000281112 001__ 281112
000281112 005__ 20250918102645.0
000281112 0247_ $$2doi$$a10.48550/ARXIV.2509.03154
000281112 037__ $$aDZNE-2025-01073
000281112 1001_ $$0P:(DE-HGF)0$$aSzustakowski, Karol$$b0
000281112 245__ $$aPreserving instance continuity and length in segmentation through connectivity-aware loss computation
000281112 260__ $$barXiv$$c2025
000281112 3367_ $$0PUB:(DE-HGF)25$$2PUB:(DE-HGF)$$aPreprint$$bpreprint$$mpreprint$$s1758110568_31854
000281112 3367_ $$2ORCID$$aWORKING_PAPER
000281112 3367_ $$028$$2EndNote$$aElectronic Article
000281112 3367_ $$2DRIVER$$apreprint
000281112 3367_ $$2BibTeX$$aARTICLE
000281112 3367_ $$2DataCite$$aOutput Types/Working Paper
000281112 520__ $$aIn many biomedical segmentation tasks, the preservation of elongated structure continuity and length is more important than voxel-wise accuracy. We propose two novel loss functions, Negative Centerline Loss and Simplified Topology Loss, that, applied to Convolutional Neural Networks (CNNs), help preserve connectivity of output instances. Moreover, we discuss characteristics of experiment design, such as downscaling and spacing correction, that help obtain continuous segmentation masks. We evaluate our approach on a 3D light-sheet fluorescence microscopy dataset of axon initial segments (AIS), a task prone to discontinuity due to signal dropout. Compared to standard CNNs and existing topology-aware losses, our methods reduce the number of segmentation discontinuities per instance, particularly in regions with missing input signal, resulting in improved instance length calculation in downstream applications. Our findings demonstrate that structural priors embedded in the loss design can significantly enhance the reliability of segmentation for biological applications.
000281112 536__ $$0G:(DE-HGF)POF4-351$$a351 - Brain Function (POF4-351)$$cPOF4-351$$fPOF IV$$x0
000281112 536__ $$0G:(DE-Juel-1)E.40401.62$$aHelmholtz AI - Helmholtz Artificial Intelligence  Coordination Unit – Local Unit FZJ (E.40401.62)$$cE.40401.62$$x1
000281112 536__ $$0G:(DE-Juel-1)E54.303.11$$aHelmholtz AI Consultant Team FB Information (E54.303.11)$$cE54.303.11$$x2
000281112 588__ $$aDataset connected to DataCite
000281112 650_7 $$2Other$$aComputer Vision and Pattern Recognition (cs.CV)
000281112 650_7 $$2Other$$aFOS: Computer and information sciences
000281112 650_7 $$2Other$$aI.4.6; I.2.10
000281112 7001_ $$0P:(DE-2719)9003099$$aFrank, Luk$$b1
000281112 7001_ $$0P:(DE-2719)2812406$$aEsser, Julia$$b2
000281112 7001_ $$0P:(DE-2719)9001219$$aGründemann, Jan$$b3
000281112 7001_ $$0P:(DE-HGF)0$$aPiraud, Marie$$b4
000281112 773__ $$a10.48550/ARXIV.2509.03154
000281112 8564_ $$uhttps://pub.dzne.de/record/281112/files/DZNE-2025-01073_Restricted.pdf
000281112 8564_ $$uhttps://pub.dzne.de/record/281112/files/DZNE-2025-01073_Restricted.pdf?subformat=pdfa$$xpdfa
000281112 909CO $$ooai:pub.dzne.de:281112$$pVDB
000281112 9101_ $$0I:(DE-HGF)0$$6P:(DE-HGF)0$$aHelmholtz Munich$$b0
000281112 9101_ $$0I:(DE-HGF)0$$6P:(DE-HGF)0$$a Helmholtz AI$$b0
000281112 9101_ $$0I:(DE-588)1065079516$$6P:(DE-2719)9003099$$aDeutsches Zentrum für Neurodegenerative Erkrankungen$$b1$$kDZNE
000281112 9101_ $$0I:(DE-588)1065079516$$6P:(DE-2719)2812406$$aDeutsches Zentrum für Neurodegenerative Erkrankungen$$b2$$kDZNE
000281112 9101_ $$0I:(DE-588)1065079516$$6P:(DE-2719)9001219$$aDeutsches Zentrum für Neurodegenerative Erkrankungen$$b3$$kDZNE
000281112 9101_ $$0I:(DE-HGF)0$$6P:(DE-HGF)0$$a Helmholtz Munich$$b4
000281112 9101_ $$0I:(DE-HGF)0$$6P:(DE-HGF)0$$a Helmholtz AI$$b4
000281112 9131_ $$0G:(DE-HGF)POF4-351$$1G:(DE-HGF)POF4-350$$2G:(DE-HGF)POF4-300$$3G:(DE-HGF)POF4$$4G:(DE-HGF)POF$$aDE-HGF$$bGesundheit$$lNeurodegenerative Diseases$$vBrain Function$$x0
000281112 9141_ $$y2025
000281112 920__ $$lyes
000281112 9201_ $$0I:(DE-2719)5000069$$kAG Gründemann$$lNeural Circuit Computations$$x0
000281112 980__ $$apreprint
000281112 980__ $$aVDB
000281112 980__ $$aI:(DE-2719)5000069
000281112 980__ $$aUNRESTRICTED