子宫内膜异位症影响约10%出生时被指派为女性的个体,其诊断和管理颇具挑战。诊断通常需要通过腹腔镜手术或分析T1/T2 MRI图像来识别该疾病的多种征象,后者更快捷、成本更低,但准确性较差。子宫内膜异位症的一个关键诊断征象是子宫直肠陷凹(Pouch of Douglas,POD)闭塞。然而,即使是经验丰富的临床医生,也难以根据MRI图像准确分类POD闭塞,这使得训练可靠的AI模型变得复杂。本文提出人机协作多模态多评估者学习(Human-AI Collaborative Multi-modal Multi-rater Learning,HAICOMM)方法来应对上述挑战。HAICOMM是首个探索该问题三个重要方面的方法:1)多评估者学习,从每个训练样本可用的多个“噪声”标签中提取更干净的标签;2)多模态学习,在训练和测试中利用T1/T2 MRI图像;3)人机协作,构建一个利用临床医生和AI模型预测结果的系统,以提供比独立临床医生和AI模型更准确的分类。在我们为验证该方法而收集的多评估者T1/T2 MRI子宫内膜异位症数据集上,所提出的HAICOMM模型表现优于临床医生集成、噪声标签学习模型以及多评估者学习方法。
英文摘要:
Endometriosis, affecting about 10% of individuals assigned female at birth, is challenging to diagnose and manage. Diagnosis typically involves the identification of various signs of the disease using either laparoscopic surgery or the analysis of T1/T2 MRI images, with the latter being quicker and cheaper but less accurate. A key diagnostic sign of endometriosis is the obliteration of the Pouch of Douglas (POD). However, even experienced clinicians struggle with accurately classifying POD obliteration from MRI images, which complicates the training of reliable AI models. In this paper, we introduce the Human-AI Collaborative Multi-modal Multi-rater Learning (HAICOMM) methodology to address the challenge above. HAICOMM is the first method that explores three important aspects of this problem: 1) multi-rater learning to extract a cleaner label from the multiple "noisy" labels available per training sample; 2) multi-modal learning to leverage the presence of T1/T2 MRI images for training and testing; and 3) human-AI collaboration to build a system that leverages the predictions from clinicians and the AI model to provide more accurate classification than standalone clinicians and AI models. Presenting results on the multi-rater T1/T2 MRI endometriosis dataset that we collected to validate our methodology, the proposed HAICOMM model outperforms an ensemble of clinicians, noisy-label learning models, and multi-rater learning methods.
Distilling Missing Modality Knowledge from Ultrasound for Endometriosis Diagnosis with Magnetic Resonance Images
Australian Institute for Machine Learning, University of Adelaide(阿德莱德大学澳大利亚机器学习研究所);Flinders Health and Medical Research Institute, Flinders University(弗林德斯大学弗林德斯健康与医学研究所);Robinson Research Institute, University of Adelaide(阿德莱德大学罗宾逊研究所);Centre for Vision, Speech and Signal Processing, University of Surrey(萨里大学视觉、语音与信号处理中心)
Multi-modal Disease Classification in Incomplete Datasets Using Geometric Matrix Completion
Technical University of Munich (TUM)(慕尼黑工业大学);German Center for Vertigo and Balance Disorders (DSGZ), Ludwig-Maximilians-Universität (LMU)(德国眩晕与平衡障碍中心(DSGZ),路德维希-马克西米利安大学(LMU))
A Pathology Foundation Model for Gastric Cancer with Real-World Validation
Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong SAR, China(计算机科学与工程系,香港科技大学,香港特别行政区,中国);Department of Pathology, Nanfang Hospital, Southern Medical University, Guangzhou, China(病理学系,南方医科大学南芳医院,广州,中国);Department of Pathology, School of Basic Medical Sciences, Southern Medical University, Guangzhou, China(病理学系,南方医科大学基础医学学院,广州,中国);Guangdong Province Key Laboratory of Molecular Tumor Pathology, Guangzhou, China(广东省分子肿瘤病理学重点实验室,广州,中国);Department of Anatomical and Cellular Pathology, The Chinese University of Hong Kong, Hong Kong SAR, China(解剖学与细胞病理学系,香港中文大学,香港特别行政区,中国);Pathology Artificial Intelligence Development and Assessment Laboratory, State Key Laboratory of Translational Oncology, The Chinese University of Hong Kong, Hong Kong SAR, China(病理人工智能发展与评估实验室,转化肿瘤学国家重点实验室,香港中文大学,香港特别行政区,中国)
Calibrate the inter-observer segmentation uncertainty via diagnosis-first principle
Baidu Inc.(百度公司);University of Electronic Science and Technology of China(电子科技大学);South China University of Technology(华南理工大学);Shenzhen Eye Hospital, Jinan University(暨南大学附属深圳眼科医院);Institute for Infocomm Research, A*STAR(新加坡科技研究局资讯通信研究院)
在医学图像上,许多组织/病变可能是模糊的。这就是为什么医学分割通常由一组临床专家进行注释,以减轻个人偏见。然而,这种临床常规也给机器学习算法的应用带来了新的挑战。如果没有明确的基础事实,就很难训练和评估深度学习模型。当注释是从不同的评分者收集的,一个常见的选择是多数投票。然而,这种策略忽略了评分员熟练程度之间的差异。在本文中,我们认为预测的分割与校准的观察者间的不确定性的任务。我们注意到,在临床实践中,医学图像分割通常用于辅助疾病诊断。受此启发,我们提出了诊断优先原则,即以疾病诊断为标准来校准观察者间的分割不确定性。根据这一思想,提出了一个诊断优先分割框架(DiFF),用于从原始图像中估计诊断优先分割,DiFF将首先学习将多评价器分割标签融合到一个单一的地面真值,以最大限度地提高疾病诊断性能。我们将融合后的地面真值称为诊断第一地面真值(Diagnosis First Ground-truth,DF-GT),并进一步提出Take and Give模型从原始图像中分割DF-GT。我们验证了三个不同的医疗分割任务的有效性:眼底图像的OD/OC分割,超声图像上的甲状腺结节分割,皮肤镜图像上的皮肤病变分割。实验结果表明,所提出的DiFF是能够显着促进相应的疾病诊断,这优于以前的国家的最先进的多评价者学习方法。
Weakly Supervised Medical Image Segmentation With Soft Labels and Noise Robust Loss
University of Calgary(卡尔加里大学);McCaig Institute for Bone and Joint Health(麦凯格骨骼与关节健康研究所);University of Alberta(阿尔伯塔大学);Alberta Machine Intelligence Institute (AMII)(阿尔伯塔机器智能研究所)
Pseudo-D: Informing Multi-View Uncertainty Estimation with Calibrated Neural Training Dynamics
Department of Electrical and Computer Engineering, University of British Columbia, Vancouver, Canada(电气与计算机工程系,不列颠哥伦比亚大学,温哥华,加拿大);Division of Cardiology, Vancouver General Hospital, Vancouver, Canada(心血管科,温哥华总医院,温哥华,加拿大)
Transformer-based Model for Oral Epithelial Dysplasia Segmentation
University of Warwick(华威大学);University of Sheffield(谢菲尔德大学);University of Campinas(坎皮纳斯大学);Queen’s University Belfast(贝尔法斯特女王大学);University of Birmingham(伯明翰大学)
A multi-path 2.5 dimensional convolutional neural network system for segmenting stroke lesions in brain MRI images
New Jersey Institute of Technology(新泽西理工学院);Kessler Foundation(凯斯勒基金会);Rutgers – New Jersey Medical School(罗格斯大学新泽西医学院);Medical College of Wisconsin(威斯康星医学院);Rutgers University – Newark(罗格斯大学纽瓦克分校)