SignificanceManual annotations are necessary for training supervised learning algorithms for object detection and instance segmentation. These manual annotations are difficult to acquire, noisy, and inconsistent across readers.AimThe goal of this work is to describe and demonstrate multireader generalizations of the Jaccard and Sørensen indices for object detection and instance segmentation.ApproachThe multireader Jaccard and Sørensen indices are described in terms of “calls,” “objects,” and number of readers. These generalizations reduce to the equations defined by confusion matrix variables in the two-reader case. In a test set of 50 cell microscopy images, we use these generalizations to assess reader variability and compare the performance of an object detection network (Yolov5) and an instance segmentation algorithm (Cellpose2.0) with a group of five human readers using the Mann–Whitney U-test with Bonferroni correction for multiplicity.ResultsThe multireader generalizations were statistically different from the mean of pairwise comparisons of readers (p < 0.0001). Further, these multireader generalizations informed when a reader was performing differently than the group. Finally, these generalizations show that Yolov5 and Cellpose2.0 performed similarly to the pool of human readers. The lower bound of the one-sided 90% confidence interval for the difference in the multireader Jaccard index between the pool of human readers and the pool of human readers plus an algorithm were −0.019 and −0.016 for Yolov5 and Cellpose2.0, respectively.ConclusionsMultireader generalizations of the Jaccard and Sørensen indices provide metrics for characterizing the agreement of an arbitrary number of readers on object detection and instance segmentation tasks.
Deep convolutional neural networks (CNNs) have demonstrated high accuracy in a wide range of computer vision applications, including medical and biological imaging. Many CNNs are fully supervised learning algorithms, and their performance is directly associated with the quality of the training data labels, which are human-defined. In this work, we investigate the fidelity of human-defined truth for cell detection, segmentation, and classification tasks in multiplex microscopy images. We compare manual annotations from human readers on three tasks. Readers were asked to (1) segment all cells in single-channel fluorescence images of a pannuclear stain (DAPI), (2) segment cells in two-channel fluorescence images (CD20/DAPI), only identifying cells with both nuclear signal (DAPI) and signal from a cell surface marker (CD20), and (3) segment two separate cell classes in three-channel fluorescence images (CD3/CD4/DAPI). In this third task, readers were asked to identify cells that had nuclear signal and were CD3+/CD4- and CD3+/CD4+. By comparing these manual segmentations within and between readers, we demonstrate that human readers show the least variability in single-channel DAPI segmentation (p<<0.05, F test for equal variance). We also compared the agreement of human readers with one another to the agreement of an object-detection network, Yolov5, on cell detection in DAPI images. All pairwise comparisons of human readers with other human readers yielded an average F1-score of 0.83±0.14, and comparisons of Yolov5 with human readers yielded an average F1-score of 0.84±0.12 (p=0.26, Welch’s T test). We therefore demonstrate that out of the provided tasks, DAPI detection provides the highest fidelity ground truth, and were unable to show a difference between Yolov5 and human readers in this task.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.