Research progress of artificial intelligence-assisted pathological image analysis in the diagnosis of prostate cancer

doi:10.3877/cma.j.issn.1674-3253.2026.01.001

Abstract

Abstract:

Prostate cancer is one of the most common malignant tumors in the male urogenital system, and its early and accurate diagnosis is crucial for clinical decision-making. Pathological imaging is the gold standard for diagnosing prostate cancer, and recent advancements in artificial intelligence (AI) technologies have significantly improved the analysis of pathological images. These innovative techniques permeate the entire process of prostate cancer pathological diagnosis, demonstrating broad application prospects. This paper systematically reviews the applications and performance of AI models in clinical tasks such as pathological image segmentation and risk stratification for prostate cancer, discusses the advantages and limitations of AI technologies, and looks ahead to future research directions.

Key words: Prostate cancer, Artificial intelligence, Deep learning, Computational pathology, Foundation models

Haonan Mei, Rui Yang, Xiuheng Liu. Research progress of artificial intelligence-assisted pathological image analysis in the diagnosis of prostate cancer[J]. Chinese Journal of Endourology(Electronic Edition), 2026, 20(01): 1-7.

/ / Recommend

Add to citation manager EndNote|Ris|BibTeX

URL: https://zhqjmnmwkzz.cma-cmc.com.cn/EN/10.3877/cma.j.issn.1674-3253.2026.01.001

