Abstract
Background Panoptic segmentation networks are a newer class of image segmentation algorithms that are constrained to understand the difference between instance-type objects (objects that are discrete countable entities, such as renal tubules) and group-type objects (uncountable, amorphous regions of texture such as renal interstitium). This class of deep networks has unique advantages for biological datasets, particularly in computational pathology.
Methods We collected 126 periodic acid Schiff whole slide images of native diabetic nephropathy, lupus nephritis, and transplant surveillance kidney biopsies, and fully annotated them for the following micro-compartments: interstitium, glomeruli, globally sclerotic glomeruli, tubules, and arterial tree (arteries/arterioles). Using this data, we trained a panoptic feature pyramid network. We compared performance of the network against a renal pathologist’s annotations, and the method’s transferability to other computational pathology domain tasks was investigated.
Results The panoptic feature pyramid networks showed high performance as compared to renal pathologist for all of the annotated classes in a testing set of transplant kidney biopsies. The network was not only able to generalize its object understanding across different stains and species of kidney data, but also across several organ types.
Conclusions Panoptic networks have unique advantages for computational pathology; namely, these networks internally model structural morphology, which aids bootstrapping of annotations for new computational pathology tasks.
Competing Interest Statement
The authors have declared no competing interest.
