Limits...
Automated detection and segmentation of synaptic contacts in nearly isotropic serial electron microscopy images.

Kreshuk A, Straehle CN, Sommer C, Koethe U, Cantoni M, Knott G, Hamprecht FA - PLoS ONE (2011)

Bottom Line: The statistical learning is performed on geometrical features of 3D neighborhoods of each voxel and can fully exploit the high z-resolution of the data.On a quantitative validation dataset of 111 synapses in 409 images of 1948×1342 pixels with manual annotations by three independent experts the error rate of the algorithm was found to be comparable to that of the experts (0.92 recall at 0.89 precision).Our software offers a convenient interface for labeling the training data and the possibility to visualize and proofread the results in 3D.

View Article: PubMed Central - PubMed

Affiliation: Interdisciplinary Center for Scientific Computing, University of Heidelberg, Heidelberg, Germany.

ABSTRACT
We describe a protocol for fully automated detection and segmentation of asymmetric, presumed excitatory, synapses in serial electron microscopy images of the adult mammalian cerebral cortex, taken with the focused ion beam, scanning electron microscope (FIB/SEM). The procedure is based on interactive machine learning and only requires a few labeled synapses for training. The statistical learning is performed on geometrical features of 3D neighborhoods of each voxel and can fully exploit the high z-resolution of the data. On a quantitative validation dataset of 111 synapses in 409 images of 1948×1342 pixels with manual annotations by three independent experts the error rate of the algorithm was found to be comparable to that of the experts (0.92 recall at 0.89 precision). Our software offers a convenient interface for labeling the training data and the possibility to visualize and proofread the results in 3D. The source code, the test dataset and the ground truth annotation are freely available on the website http://www.ilastik.org/synapse-detection.

Show MeSH
Error examples.A, B, C: false negative decisions of the human observers, D, E, F: false positive detections of the human observers, shown as yellow “ball” labels in the image center, G, H, I: false negative decisions of the algorithm, J, K, L false positive decisions of the algorithm.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3198725&req=5

pone-0024899-g005: Error examples.A, B, C: false negative decisions of the human observers, D, E, F: false positive detections of the human observers, shown as yellow “ball” labels in the image center, G, H, I: false negative decisions of the algorithm, J, K, L false positive decisions of the algorithm.

Mentions: The results show that with an adequate selection of appearance features, synapses are sufficiently different from other structures in neural tissue to allow for reliable automated detection in nearly isotropic FIB/SEM serial images. Fig. 5 illustrates typical false negative and false positive detections of the humans and of the algorithm, which have different causes. The false positives of the algorithm are mostly caused by myelinated membranes or very dark lines located near mitochondria (Fig. 5J, 5K, 5L). Similarly, most of the false negative detections also stem from synapses located very close to myelinated membranes. In the probability maps, they become connected to the large false positives caused by these membranes, and these large connected components are then filtered out based on the size criterion (Fig. 5G). Since ilastik provides a convenient summary report of all detected synapses (Fig. 4) and reduces the data from millions of voxels to just dozens of synapse candidates, the false positives for the entire stack can easily be discarded by a human in just a few minutes of additional proofreading.


Automated detection and segmentation of synaptic contacts in nearly isotropic serial electron microscopy images.

Kreshuk A, Straehle CN, Sommer C, Koethe U, Cantoni M, Knott G, Hamprecht FA - PLoS ONE (2011)

Error examples.A, B, C: false negative decisions of the human observers, D, E, F: false positive detections of the human observers, shown as yellow “ball” labels in the image center, G, H, I: false negative decisions of the algorithm, J, K, L false positive decisions of the algorithm.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3198725&req=5

pone-0024899-g005: Error examples.A, B, C: false negative decisions of the human observers, D, E, F: false positive detections of the human observers, shown as yellow “ball” labels in the image center, G, H, I: false negative decisions of the algorithm, J, K, L false positive decisions of the algorithm.
Mentions: The results show that with an adequate selection of appearance features, synapses are sufficiently different from other structures in neural tissue to allow for reliable automated detection in nearly isotropic FIB/SEM serial images. Fig. 5 illustrates typical false negative and false positive detections of the humans and of the algorithm, which have different causes. The false positives of the algorithm are mostly caused by myelinated membranes or very dark lines located near mitochondria (Fig. 5J, 5K, 5L). Similarly, most of the false negative detections also stem from synapses located very close to myelinated membranes. In the probability maps, they become connected to the large false positives caused by these membranes, and these large connected components are then filtered out based on the size criterion (Fig. 5G). Since ilastik provides a convenient summary report of all detected synapses (Fig. 4) and reduces the data from millions of voxels to just dozens of synapse candidates, the false positives for the entire stack can easily be discarded by a human in just a few minutes of additional proofreading.

Bottom Line: The statistical learning is performed on geometrical features of 3D neighborhoods of each voxel and can fully exploit the high z-resolution of the data.On a quantitative validation dataset of 111 synapses in 409 images of 1948×1342 pixels with manual annotations by three independent experts the error rate of the algorithm was found to be comparable to that of the experts (0.92 recall at 0.89 precision).Our software offers a convenient interface for labeling the training data and the possibility to visualize and proofread the results in 3D.

View Article: PubMed Central - PubMed

Affiliation: Interdisciplinary Center for Scientific Computing, University of Heidelberg, Heidelberg, Germany.

ABSTRACT
We describe a protocol for fully automated detection and segmentation of asymmetric, presumed excitatory, synapses in serial electron microscopy images of the adult mammalian cerebral cortex, taken with the focused ion beam, scanning electron microscope (FIB/SEM). The procedure is based on interactive machine learning and only requires a few labeled synapses for training. The statistical learning is performed on geometrical features of 3D neighborhoods of each voxel and can fully exploit the high z-resolution of the data. On a quantitative validation dataset of 111 synapses in 409 images of 1948×1342 pixels with manual annotations by three independent experts the error rate of the algorithm was found to be comparable to that of the experts (0.92 recall at 0.89 precision). Our software offers a convenient interface for labeling the training data and the possibility to visualize and proofread the results in 3D. The source code, the test dataset and the ground truth annotation are freely available on the website http://www.ilastik.org/synapse-detection.

Show MeSH