https://zhqjmnmwkzz.cma-cmc.com.cn/EN/Y2026/V20/I01/1

Figures/Tables 2

References 50

[1]	Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024[J]. CA A Cancer J Clin, 2024, 74(1): 12-49. DOI: 10.3322/caac.21820.
[2]	Han B, Zheng R, Zeng H, et al. Cancer incidence and mortality in China, 2022[J]. J Natl Cancer Cent, 2024, 4(1): 47-53. DOI: 10.1016/j.jncc.2024.01.006.
[3]	Ronneberger O, et al. Medical image computing and computer-assisted intervention-MICCAI 2015 [M]. Lecture Notes in Computer Science, 2015, 9351: 234-241.
[4]	Ren J, Sadimin E, Foran DJ, et al. Computer aided analysis of prostate histopathology images to support a refined Gleason grading system[C]//Medical Imaging 2017: Image Processing，Orlando, Florida, USA. SPIE, 2017: 101331V</first_page>. DOI: 10.1117/12.2253887.
[5]	Li W, Li J, Sarma KV, et al. Path R-CNN for prostate cancer diagnosis and gleason grading of histological images[J]. IEEE Trans Med Imag, 2019, 38(4): 945-954. DOI: 10.1109/TMI.2018.2875868.
[6]	Salvi M, Bosco M, Molinaro L, et al. A hybrid deep learning approach for gland segmentation in prostate histopathological images[J]. Artif Intell Med, 2021, 115: 102076. DOI: 10.1016/j.artmed.2021.102076.
[7]	Silva-Rodríguez J, Colomer A, Naranjo V. WeGleNet: a weakly-supervised convolutional neural network for the semantic segmentation of Gleason grades in prostate histology images[J]. Comput Med Imaging Graph, 2021, 88: 101846. DOI: 10.1016/j.compmedimag.2020.101846.
[8]	Bukowy JD, Foss H, McGarry SD, et al. Accurate segmentation of prostate cancer histomorphometric features using a weakly supervised convolutional neural network[J]. J Med Imaging (Bellingham), 2020, 7(5): 057501. DOI: 10.1117/1.jmi.7.5.057501.
[9]	Pantanowitz L, Quiroga-Garza GM, Bien L, et al. An artificial intelligence algorithm for prostate cancer diagnosis in whole slide images of core needle biopsies: a blinded clinical validation and deployment study[J]. Lancet Digit Health, 2020, 2(8): e407-e416. DOI: 10.1016/S2589-7500(20)30159-X.
[10]	Bulten W, Pinckaers H, van Boven H, et al. Automated deep-learning system for Gleason grading of prostate cancer using biopsies: a diagnostic study[J]. Lancet Oncol, 2020, 21(2): 233-241. DOI: 10.1016/s1470-2045(19)30739-9.
[11]	Ström P, Kartasalo K, Olsson H, et al. Artificial intelligence for diagnosis and grading of prostate cancer in biopsies: a population-based, diagnostic study[J]. Lancet Oncol, 2020, 21(2): 222-232. DOI: 10.1016/s1470-2045(19)30738-7.
[12]	Duenweg SR, Brehler M, Bobholz SA, et al. Comparison of a machine and deep learning model for automated tumor annotation on digitized whole slide prostate cancer histology[J]. PLoS One, 2023, 18(3): e0278084. DOI: 10.1371/journal.pone.0278084.
[13]	Campanella G, Hanna MG, Geneslaw L, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images[J]. Nat Med, 2019, 25(8): 1301-1309. DOI: 10.1038/s41591-019-0508-1.
[14]	Fogarty R, Goldgof D, Hall L, et al. Classifying malignancy in prostate glandular structures from biopsy scans with deep learning[J]. Cancers (Basel), 2023, 15(8): 2335. DOI: 10.3390/cancers15082335.
[15]	Inamdar MA, Raghavendra U, Gudigar A, et al. A novel attention-based model for semantic segmentation of prostate glands using histopathological images[J]. IEEE Access, 2023, 11: 108982-108994.
[16]	Patkar S, Harmon S, Sesterhenn I, et al. A selective CutMix approach improves generalizability of deep learning-based grading and risk assessment of prostate cancer[J]. J Pathol Inform, 2024, 15: 100381. DOI: 10.1016/j.jpi.2024.100381.
[17]	Kondejkar T, Al-Heejawi SMA, Breggia A, et al. Multi-scale digital pathology patch-level prostate cancer grading using deep learning: use case evaluation of DiagSet dataset[J]. Bioengineering (Basel), 2024, 11(6): 624. DOI: 10.3390/bioengineering11060624.
[18]	Shao Y, Bazargani R, Karimi D, et al. Prostate cancer risk stratification by digital histopathology and deep learning[J]. JCO Clin Cancer Inform, 2024, 8: e2300184. DOI: 10.1200/CCI.23.00184.
[19]	Salvi M, Manini C, López JI, et al. Deep learning approach for accurate prostate cancer identification and stratification using combined immunostaining of cytokeratin, p63, and racemase[J]. Comput Med Imaging Graph, 2023, 109: 102288. DOI: 10.1016/j.compmedimag.2023.102288.
[20]	Karageorgos GM, Cho S, McDonough E, et al. Deep learning-based automated pipeline for blood vessel detection and distribution analysis in multiplexed prostate cancer images[J]. Front Bioinform, 2023, 3: 1296667. DOI: 10.3389/fbinf.2023.1296667.
[21]	Martell MT, Haven NJM, Cikaluk BD, et al. Deep learning-enabled realistic virtual histology with ultraviolet photoacoustic remote sensing microscopy[J]. Nat Commun, 2023, 14(1): 5967. DOI: 10.1038/s41467-023-41574-2.
[22]	Wong PF, McNeil C, Wang Y, et al. Clinical-grade validation of an autofluorescence virtual staining system with human experts and a deep learning system for prostate cancer[J]. Mod Pathol, 2024, 37(11): 100573. DOI: 10.1016/j.modpat.2024.100573.
[23]	Katsamenis OL, Olding M, Warner JA, et al. X-ray micro-computed tomography for nondestructive three-dimensional (3D) X-ray histology[J]. Am J Pathol, 2019, 189(8): 1608-1620. DOI: 10.1016/j.ajpath.2019.05.004.
[24]	Lin L, Wang LV. The emerging role of photoacoustic imaging in clinical oncology[J]. Nat Rev Clin Oncol, 2022, 19(6): 365-384. DOI: 10.1038/s41571-022-00615-3.
