的灰区患者中,DL模型对csPCa的诊断性能显著优于PI-RADS评分,提示DL在灰区中具有潜在增益价值[43]。在PI-RADS 3病灶中,DL亦显示出降低不确定性的潜力。通过直接学习影像空间特征,DL模型可能对视觉上模糊的病灶进行概率化风险输出,使“类别判断”转化为“连续概率预测”,从而为活检决策提供更加细化的依据。此外,国际前列腺AI挑战(PI-CAI)进一步验证了DL在大规模、多中心数据集上的性能。该挑战结果显示,领先的DL模型在csPCa检测中的综合表现已达到甚至超过多数人类读片者的水平,表明AI辅助判读具有现实临床可行性[44]

DL的另一优势在于其可实现自动病灶检测、分割与风险评分一体化,从而构建完整的决策支持系统。与影像组学依赖手工ROI不同,DL模型可通过全图卷积网络分析整个前列腺区域,减少人为勾画误差并提高工作流效率。然而,DL模型亦面临关键挑战。首先,模型训练需要大量标注数据,而不同中心之间MRI协议差异可能导致性能下降。其次,DL模型的“黑箱”特性限制了其可解释性。尽管Grad-CAM等可视化技术可提供一定程度的解释支持,但其仍难以完全揭示深层特征与肿瘤生物学之间的对应关系[45]。因此,在PSA灰区管理中,DL的定位并非替代放射科医师或PI-RADS体系,而是作为自动化风险评估工具、决策支持系统、降低阅片差异的标准化平台。

5.3. 多模态融合模型

在PSA灰区这一高度异质性的临床场景中,单一模态往往只能捕获风险谱的局部维度,而多模态融合模型通过整合互补信息,持续表现出更优的判别性能与风险校准能力。传统临床nomogram基于PSA、年龄及体积等变量构建风险基线;MRI引入后,PI-RADS与PSAD的联合已显著提高csPCa预测能力。然而,结构化变量无法充分反映肿瘤内部异质性。影像组学通过高维特征提取补充组织表型信息,而DL进一步捕获跨序列与跨尺度的潜在表征。将临床变量、影像分级、影像组学特征及AI风险评分整合的模型,在多项研究中均显著优于单一PI-RADS或单变量模型[46]。多模态模型“几乎总是最优”的原因可能在于:第一,不同模态反映不同生物层级;第二,融合机制可部分抵消单源噪声与偏倚;第三,连续概率输出优于离散分级,更适用于个体化活检决策。

MRI之外,其他模态正逐步纳入融合框架。PHI在PSA灰区人群中优于传统PSA,可作为进入MRI路径前的风险过滤工具[47]。在PI-RADS 3或分期评估场景中,PSMA PET/CT提供分子水平信息,其与mpMRI及临床变量构建的联合模型显示出显著提高的csPCa或包膜外侵犯预测能力[48]。因此,PSA灰区的未来诊断路径并非“PSA或MRI”的线性选择,而是向“临床 + 影像 + AI + 分子标志物”的综合风险生态系统演进。在这一框架下,PI-RADS不再是终点,而是成为多模态概率整合中的一个结构化输入变量。

6. 总结与未来方向

当前多数影像组学与DL研究仍基于回顾性、单中心数据,其局限并不仅在于样本规模不足,更在于特征可重复性与跨设备一致性尚未得到系统验证。影像特征对扫描参数、重建算法及ROI分割方式高度敏感,不同厂家与不同磁场强度间的分布差异可导致显著的域偏移,从而削弱模型的外部泛化能力[49]。尽管IBSI标准对特征定义与计算流程进行了规范化,但标准化本身并不能完全消除设备与协议差异带来的系统性偏差[50]。因此,未来研究应在前瞻性、多中心框架下开展,并将特征重复性评估、跨扫描仪一致性分析及特征协调或域适应策略纳入模型构建流程,使“跨域稳定性”成为与AUC同等重要的核心评价指标。唯有如此,影像组学与DL模型方能真正实现从研究工具向临床决策支持系统的转化。

风险评估方式也需要从离散分级逐步过渡到连续概率预测。相比简单的类别判断,基于影像与生物标志物融合的连续风险模型更有助于个体化决策,并可通过决策曲线等方法评估临床净获益。

未来的风险模型还应整合多维信息,包括MRI特征、液体活检或分子标志物,以及纵向随访数据。通过多模态与时间维度的结合,有望实现从单次诊断向动态风险管理的转变。

随着影像在风险评估中的作用不断增强,放射科医生的角色亦将拓展。除了影像判读,还需参与AI结果解读、模型质量控制及多学科决策过程。如何在保证标准化与可解释性的前提下推进临床应用,是下一阶段的重要课题。

