Gspcnet:基于多模态数据的前列腺癌 Gleason 评分

13 Pages Posted: 20 May 2025

See all articles by Zheng Gong

Zheng Gong

affiliation not provided to SSRN

Meiqi Yang

affiliation not provided to SSRN

Shuai Ji

affiliation not provided to SSRN

Wei Li

affiliation not provided to SSRN

Lun Zhao

Liaoning University of Traditional Chinese Medicine

Xue-Song Wang

affiliation not provided to SSRN

Peng Wang

affiliation not provided to SSRN

zhongyuan Liu

affiliation not provided to SSRN

Abstract

前列腺癌的诊断通常取决于前列腺磁共振成像 (MRI) 和随后的活检程序。然而,前列腺 MRI 的有限性能会影响活检的阳性预测值。因此,已经集成了深度学习方法以提高整体诊断效果。尽管如此,前列腺癌患者数据表现出多模态,仅依靠 mpMRI 可能无法全面反映患者的病情。因此,有必要根据多模态数据开发格里森分级分数。然而,目前缺乏关于前列腺癌的多样化多模式数据。为了解决这一差距,我们提出了一个简单但高度可解释的诊断网络 (GSPCnet),该网络整合了来自前列腺癌患者的多模态数据以预测 Gleason 分级。具体来说,我们采用改进的 ResNet50 网络对所有患者图像进行分类。使用前列腺特异性抗原 (PSA) 参数过滤结果,将处理后的序列数据中的最高等级作为单个序列的分析结果。根据三个序列 (DWI、T2 和 DCE) 的结果,评估患者的 Gleason 分级。在适当的授权下,我们构建了一个由 215 名不同阶段的前列腺癌患者组成的多模式数据集。每个患者记录包括 DWI、T2 和 DCE 序列以及 PSA 参数。选择具有清晰可见病灶的图像作为分类的训练和测试数据集。数据集以 7:3 的比例分为训练集和测试集。经过多轮训练,我们的分类网络实现了以下精度:DWI 序列准确度为 0.776 (二次加权,95% 置信区间 [CI] 0.764–0.785),T2 序列准确度为 0.881 (二次加权,95% CI 0.876–0.882),DCE 序列准确度为 0.921 (二次加权,95% CI 0.920–0.926)。然后,我们为每个分类添加了 5 名患者的数据,经过 20 次随机试验,计算出患者的诊断准确性为 0.822 (二次加权,95% CI 0.798–0.846)。这些结果强调了我们的网络在活检前评估中的效用,而我们网络的高度可解释性为临床医生做出诊断提供了重要帮助。

Note:
Funding Information: This study was financially supported by Liaoning Provincial Department of Medical and Engineering Interdisciplinary Joint Fund (grant number 2022-YGJC-48).

Conflict of Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Ethical Approval: This study was approved by the Ethics Committee of Shengjing Hospital Affiliated China Medical University(No. 2024PS1244K).

Keywords: Multimodal information processing, Mp-MRI, PSA, Prostate Cancer, Gleason Score

Suggested Citation

Gong, Zheng and Yang, Meiqi and Ji, Shuai and Li, Wei and Zhao, Lun and Wang, Xue-Song and Wang, Peng and Liu, zhongyuan, Gspcnet:基于多模态数据的前列腺癌 Gleason 评分. Available at SSRN: https://ssrn.com/abstract=5252399 or http://dx.doi.org/10.2139/ssrn.5252399

Zheng Gong

affiliation not provided to SSRN ( email )

No Address Available

Meiqi Yang

affiliation not provided to SSRN ( email )

No Address Available

Shuai Ji

affiliation not provided to SSRN ( email )

No Address Available

Wei Li

affiliation not provided to SSRN ( email )

No Address Available

Lun Zhao

Liaoning University of Traditional Chinese Medicine ( email )

Xue-Song Wang

affiliation not provided to SSRN ( email )

No Address Available

Peng Wang

affiliation not provided to SSRN ( email )

No Address Available

Zhongyuan Liu (Contact Author)

affiliation not provided to SSRN ( email )

No Address Available

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
4
Abstract Views
34
PlumX Metrics