[25]	Tu H, Liu Y, Turchinovich D, et al. Stain-free histopathology by programmable supercontinuum pulses[J]. Nature Photon, 2016, 10(8): 534-540. DOI: 10.1038/nphoton.2016.94.
[26]	Huisken J, Swoger J, Del Bene F, et al. Optical sectioning deep inside live embryos by selective plane illumination microscopy[J]. Science, 2004, 305(5686): 1007-1009. DOI: 10.1126/science.1100035.
[27]	Tanaka N, Kanatani S, Tomer R, et al. Whole-tissue biopsy phenotyping of three-dimensional tumours reveals patterns of cancer heterogeneity[J]. Nat Biomed Eng, 2017, 1(10): 796-806. DOI: 10.1038/s41551-017-0139-0.
[28]	Glaser AK, Reder NP, Chen Y, et al. Light-sheet microscopy for slide-free non-destructive pathology of large clinical specimens[J]. Nat Biomed Eng, 2017, 1(7): 0084. DOI: 10.1038/s41551-017-0084.
[29]	Glaser AK, Reder NP, Chen Y, et al. Multi-immersion open-top light-sheet microscope for high-throughput imaging of cleared tissues. Nat Commun[J]. 2019, 10(1): 2781. DOI: 10.1038/s41467-019-10534-0.
[30]	Glaser AK, Bishop KW, Barner LA, et al. A hybrid open-top light-sheet microscope for versatile multi-scale imaging of cleared tissues[J]. Nat Methods, 2022, 19(5): 613-619. DOI: 10.1038/s41592-022-01468-5.
[31]	Xie W, Reder NP, Koyuncu C, et al. Prostate cancer risk stratification via nondestructive 3d pathology with deep learning-assisted gland analysis[J]. Cancer Res, 2022, 2(2): 334-345. DOI: 10.1158/0008-5472.CAN-21-2843.
[32]	Serafin R, Koyuncu C, Xie W, et al. Nondestructive 3D pathology with analysis of nuclear features for prostate cancer risk assessment[J]. J Pathol, 2023, 260(4):390-401. DOI: 10.1002/path.6090.
[33]	Song AH, Williams M, Williamson DFK, et al. Weakly Supervised AI for Efficient Analysis of 3D Pathology Samples[J]. ArXiv, 2024, 187(10):2502-2520.e17. DOI: 10.1016/j.cell.2024.03.035.
[34]	Chen RJ, Ding T, Lu MY, et al. Towards a general-purpose foundation model for computational pathology[J]. Nat Med, 2024, 30(3): 850-862. DOI: 10.1038/s41591-024-02857-3.
[35]	Lu MY, Chen B, Williamson DFK, et al. A visual-language foundation model for computational pathology[J]. Nat Med, 2024, 30(3): 863-874. DOI: 10.1038/s41591-024-02856-4.
[36]	Xu H, Usuyama N, Bagga J, et al. A whole-slide foundation model for digital pathology from real-world data[J]. Nature, 2024, 630(8015): 181-188. DOI: 10.1038/s41586-024-07441-w.
[37]	Vorontsov E, Bozkurt A, Casson A, et al. A foundation model for clinical-grade computational pathology and rare cancers detection[J]. Nat Med, 2024, 30(10): 2924-2935. DOI: 10.1038/s41591-024-03141-0.
[38]	Wang X, Yang S, Zhang J, et al. Transformer-based unsupervised contrastive learning for histopathological image classification[J]. Med Image Anal, 2022, 81: 102559. DOI: 10.1016/j.media.2022.102559.
[39]	Ding T, Wagner SJ, Song AH, et al. Multimodal whole slide foundation model for pathology[J]. 2024, arXiv:2411.19666v1.
[40]	Hua S, Yan F, Shen T, et al. PathoDuet: foundation models for pathological slide analysis of H&E and IHC stains[J]. Med Image Anal, 2024, 97: 103289. DOI: 10.1016/j.media.2024.103289.
[41]	Huang Z, Bianchi F, Yuksekgonul M, et al. A visual–language foundation model for pathology image analysis using medical Twitter[J]. Nat Med, 2023, 29(9): 2307-2316. DOI: 10.1038/s41591-023-02504-3.
[42]	Pan X, AbdulJabbar K, Coelho-Lima J, et al. The artificial intelligence-based model ANORAK improves histopathological grading of lung adenocarcinoma[J]. Nat Cancer, 2024, 5(2): 347-363. DOI: 10.1038/s43018-023-00694-w.
[43]	Kim C, Gadgil SU, DeGrave AJ, et al. Transparent medical image AI via an image–text foundation model grounded in medical literature[J]. Nat Med, 2024, 30(4): 1154-1165. DOI: 10.1038/s41591-024-02887-x.
[44]	Wang X, Zhao J, Marostica E, et al. A pathology foundation model for cancer diagnosis and prognosis prediction[J]. Nature, 2024, 634(8035): 970-978. DOI: 10.1038/s41586-024-07894-z.
[45]	Marrón-Esquivel JM, Duran-Lopez L, Linares-Barranco A, et al. A comparative study of the inter-observer variability on Gleason grading against Deep Learning-based approaches for prostate cancer[J]. Comput Biol Med, 2023, 159: 106856. DOI: 10.1016/j.compbiomed.2023.106856.
[46]	Raciti P, Sue J, Retamero JA, et al. Clinical validation of artificial intelligence-augmented pathology diagnosis demonstrates significant gains in diagnostic accuracy in prostate cancer detection[J]. Arch Pathol Lab Med, 2023, 147(10): 1178-1185. DOI: 10.5858/arpa.2022-0066-oa.
[47]	Eminaga O, Abbas M, Kunder C, et al. Critical evaluation of artificial intelligence as a digital twin of pathologists for prostate cancer pathology[J]. Sci Rep, 2024, 14: 5284. DOI: 10.1038/s41598-024-55228-w.
[48]	Arvidsson I, Svanemur E, Marginean F, et al. Artificial intelligence for detection of prostate cancer in biopsies during active surveillance[J]. BJU Int, 2024, 134(6): 1001-1009. DOI: 10.1111/bju.16456.
[49]	Lee SH, Jang HJ. Deep learning-based prediction of molecular cancer biomarkers from tissue slides: a new tool for precision oncology[J]. Clin Mol Hepatol, 2022, 28(4): 754-772. DOI: 10.3350/cmh.2021.0394.
[50]	Lu MY, Chen B, Williamson DFK, et al. A multimodal generative AI copilot for human pathology[J]. Nature, 2024, 634(8033): 466-473. DOI: 10.1038/s41586-024-07618-3.