NOTES

*通讯作者。

参考文献

[1] Siegel, R.L., Kratzer, T.B., Giaquinto, A.N., Sung, H. and Jemal, A. (2025) Cancer Statistics, 2025. CA: A Cancer Journal for Clinicians, 75, 10-45. [Google Scholar] [CrossRef] [PubMed]
[2] Bray, F., Laversanne, M., Sung, H., Ferlay, J., Siegel, R.L., Soerjomataram, I., et al. (2024) Global Cancer Statistics 2022: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA: A Cancer Journal for Clinicians, 74, 229-263. [Google Scholar] [CrossRef] [PubMed]
[3] Guo, E., Xu, L., Zhang, D., Zhang, J., Zhang, X., Bai, X., et al. (2024) Diagnostic Performance of MRI in Detecting Prostate Cancer in Patients with Prostate-Specific Antigen Levels of 4-10 ng/ml: A Systematic Review and Meta-Analysis. Insights into Imaging, 15, Article No. 147. [Google Scholar] [CrossRef] [PubMed]
[4] Loeb, S., Vellekoop, A., Ahmed, H.U., Catto, J., Emberton, M., Nam, R., et al. (2013) Systematic Review of Complications of Prostate Biopsy. European Urology, 64, 876-892. [Google Scholar] [CrossRef] [PubMed]
[5] Luo, L., Wang, R., Bai, L., Shang, J., Wang, X., Chang, R., et al. (2024) The Accuracy of Fluorine 18-Labelled Prostate-Specific Membrane Antigen PET/CT and MRI for Diagnosis of Prostate Cancer in PSA Grey Zone. British Journal of Cancer, 132, 253-258. [Google Scholar] [CrossRef] [PubMed]
[6] Liu, X., Elbanan, M.G., Luna, A., Haider, M.A., Smith, A.D., Sabottke, C.F., et al. (2022) Radiomics in Abdominopelvic Solid-Organ Oncologic Imaging: Current Status. American Journal of Roentgenology, 219, 985-995. [Google Scholar] [CrossRef] [PubMed]
[7] Shu, X., Liu, Y., Qiao, X., Ai, G., Liu, L., Liao, J., et al. (2023) Radiomic-Based Machine Learning Model for the Accurate Prediction of Prostate Cancer Risk Stratification. The British Journal of Radiology, 96, Article ID: 20220238. [Google Scholar] [CrossRef] [PubMed]
[8] Michaely, H.J., Aringhieri, G., Cioni, D. and Neri, E. (2022) Current Value of Biparametric Prostate MRI with Machine-Learning or Deep-Learning in the Detection, Grading, and Characterization of Prostate Cancer: A Systematic Review. Diagnostics, 12, Article 799. [Google Scholar] [CrossRef] [PubMed]
[9] Pellicer-Valero, O.J., Marenco Jiménez, J.L., Gonzalez-Perez, V., Casanova Ramón-Borja, J.L., Martín García, I., Barrios Benito, M., et al. (2022) Deep Learning for Fully Automatic Detection, Segmentation, and Gleason Grade Estimation of Prostate Cancer in Multiparametric Magnetic Resonance Images. Scientific Reports, 12, Article No. 2975. [Google Scholar] [CrossRef] [PubMed]
[10] Cao, R., Zhong, X., Afshari, S., Felker, E., Suvannarerg, V., Tubtawee, T., et al. (2021) Performance of Deep Learning and Genitourinary Radiologists in Detection of Prostate Cancer Using 3‐T Multiparametric Magnetic Resonance Imaging. Journal of Magnetic Resonance Imaging, 54, 474-483. [Google Scholar] [CrossRef] [PubMed]
[11] Roest, C., Fransen, S.J., Kwee, T.C. and Yakar, D. (2022) Comparative Performance of Deep Learning and Radiologists for the Diagnosis and Localization of Clinically Significant Prostate Cancer at MRI: A Systematic Review. Life, 12, Article 1490. [Google Scholar] [CrossRef] [PubMed]
[12] Stamey, T.A., Yang, N., Hay, A.R., McNeal, J.E., Freiha, F.S. and Redwine, E. (1987) Prostate-Specific Antigen as a Serum Marker for Adenocarcinoma of the Prostate. New England Journal of Medicine, 317, 909-916. [Google Scholar] [CrossRef] [PubMed]
[13] Balk, S.P., Ko, Y. and Bubley, G.J. (2003) Biology of Prostate-Specific Antigen. Journal of Clinical Oncology, 21, 383-391. [Google Scholar] [CrossRef] [PubMed]
