推荐hgial · 2021-09
“It's unlikely they're making an elementary ML mistake here.”
推荐,认为作者不太可能犯基本的 ML 错误,为论文方法辩护
深度学习能从医学影像预测自述种族,即使影像被裁剪加噪
研究表明,标准深度学习模型能在多种医学影像模态上以高精度预测自述种族,在外部验证以及针对临床任务优化模型时依然成立;这并非由疾病分布等简单代理变量造成,且在所有解剖区域和频谱上都存在。作者强调模型能预测种族本身并非问题所在,关键在于临床专家做不到这一点,因此给医学影像模型的部署带来巨大风险,缓解也将很困难。
评论者围绕其可信度争论:有人认为作者不太可能犯基本的 ML 错误,也有人批评它可能只是检测到不同设备和操作规程的差异,或只是做了些简单的图像处理。
“It's unlikely they're making an elementary ML mistake here.”
“What it appears they are detecting is the difference between different machines and practices.”
“The paper is mostly applying trivial image processing functions and seeing how some software they don't understand is responding.”
“A conceptually related finding that computer vsion models can effectively classify patient race from significantly noised x-ray images”