New neuron image dataset released to advance microscopy analysis methods
-
The dataset contains 1874 high-resolution fluorescence microscopy images of rodent brain slices showing neuronal structures, with 750 images having manual annotations.
-
The images are organized into 3 collections based on stain color (green, yellow, red), capturing diverse neuronal markers like cFOS, CTb, orexin, etc.
-
Multiple annotation formats are provided (masks, polygons, bounding boxes, dots) to enable exploring different tasks like segmentation, detection, counting.
-
Challenges include class imbalance, variability in shapes/sizes/distributions of cells, overcrowding, ambiguity in annotations.
-
Intended uses include testing model generalization, transfer learning, domain adaptation; comparing annotation types and formats; advancing methodologies for microscopy analysis.