[14] Gaudreau, P., Stagg, J., Soulières, D. and Saad, F. (2016) The Present and Future of Biomarkers in Prostate Cancer: Proteomics, Genomics, and Immunology Advancements. Biomarkers in Cancer, 8, BIC.S31802. [Google Scholar] [CrossRef] [PubMed]
[15] Barbieri, C. and Shoag, J. (2016) Clinical Variability and Molecular Heterogeneity in Prostate Cancer. Asian Journal of Andrology, 18, 543-548. [Google Scholar] [CrossRef] [PubMed]
[16] Xu, B., Chen, Y., Chen, X., Gan, L., Zhang, Y., Feng, J., et al. (2021) Metabolomics Profiling Discriminates Prostate Cancer from Benign Prostatic Hyperplasia within the Prostate-Specific Antigen Gray Zone. Frontiers in Oncology, 11, Article 730638. [Google Scholar] [CrossRef] [PubMed]
[17] Crocetto, F., Musone, M., Chianese, S., Conforti, P., Digitale Selvaggio, G., Caputo, V.F., et al. (2025) Blood and Urine-Based Biomarkers in Prostate Cancer: Current Advances, Clinical Applications, and Future Directions. The Journal of Liquid Biopsy, 9, Article ID: 100305. [Google Scholar] [CrossRef] [PubMed]
[18] Erdogan, A., Polat, S., Keskin, E. and Turan, A. (2019) Is Prostate Volume Better than PSA Density and Free/Total PSA Ratio in Predicting Prostate Cancer in Patients with PSA 2.5-10 ng/ml and 10.1-30 ng/ml? The Aging Male, 23, 59-65. [Google Scholar] [CrossRef] [PubMed]
[19] Liu, J., Dong, B., Qu, W., Wang, J., Xu, Y., Yu, S., et al. (2020) Using Clinical Parameters to Predict Prostate Cancer and Reduce the Unnecessary Biopsy among Patients with PSA in the Gray Zone. Scientific Reports, 10, Article No. 5157. [Google Scholar] [CrossRef] [PubMed]
[20] Wysock, J.S., Becher, E., Persily, J., Loeb, S. and Lepor, H. (2020) Concordance and Performance of 4Kscore and SelectMDx for Informing Decision to Perform Prostate Biopsy and Detection of Prostate Cancer. Urology, 141, 119-124. [Google Scholar] [CrossRef] [PubMed]
[21] De Kok, J.B., Verhaegh, G.W., Roelofs, R.W., et al. (2002) DD3(PCA3), a Very Sensitive and Specific Marker to Detect Prostate Tumors. Cancer research, 62, 2695-2698.
[22] Hendriks, R.J., van der Leest, M.M.G., Israël, B., Hannink, G., YantiSetiasti, A., Cornel, E.B., et al. (2021) Clinical Use of the SelectMDx Urinary-Biomarker Test with or without mpMRI in Prostate Cancer Diagnosis: A Prospective, Multicenter Study in Biopsy-Naïve Men. Prostate Cancer and Prostatic Diseases, 24, 1110-1119. [Google Scholar] [CrossRef] [PubMed]
[23] McKiernan, J., Donovan, M.J., O’Neill, V., Bentink, S., Noerholm, M., Belzer, S., et al. (2016) A Novel Urine Exosome Gene Expression Assay to Predict High-Grade Prostate Cancer at Initial Biopsy. JAMA Oncology, 2, 882-889. [Google Scholar] [CrossRef] [PubMed]
[24] Ahmed, H.U., El-Shater Bosaily, A., Brown, L.C., Gabe, R., Kaplan, R., Parmar, M.K., et al. (2017) Diagnostic Accuracy of Multi-Parametric MRI and TRUS Biopsy in Prostate Cancer (PROMIS): A Paired Validating Confirmatory Study. The Lancet, 389, 815-822. [Google Scholar] [CrossRef] [PubMed]
[25] Kasivisvanathan, V., Rannikko, A.S., Borghi, M., Panebianco, V., Mynderse, L.A., Vaarala, M.H., et al. (2018) MRI-Targeted or Standard Biopsy for Prostate-Cancer Diagnosis. New England Journal of Medicine, 378, 1767-1777. [Google Scholar] [CrossRef] [PubMed]
[26] Sathianathen, N.J., Omer, A., Harriss, E., Davies, L., Kasivisvanathan, V., Punwani, S., et al. (2020) Negative Predictive Value of Multiparametric Magnetic Resonance Imaging in the Detection of Clinically Significant Prostate Cancer in the Prostate Imaging Reporting and Data System Era: A Systematic Review and Meta-Analysis. European Urology, 78, 402-414. [Google Scholar] [CrossRef] [PubMed]
