Some Reasearches On Brain-wide Tracing Of Neuronal Population

Brain neural network research is helpful to understand the working principle of brain,and is also the the basis of pathogenesis and treatment of brain disease and inspire artificial intelligence research.As neural network consists of neurons,quantifying neuron topology and geometric features,namely neuronal tracing is great significance for brain neural network research.Owing to the development of labelling and microscopy imaging techniques,we are able to acquire lots of brain-wide image datasets.These image datasets include many neurons which form neuronal poplation and can provide abundant information.However recent automatic neuronal tracing tools can trace the morphology of neurons but cannot get the relationship between neurites and neurons,thus cannot separate single neuron and be applied to neuronal population tracing.A few commercial softwares can trace neuronal population semi-automatically,but is labor intensive.It is very important and urgent to develop brain-wide neuronal population tracing tool.There are four difficulties of high precision brain-wide neuronal population tracing.(1)Interfered by dense neuritis signal,it is hard to get the relationship between neurites and neurons,thus cannot separate single neuron.(2)It is lack of brain-wide neuronal population tracing framework of several hundred GB or TB-level size.(3)Owing to the tree structure of neuron,small mistakes of neuron branch will have great bad effect on all the children branches and cause lots of mistakes at whole brain scale.(4)It is very difficult to find these mistakes in TB-level size area.We propose neuronal population tracing method and build a set of brain-wide neuronal population tracing solution to conquer the above difficulties.We have developed premilinary brain-wide neuronal population tracing tool.(1)For difficulty 1,we propose local dense neuronal population tracing method.It includes automatic neuronal population segment method and revision with selective display mode.Automatic neuronal population segment method can get the links of neuritis and build the relationship of neuritis and neurons automatically,then separate single neuron from neuronal population.Revision with selective display mode can build the tree structure of neurons and display neuron branches selectively,conquering the difficulty that annotator cannot observe the dense neuronal population correctly.(2)We develop brain-wide neuronal population tracing tool named GTree.GTree provide a set of brain-wide neuronal population tracing solution based on local dense neuronal population tracing method,conquering difficulty 2,3 and 4.(2.1)GTree based on big data framework GHD and TDat,has ability to process big image dataset of which the volume is from 10 GB to PB-level size.(2.2)GTree provide put the following solution into practice: automatic image block division,neurite tracing,trace results merging and online revision.(2.3)GTree build fast brain-wide neuronal population traverse system and multiple neuron tracing results online integration system.These systems avoid the precision collapse of neuron branches trace errors.GTree achives good performance.The accuracy of tracing result traced by GTree is higher than commercial tracing software.GTree can trace brain-wide neuronal population priliminarily.For example,GTree cost about 10 hours to get single neuron tracing result from neuronal population imaged by TDI-fMOST,about 0.6 cm per hour.GTree is 5 times faster than commercial tracing software.GTree has been applied to tracing brain-wide neuronal population preliminarily,which are imaged by 2p-fMOST,TDI-fMOST,serial two-photo tomography,light sheet tomography and so on.GTree has been used in Britton Chance Center for Biomedical Photonics,Wuhan National Laboratory for Optoelectronics,Huazhong University of Science and Technology;Department of Neurobiology School of Basic Medicine and Tongji Medical College,Yunyun Han group;Cold Spring Harbor Laboratory Osten group and so on.GTree has advantage in the field of brain-wide neuronal population reconstruction,and is expected to play an important role in brain neural network research.