Irmakci Ismail, Nateghi Ramin, Zhou Rujoi, Ross Ashley E, Yang Ximing J, Cooper Lee A D, Goldstein Jeffery A
medRxiv. 2023 May 2:2023.04.28.23289287. doi: 10.1101/2023.04.28.23289287.
Machine learning (ML) models are poised to transform surgical pathology practice. The most successful use attention mechanisms to examine whole slides, identify which areas of tissue are diagnostic, and use them to guide diagnosis. Tissue contaminants, such as floaters, represent unexpected tissue. While human pathologists are extensively trained to consider and detect tissue contaminants, we examined their impact on ML models. We trained 4 whole slide models. Three operate in placenta for 1) detection of decidual arteriopathy (DA), 2) estimation of gestational age (GA), and 3) classification of macroscopic placental lesions. We also developed a model to detect prostate cancer in needle biopsies. We designed experiments wherein patches of contaminant tissue are randomly sampled from known slides and digitally added to patient slides and measured model performance. We measured the proportion of attention given to contaminants and examined the impact of contaminants in T-distributed Stochastic Neighbor Embedding (tSNE) feature space. Every model showed performance degradation in response to one or more tissue contaminants. DA detection balanced accuracy decreased from 0.74 to 0.69 +/- 0.01 with addition of 1 patch of prostate tissue for every 100 patches of placenta (1% contaminant). Bladder, added at 10% contaminant raised the mean absolute error in estimating gestation age from 1.626 weeks to 2.371 +/ 0.003 weeks. Blood, incorporated into placental sections, induced false negative diagnoses of intervillous thrombi. Addition of bladder to prostate cancer needle biopsies induced false positives, a selection of high-attention patches, representing 0.033mm, resulted in a 97% false positive rate when added to needle biopsies. Contaminant patches received attention at or above the rate of the average patch of patient tissue. Tissue contaminants induce errors in modern ML models. The high level of attention given to contaminants indicates a failure to encode biological phenomena. Practitioners should move to quantify and ameliorate this problem.
机器学习（ML）模型有望改变外科病理学实践。最成功的模型使用注意力机制来检查整张切片，识别哪些组织区域具有诊断价值，并利用这些区域来指导诊断。组织污染物，如漂浮物，代表意外的组织。虽然人类病理学家经过广泛培训以考虑和检测组织污染物，但我们研究了它们对ML模型的影响。我们训练了4个整张切片模型。其中3个用于胎盘，分别用于1）蜕膜动脉病（DA）的检测、2）孕周（GA）的估计以及3）宏观胎盘病变的分类。我们还开发了一个用于检测针吸活检中前列腺癌的模型。我们设计了实验，从已知切片中随机采样污染物组织块，并将其数字添加到患者切片中，然后测量模型性能。我们测量了给予污染物的注意力比例，并在T分布随机邻域嵌入（tSNE）特征空间中检查了污染物的影响。每个模型在面对一种或多种组织污染物时都表现出性能下降。每100个胎盘组织块添加1个前列腺组织块（1%污染物）时，DA检测的平衡准确率从0.74降至0.69±0.01。添加10%的膀胱组织污染物会使孕周估计的平均绝对误差从1.626周增加到2.371±0.003周。将血液掺入胎盘切片会导致绒毛间隙血栓的假阴性诊断。在前列腺癌针吸活检中添加膀胱组织会导致假阳性，选择代表0.033毫米的高注意力组织块添加到针吸活检中时，假阳性率达到97%。污染物组织块获得的注意力与患者组织的平均组织块相当或更高。组织污染物会在现代ML模型中引发错误。对污染物给予的高度关注表明未能对生物学现象进行编码。从业者应着手量化并改善这一问题。