文献	样本量	模型与方法	临床事件与结论	核心指标
Pantanowitz等^[9]	4 677张前列腺穿刺活检WSIs	多层CNN网络	检测前列腺癌、分辨高低级别前列腺癌、评估神经侵犯	AUC依次为：0.991、0.941、0.957
Bulten等^[10]	1 243例患者5 759次活检	基于CycleGANs的无监督归一化算法；基于U-Net的深度学习系统	检测PCa；Gleason分级系统	诊断效能AUC达到0.990 Gleason分级AUC：0.978、0.974
Ström等^[11]	1 289例患者6 682次活检	ImageNet预先训练Inception V3模型	检测PCa；计算穿刺活检的肿瘤负荷	诊断PCa：AUC为0.997，穿刺组织肿瘤负荷：相关性0.96
Duenweg等^[12]	47例前列腺根治术患者（前瞻性）	病理组学ATARI模型 ImageNet预训练ResNet-101	检测PCa；分辨高风险PCa与低风险PCa；Patch级Gleason分级	识别PCa准确性：88%，高、低级别Gleason分级准确性：72%
Campanella等^[13]	24 859张前列腺WSIs 9 962张皮肤癌WSIs 9 894张乳腺癌WSIs	仅采用诊断报告作为弱监督标签训练神经网络	多种肿瘤的识别与诊断	AUC：0.98
Fogarty等^[14]	52例患者150张穿刺活检PANDAS数据	基于ImageNet预训练VGG-16模型权重，分两步，依次训练MLP并再微调整体权重	检测PCa；识别Gleason3/4级以上病变	Patch水平以91%的准确性区分Gleason 3/4级及以上病变与良性标本
Inamdar等^[15]	1 000张WSI构建训练集，500张WSI构建测试集	ImageNet预训练并冻结的ResNet-50编码器，基于U-Net和注意力机制的解码器实现前列腺组织语义分割	分割肿瘤区域与腺体区域	腺体和肿瘤组织Dice系数0.92和0.91，平均0.9168
Patkar等^[16]	580张根治术切片，6 218张穿刺活检	癌症检测：MobileNetV3 Gleason分级：基于CutMix训练策略，ResNet50癌症分级	检测PCa；Patch级Gleason分级；评估不同Gleason比例与RFS、MFS、OS之间的关系	Patch级别肿瘤检测：准确率95.2%，RFS、MFS和OS：一致性0.69、0.72和0.64
Kondejkar等^[17]	430张完整标注的WSI，4 675张仅二分类诊断的WSI，46张由病理学家独立诊断的WSI	利用不同层数的ResNet，在不同放大倍率的Patch中检测肿瘤并进行分类	检测PCa；Patch级Gleason分级	精确度>0.9999
Shao等^[18]	502位RP术后患者	CCHEK（CAPRA-S+CNN HE & Ki-67）：将临床病理学数据HE和Ki-67染色组织微阵列的计算特征相结合，进行癌症风险分层	预测BCR、OS	风险分层和预后预测效果优

