Large Scale Image Segmentation with Structured Loss Based Deep Learning for Connectome Reconstruction

Large Scale Image Segmentation with Structured Loss Based Deep Learning for Connectome Reconstruction

We present a method combining affinity prediction with region agglomeration, which improves significantly upon the state of the art of neuron segmentation from electron microscopy (EM) in accuracy and scalability. Our method consists of a 3D U-Net, trained to predict affinities between voxels, follo...

Description complète

Détails bibliographiques
Publié dans:IEEE transactions on pattern analysis and machine intelligence. - 1979. - 41(2019), 7 vom: 26. Juli, Seite 1669-1680
Auteur principal: Funke, Jan (Auteur)
Autres auteurs: Tschopp, Fabian, Grisaitis, William, Sheridan, Arlo, Singh, Chandan, Saalfeld, Stephan, Turaga, Srinivas C
Format: Article en ligne
Langue:English
Publié: 2019
Accès à la collection:IEEE transactions on pattern analysis and machine intelligence
Sujets:Journal Article
LEADER 01000caa a22002652c 4500
001 NLM286362511
003 DE-627
005 20250223192050.0
007 cr uuu---uuuuu
008 231225s2019 xx |||||o 00| ||eng c
024 7 |a 10.1109/TPAMI.2018.2835450  |2 doi 
028 5 2 |a pubmed25n0954.xml 
035 |a (DE-627)NLM286362511 
035 |a (NLM)29993708 
040 |a DE-627  |b ger  |c DE-627  |e rakwb 
041 |a eng 
100 1 |a Funke, Jan  |e verfasserin  |4 aut 
245 1 0 |a Large Scale Image Segmentation with Structured Loss Based Deep Learning for Connectome Reconstruction 
264 1 |c 2019 
336 |a Text  |b txt  |2 rdacontent 
337 |a ƒaComputermedien  |b c  |2 rdamedia 
338 |a ƒa Online-Ressource  |b cr  |2 rdacarrier 
500 |a Date Completed 04.02.2020 
500 |a Date Revised 04.02.2020 
500 |a published: Print-Electronic 
500 |a Citation Status MEDLINE 
520 |a We present a method combining affinity prediction with region agglomeration, which improves significantly upon the state of the art of neuron segmentation from electron microscopy (EM) in accuracy and scalability. Our method consists of a 3D U-Net, trained to predict affinities between voxels, followed by iterative region agglomeration. We train using a structured loss based on Malis, encouraging topologically correct segmentations obtained from affinity thresholding. Our extension consists of two parts: First, we present a quasi-linear method to compute the loss gradient, improving over the original quadratic algorithm. Second, we compute the gradient in two separate passes to avoid spurious gradient contributions in early training stages. Our predictions are accurate enough that simple learning-free percentile-based agglomeration outperforms more involved methods used earlier on inferior predictions. We present results on three diverse EM datasets, achieving relative improvements over previous results of 27, 15, and 250 percent. Our findings suggest that a single method can be applied to both nearly isotropic block-face EM data and anisotropic serial sectioned EM data. The runtime of our method scales linearly with the size of the volume and achieves a throughput of $\sim$∼ 2.6 seconds per megavoxel, qualifying our method for the processing of very large datasets 
650 4 |a Journal Article 
700 1 |a Tschopp, Fabian  |e verfasserin  |4 aut 
700 1 |a Grisaitis, William  |e verfasserin  |4 aut 
700 1 |a Sheridan, Arlo  |e verfasserin  |4 aut 
700 1 |a Singh, Chandan  |e verfasserin  |4 aut 
700 1 |a Saalfeld, Stephan  |e verfasserin  |4 aut 
700 1 |a Turaga, Srinivas C  |e verfasserin  |4 aut 
773 0 8 |i Enthalten in  |t IEEE transactions on pattern analysis and machine intelligence  |d 1979  |g 41(2019), 7 vom: 26. Juli, Seite 1669-1680  |w (DE-627)NLM098212257  |x 1939-3539  |7 nnas 
773 1 8 |g volume:41  |g year:2019  |g number:7  |g day:26  |g month:07  |g pages:1669-1680 
856 4 0 |u http://dx.doi.org/10.1109/TPAMI.2018.2835450  |3 Volltext 
912 |a GBV_USEFLAG_A 
912 |a SYSFLAG_A 
912 |a GBV_NLM 
912 |a GBV_ILN_350 
951 |a AR 
952 |d 41  |j 2019  |e 7  |b 26  |c 07  |h 1669-1680