[27] Kuhl, C.K., Bruhn, R., Krämer, N., Nebelung, S., Heidenreich, A. and Schrading, S. (2017) Abbreviated Biparametric Prostate MR Imaging in Men with Elevated Prostate-Specific Antigen. Radiology, 285, 493-505. [Google Scholar] [CrossRef] [PubMed]
[28] Mazzone, E., Stabile, A., Pellegrino, F., Basile, G., Cignoli, D., Cirulli, G.O., et al. (2021) Positive Predictive Value of Prostate Imaging Reporting and Data System Version 2 for the Detection of Clinically Significant Prostate Cancer: A Systematic Review and Meta-analysis. European Urology Oncology, 4, 697-713. [Google Scholar] [CrossRef] [PubMed]
[29] Schoots, I.G. and Padhani, A.R. (2020) Risk‐adapted Biopsy Decision Based on Prostate Magnetic Resonance Imaging and Prostate‐Specific Antigen Density for Enhanced Biopsy Avoidance in First Prostate Cancer Diagnostic Evaluation. BJU International, 127, 175-178. [Google Scholar] [CrossRef] [PubMed]
[30] Frisbie, J.W., Van Besien, A.J., Lee, A., Xu, L., Wang, S., Choksi, A., et al. (2022) PSA Density Is Complementary to Prostate MP-MRI PI-RADS Scoring System for Risk Stratification of Clinically Significant Prostate Cancer. Prostate Cancer and Prostatic Diseases, 26, 347-352. [Google Scholar] [CrossRef] [PubMed]
[31] Manetta, R., Palumbo, P., Gianneramo, C., Bruno, F., Arrigoni, F., Natella, R., et al. (2019) Correlation between ADC Values and Gleason Score in Evaluation of Prostate Cancer: Multicentre Experience and Review of the Literature. Gland Surgery, 8, S216-S222. [Google Scholar] [CrossRef] [PubMed]
[32] Yan, X., Ma, K., Zhu, L., Pan, Y., Wang, Y., Shi, J., et al. (2024) The Value of Apparent Diffusion Coefficient Values in Predicting Gleason Grading of Low to Intermediate-Risk Prostate Cancer. Insights into Imaging, 15, Article No. 137. [Google Scholar] [CrossRef] [PubMed]
[33] Hectors, S.J., Chen, C., Chen, J., Wang, J., Gordon, S., Yu, M., et al. (2021) Magnetic Resonance Imaging Radiomics‐based Machine Learning Prediction of Clinically Significant Prostate Cancer in Equivocal PI‐RADS 3 Lesions. Journal of Magnetic Resonance Imaging, 54, 1466-1473. [Google Scholar] [CrossRef] [PubMed]
[34] Hou, Y., Bao, M., Wu, C., Zhang, J., Zhang, Y. and Shi, H. (2020) A Radiomics Machine Learning-Based Redefining Score Robustly Identifies Clinically Significant Prostate Cancer in Equivocal PI-RADS Score 3 Lesions. Abdominal Radiology, 45, 4223-4234. [Google Scholar] [CrossRef] [PubMed]
[35] Lu, F., Zhao, Y., Wang, Z. and Feng, N. (2025) Biparametric MRI-Based Radiomics for Prediction of Clinically Significant Prostate Cancer of PI-RADS Category 3 Lesions. BMC Cancer, 25, Article No. 615. [Google Scholar] [CrossRef] [PubMed]
[36] Lambin, P., Rios-Velazquez, E., Leijenaar, R., Carvalho, S., van Stiphout, R.G.P.M., Granton, P., et al. (2012) Radiomics: Extracting More Information from Medical Images Using Advanced Feature Analysis. European Journal of Cancer, 48, 441-446. [Google Scholar] [CrossRef] [PubMed]
[37] Min, X., Li, M., Dong, D., Feng, Z., Zhang, P., Ke, Z., et al. (2019) Multi-Parametric MRI-Based Radiomics Signature for Discriminating between Clinically Significant and Insignificant Prostate Cancer: Cross-Validation of a Machine Learning Method. European Journal of Radiology, 115, 16-21. [Google Scholar] [CrossRef] [PubMed]
[38] Zhou, C., Zhang, Y., Guo, S., Wang, D., Lv, H., Qiao, X., et al. (2023) Multiparametric MRI Radiomics in Prostate Cancer for Predicting Ki-67 Expression and Gleason Score: A Multicenter Retrospective Study. Discover Oncology, 14, Article No. 133. [Google Scholar] [CrossRef] [PubMed]
