[Skip to Content]
[Skip to Content Landing]
Views 576
Citations 0
Comment & Response
April 24, 2018

Machine Learning Compared With Pathologist Assessment—Reply

Author Affiliations
  • 1Diagnostic Image Analysis Group, Department of Radiology and Nuclear Medicine, Radboud University Medical Center, Nijmegen, the Netherlands
  • 2Department of Pathology, Radboud University Medical Center, Nijmegen, the Netherlands
JAMA. 2018;319(16):1726. doi:10.1001/jama.2018.1478

In Reply We agree that the 3 criteria pointed out by Dr van Smeden and colleagues are important when comparing algorithms and physicians in diagnostic studies. However, we do not agree that our study deviated from these criteria.

First, the concern was raised that the physicians should have had access to all regular diagnostic information, including relevant additional diagnostic testing. The aim of our study was not to replace pathologists for diagnosing lymph node metastases in a clinical setting but to investigate the potential of machine learning algorithms for detection of metastases in hematoxylin and eosin–stained sentinel lymph node tissue sections and compare these with the performance of pathologists. Such detection is a diagnostic task that is common in pathology practice, for which the pathologists participating in our study were highly qualified. The potential utility of deep learning algorithms for pathological diagnosis in a clinical setting requires further assessment, which is beyond the scope of our study.

×