基于生成模型与求和-乘积网络的乳腺X线摄影扫描异常检测
计算机视觉与模式识别
2022-12-12 v1 机器学习
摘要
仅以健康数据训练的无监督异常检测模型近年日益重要，因为医学数据标注是一项繁琐任务。自编码器与生成对抗网络是用于学习数据分布的标准异常检测方法。然而，在测试样本似然的推断与评估方面它们存在不足。我们提出一种生成模型与概率图模型的新颖组合。在经自编码器编码图像样本后，数据分布由随机与张量化求和-乘积网络建模，确保测试时的精确高效推断。我们在乳腺X线图像上采用分块处理，评估不同自编码器架构与随机及张量化求和-乘积网络的结合，观察到优于单独使用各模型及医学数据异常检测 state-of-the-art 的性能。
引用
@article{arxiv.2210.06188,
title = {Anomaly Detection using Generative Models and Sum-Product Networks in Mammography Scans},
author = {Marc Dietrichstein and David Major and Martin Trapp and Maria Wimmer and Dimitrios Lenis and Philip Winter and Astrid Berg and Theresa Neubauer and Katja Bühler},
journal= {arXiv preprint arXiv:2210.06188},
year = {2022}
}
备注
Submitted to DGM4MICCAI 2022 Workshop. This preprint has not undergone peer review (when applicable) or any post-submission improvements or corrections. The Version of Record of this contribution is published in LNCS 13609, and is available online at https://doi.org/10.1007/978-3-031-18576-2_8