文献	样本量	模型与方法	临床事件与结论	核心指标
Pantanowitz等^[9]	4 677张前列腺穿刺活检WSIs	多层CNN网络	检测前列腺癌、分辨高低级别前列腺癌、评估神经侵犯	AUC依次为：0.991、0.941、0.957
Bulten等^[10]	1 243例患者5 759次活检	基于CycleGANs的无监督归一化算法；基于U-Net的深度学习系统	检测PCa；Gleason分级系统	诊断效能AUC达到0.990 Gleason分级AUC：0.978、0.974
Ström等^[11]	1 289例患者6 682次活检	ImageNet预先训练Inception V3模型	检测PCa；计算穿刺活检的肿瘤负荷	诊断PCa：AUC为0.997，穿刺组织肿瘤负荷：相关性0.96
Duenweg等^[12]	47例前列腺根治术患者（前瞻性）	病理组学ATARI模型 ImageNet预训练ResNet-101	检测PCa；分辨高风险PCa与低风险PCa；Patch级Gleason分级	识别PCa准确性：88%，高、低级别Gleason分级准确性：72%
Campanella等^[13]	24 859张前列腺WSIs 9 962张皮肤癌WSIs 9 894张乳腺癌WSIs	仅采用诊断报告作为弱监督标签训练神经网络	多种肿瘤的识别与诊断	AUC：0.98
Fogarty等^[14]	52例患者150张穿刺活检PANDAS数据	基于ImageNet预训练VGG-16模型权重，分两步，依次训练MLP并再微调整体权重	检测PCa；识别Gleason3/4级以上病变	Patch水平以91%的准确性区分Gleason 3/4级及以上病变与良性标本
Inamdar等^[15]	1 000张WSI构建训练集，500张WSI构建测试集	ImageNet预训练并冻结的ResNet-50编码器，基于U-Net和注意力机制的解码器实现前列腺组织语义分割	分割肿瘤区域与腺体区域	腺体和肿瘤组织Dice系数0.92和0.91，平均0.9168
Patkar等^[16]	580张根治术切片，6 218张穿刺活检	癌症检测：MobileNetV3 Gleason分级：基于CutMix训练策略，ResNet50癌症分级	检测PCa；Patch级Gleason分级；评估不同Gleason比例与RFS、MFS、OS之间的关系	Patch级别肿瘤检测：准确率95.2%，RFS、MFS和OS：一致性0.69、0.72和0.64
Kondejkar等^[17]	430张完整标注的WSI，4 675张仅二分类诊断的WSI，46张由病理学家独立诊断的WSI	利用不同层数的ResNet，在不同放大倍率的Patch中检测肿瘤并进行分类	检测PCa；Patch级Gleason分级	精确度>0.9999
Shao等^[18]	502位RP术后患者	CCHEK（CAPRA-S+CNN HE & Ki-67）：将临床病理学数据HE和Ki-67染色组织微阵列的计算特征相结合，进行癌症风险分层	预测BCR、OS	风险分层和预后预测效果优