[39] Liu, S., Deng, J., Dong, D., Fang, M., Ye, Z., Hu, Y., et al. (2023) Deep Learning‐Based Radiomics Model Can Predict Extranodal Soft Tissue Metastasis in Gastric Cancer. Medical Physics, 51, 267-277. [Google Scholar] [CrossRef] [PubMed]
[40] Wibmer, A., Hricak, H., Gondo, T., Matsumoto, K., Veeraraghavan, H., Fehr, D., et al. (2015) Haralick Texture Analysis of Prostate MRI: Utility for Differentiating Non-Cancerous Prostate from Prostate Cancer and Differentiating Prostate Cancers with Different Gleason Scores. European Radiology, 25, 2840-2850. [Google Scholar] [CrossRef] [PubMed]
[41] Sanduleanu, S., Woodruff, H.C., de Jong, E.E.C., van Timmeren, J.E., Jochems, A., Dubois, L., et al. (2018) Tracking Tumor Biology with Radiomics: A Systematic Review Utilizing a Radiomics Quality Score. Radiotherapy and Oncology, 127, 349-360. [Google Scholar] [CrossRef] [PubMed]
[42] Zhao, L., Bao, J., Qiao, X., Jin, P., Ji, Y., Li, Z., et al. (2022) Predicting Clinically Significant Prostate Cancer with a Deep Learning Approach: A Multicentre Retrospective Study. European Journal of Nuclear Medicine and Molecular Imaging, 50, 727-741. [Google Scholar] [CrossRef] [PubMed]
[43] Yang, C., Li, B., Luan, Y., Wang, S., Bian, Y., Zhang, J., et al. (2024) Deep Learning Model for the Detection of Prostate Cancer and Classification of Clinically Significant Disease Using Multiparametric MRI in Comparison to Pi-Rads Score. Urologic Oncology: Seminars and Original Investigations, 42, 158.e17-158.e27. [Google Scholar] [CrossRef] [PubMed]
[44] Saha, A., Bosma, J.S., Twilt, J.J., van Ginneken, B., Bjartell, A., Padhani, A.R., et al. (2024) Artificial Intelligence and Radiologists in Prostate Cancer Detection on MRI (PI-CAI): An International, Paired, Non-Inferiority, Confirmatory Study. The Lancet Oncology, 25, 879-887. [Google Scholar] [CrossRef] [PubMed]
[45] Lundervold, A.S. and Lundervold, A. (2019) An Overview of Deep Learning in Medical Imaging Focusing on MRI. Zeitschrift für Medizinische Physik, 29, 102-127. [Google Scholar] [CrossRef] [PubMed]
[46] Chen, T., Hu, W., Zhang, Y., Wei, C., Zhao, W., Shen, X., et al. (2025) A Multimodal Deep Learning Nomogram for the Identification of Clinically Significant Prostate Cancer in Patients with Gray-Zone PSA Levels: Comparison with Clinical and Radiomics Models. Academic Radiology, 32, 864-876. [Google Scholar] [CrossRef] [PubMed]
[47] Gnanapragasam, V.J., Burling, K., George, A., Stearn, S., Warren, A., Barrett, T., et al. (2016) The Prostate Health Index Adds Predictive Value to Multi-Parametric MRI in Detecting Significant Prostate Cancers in a Repeat Biopsy Population. Scientific Reports, 6, Article No. 35364. [Google Scholar] [CrossRef] [PubMed]
[48] Xu, J., Chen, H., Chen, L., Li, T., Lin, H., Bian, S., et al. (2025) The Predictive Value of Multiparametric MRI Combined with [18F]PSMA-1007 PET/CT for the Pathological Upgrade in Prostate Cancer: A Multicenter Study. European Journal of Nuclear Medicine and Molecular Imaging, 52, 4425-4433. [Google Scholar] [CrossRef] [PubMed]
[49] Demircioğlu, A. (2025) Reproducibility and Interpretability in Radiomics: A Critical Assessment. Diagnostic and Interventional Radiology, 31, 321-328.
[50] Aguirre-Meneses, H., Stoehr-Muñoz, P., Molina-Gonzalez, M., Nuñez-Gaona, M. and Roldan-Valadez, E. (2025) Radiomics and the Image Biomarker Standardisation Initiative (IBSI): A Narrative Review Using a Six-Question Map and Implementation Framework for Reproducible Imaging Biomarkers. Cureus, 17, e95335. [Google Scholar] [CrossRef