通过深度学习来分类脏异种移植排斥在整个幻灯片上的 histopathologic图像
Yongrong Ye1, Liubing Xia1, Shicong Yang2
1Department of Kidney Transplantation, The Third Affiliated Hospital of Sun Yat-sen University, Guangzhou, China.
Frontiers in immunology
|July 22, 2024
概括
一个新的深度学习系统,脏排斥人工智能模型 (RRAIM),准确地检测和分类脏移植排斥并预测预后,在初步评估中超过人类病理学家.
科学领域:
- 腎臟病學 (nephrology) 是一種醫學專業.
- 病理学 病理学 病理学
- 人工智能的人工智能
- 医疗成像医学成像
背景情况:
- 目前的移植拒绝诊断依赖于主观手动组织病理学,导致可重现性问题.
- 需要客观和可重复的方法来诊断移植排斥.
- 整个幻灯片图像 (WSI) 的自动评估提供了一个潜在的解决方案.
研究的目的:
- 开发一个深度学习系统,用于自动评估脏全移植活检WSIs.
- 为了实现移植排斥的自动检测和亚型化.
- 预测移植拒绝的预后.
主要方法:
- 收集和分析了302个脏异种移植活检中的H&E染色的WSIs.
- 使用多实例学习和卷积神经网络 (CNN) 来进行特征提取和分类.
- 使用AUC,混矩阵和病理学家机器比较来评估模型性能.
主要成果:
- 脏排斥人工智能模型 (RRAIM) 在排斥检测和亚型化方面实现了0.798的3类AUC.
- 在一个独立的测试组中,RRAIM的表现超过了三个移植病理学家的表现.
- 预后模型准确预测了移植损失 (AUC=0.936) 和治疗反应 (AUC=0.756).
结论:
- 开发了使用多实例学习进行移植拒绝诊断和预后的深度学习模型.
- 这些模型在检测,分类和预测拒绝结果方面表现强.
- 这些人工智能模型在协助病理诊断和改善患者护理方面表现有前途.
相关概念视频
01:28Kidney Transplant I: Introduction
A kidney transplant is a surgical approach that involves replacing a non-functioning kidney with a healthy one from a donor. This procedure is often a treatment option for end-stage renal disease (ESRD) patients. The method requires careful recipient selection, including evaluating various medical and psychosocial factors. These criteria vary between transplant centers but generally include assessments of the patient's overall health, adherence to medical recommendations, and lifestyle...
01:26Kidney Transplant II: Surgical Procedure
Preoperative ManagementThe primary goals of preoperative management in kidney transplantation are to optimize the patient’s metabolic state and prepare them for surgery through diet adjustments, necessary dialysis, and tailored medical treatment. This phase also involves comprehensive infection screening and patient education about the surgical procedure and postoperative care to improve outcomes and adherence.Medical ManagementA comprehensive evaluation is required for both the living donor...