模型名称	训练方法与模型架构	训练数据集
UNI^[34]	DINO v2自监督预训练	Mass-100K；1亿张Patch，10万个HE染色的WSIs，涵盖20种主要组织类型
CONCH^[35]	iBOT BERT预训练Patch编码器；CoCa视觉文本训练；多模态融合解码器	包含117万图像-标题对的病理视觉语言数据集
Prov-GigaPath^[36]	两阶段级联结构：基于LongNet切片级编码器，生成上下文化的嵌入；DINO v2自监督预训练	1.3亿个256×256 Patch上预训练的全切片病理学基础模型。171 189个WSI，涵盖超过30 000例患者的31种主要组织类型
Virchow^[37]	DINOv2自监督学习框架训练ViT-H	MSK-IMPACT；包含1 488 550 Patch，300万张WSI，涵盖17个部位
TITAN^[39]	iBOT BERT视觉预训练；基于PathChat和通义千问Qwen2-7B-Instruct生成ROI图像及描述组；CoCa框架全切片图像与病理报告对齐	Mass-340K；包含335 645张WSI、182 862份医学报告；涵盖20种器官、多种染色类型（HE占90.9%，IHC占9.1%）以及肿瘤性和非肿瘤性组织（分别占70.0%和30.0%）
CHIEF^[44]	1 500万个未标记的Patch上无监督预训练；WSI级别的弱监督预训练	1 500万张未标记的病理图像和60 530张覆盖19个癌种的WSI

模型名称	训练方法与模型架构	训练数据集
UNI^[34]	DINO v2自监督预训练	Mass-100K；1亿张Patch，10万个HE染色的WSIs，涵盖20种主要组织类型
CONCH^[35]	iBOT BERT预训练Patch编码器；CoCa视觉文本训练；多模态融合解码器	包含117万图像-标题对的病理视觉语言数据集
Prov-GigaPath^[36]	两阶段级联结构：基于LongNet切片级编码器，生成上下文化的嵌入；DINO v2自监督预训练	1.3亿个256×256 Patch上预训练的全切片病理学基础模型。171 189个WSI，涵盖超过30 000例患者的31种主要组织类型
Virchow^[37]	DINOv2自监督学习框架训练ViT-H	MSK-IMPACT；包含1 488 550 Patch，300万张WSI，涵盖17个部位
TITAN^[39]	iBOT BERT视觉预训练；基于PathChat和通义千问Qwen2-7B-Instruct生成ROI图像及描述组；CoCa框架全切片图像与病理报告对齐	Mass-340K；包含335 645张WSI、182 862份医学报告；涵盖20种器官、多种染色类型（HE占90.9%，IHC占9.1%）以及肿瘤性和非肿瘤性组织（分别占70.0%和30.0%）
CHIEF^[44]	1 500万个未标记的Patch上无监督预训练；WSI级别的弱监督预训练	1 500万张未标记的病理图像和60 530张覆盖19个癌种的WSI

Please choose a citation manager

Content to export