Renal Cell Carcinoma Detection and Subtyping Framework with Minimal Point-Based Annotation A semi-supervised algorithm for cancer region detection and subtyping of renal cell carcinoma, this is a two step classification framework.

Introduction

全幻灯片图像(WSI)中的癌性区域检测和分型是肾细胞癌(RCC)诊断的基础。开发自动化 RCC 诊断系统的主要挑战是缺乏具有精确注释的大规模数据集。在本文中,我们提出了一种框架,该框架采用半监督学习(SSL)方法通过一种称为最小点基(Min-Point)注释的新颖注释方法来准确检测癌变区域。预测结果可通过混合损失训练策略有效地用于分类模型中,以进行子类型化。注释者只需在每个WSI中标记一些癌点和非癌点。对RCC的三种重要亚型进行的实验证明,使用Min-Point注释数据集训练的癌区域检测器的性能可与使用完整癌区域描绘的数据集训练的分类器相媲美。在子类型化中,就测试f1分数而言,所提出的模型比仅使用整张幻灯片诊断标签训练的模型要好12%。我们相信,我们的“先检测后再分类”模式与Min-Point注释相结合,将为开发具有类似挑战的智能系统树立标准

Publications

  1. Renal Cell Carcinoma Detection and Subtyping with Minimal Point-Based Annotation in Whole-Slide Images
    Cite this
    Zeyu Gao, Pargorn Puttapirat, Jiangbo Shi, Chen Li, 2020/10/2 - Medical Image Computing and Computer Assisted Intervention – MICCAI 2020,Lecture Notes in Computer Science

Credits

  1. Zeyu Gao betpotti@gmail.com
    Investigator

    Computer Science and Technology, Xi'an Jiaotong University, China

Community Ratings

UsabilityEfficiencyReliabilityRated By
0 user
Sign in to rate
Summary
AccessionBT007155
Tool TypeApplication
CategoryImage analysis
PlatformsLinux/Unix
TechnologiesPython3
User InterfaceTerminal Command Line
Latest Release1.0 (May 31, 2021)
Download Count801
Country/RegionChina
Submitted ByZeyu Gao
Fundings

2018YFC0910400