Touchstone Benchmark: Are We on the Right Way for Evaluating AI Algorithms for Medical Segmentation?
Johns Hopkins University, UniBo, IIT · Johns Hopkins University · Vanderbilt University · German Cancer Research Center (DKFZ) · KU Leuven · DKFZ · Shanghai Jiaotong University · Monash University · Shanghai AI Laboratory · Deutsches Krebsforschungszentrum (DKFZ) · Deutsches Krebsforschungszentrum · German Cancer Research Center (DKFZ), Heidelberg · German Cancer Research Center · Northwestern Polytechnical University · University of Adelaide · Alibaba Group · Northwestern Polytechnical University, Research & Development Institute of Northwestern Polytechnical University in Shenzhen · Hong Kong University of Science and Technology · Hong Kong University of Science and Technology (Guangzhou) & HKUST · Universität Regensburg · RWTH Aachen, Rheinisch Westfälische Technische Hochschule Aachen · University of Regensburg · Harbin Institute of Technology (Shenzhen) · Harbin Institute of Technology · Beijing Academy of Artificial Intelligence · The Chinese University of Hong Kong · Peking University · Shanghai Jiao Tong University · The Hong Kong University of Science and Technology · Duke University · Stony Brook University · Department of Computer Science and Engineering, Hong Kong University of Science and Technology · The Hong Kong Univeristy of Science and Technology · NVIDIA · Istituto Italiano di Tecnologia · EPFL - EPF Lausanne · JHU
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
How can we test AI performance? This question seems trivial, but it isn't. Standard benchmarks often have problems such as in-distribution and small-size test sets, oversimplified metrics, unfair comparisons, and short-term outcome pressure. As a consequence, good performance on standard benchmarks does not guarantee success in real-world scenarios. To address these problems, we present Touchstone, a large-scale collaborative segmentation benchmark of 9 types of abdominal organs. This benchmark is based on 5,195 training CT scans from 76 hospitals around the world and 5,903 testing CT scans from 11 additional hospitals. This diverse test set enhances the statistical significance of benchmark results and rigorously evaluates AI algorithms across various out-of-distribution scenarios. We invited 14 inventors of 19 AI algorithms to train their algorithms, while our team, as a third party, independently evaluated these algorithms on three test sets. In addition, we also evaluated pre-existing AI frameworks---which, differing from algorithms, are more flexible and can support different algorithms—including MONAI from NVIDIA, nnU-Net from DKFZ, and numerous other open-source frameworks. We are committed to expanding this benchmark to encourage more innovation of AI algorithms for the medical domain.