A deep unrolling network based on cartoon texture decomposition for low-dose CT reconstruction

{"OBJECTIVE:":[0],"Deep":[1],"unrolling":[2,66,78,253,261,269],"network,":[3],"as":[4,104],"a":[5,73,129,155,166,259],"promising":[6],"deep":[7,65,77,141,268],"learning":[8],"approach":[9],"for":[10,144,263],"low-dose":[11],"computed":[12],"tomography":[13],"(LDCT)":[14],"reconstruction,":[15],"can":[16],"efficiently":[17],"address":[18],"the":[19,35,52,57,62,71,84,109,115,124,140,161,184,194,203,211,244,250,254,264],"issues":[20],"of":[21,56,64,100,151,163,183,233,252,267],"severe":[22],"noise":[23,182,217],"and":[24,170,179,198,202,216,236],"artifacts":[25,215],"in":[26,231],"LDCT":[27,219],"imaging.":[28],"However,":[29],"most":[30],"existing":[31],"methods":[32],"predominantly":[33],"unrolled":[34,138],"fidelity":[36,86,111,256],"term":[37,87],"to":[38,45,113,139,175,229,247],"convolutional":[39],"neural":[40],"networks":[41],"(CNNs)":[42],"while":[43,221],"failing":[44],"unroll":[46],"regularization":[47,90,105,245],"terms,":[48,106],"which":[49,226],"inevitably":[50],"limits":[51],"feature":[53],"capture":[54],"capability":[55],"network.":[58],"To":[59,158],"further":[60,159,242],"improve":[61],"performance":[63],"networks,":[67],"this":[68],"paper":[69],"proposed":[70],"CTDNet,":[72],"cartoon":[74,96],"texture":[75,97],"decomposition-based":[76],"network":[79,143],"that":[80,210],"unrolls":[81,243],"not":[82],"only":[83],"data":[85,110,255],"but":[88],"also":[89],"terms":[91,232,246],"into":[92],"CNNs.":[93],"APPROACH:":[94],"The":[95,206],"decomposition":[98],"model":[99],"Meyer":[101],"was":[102,121,136,173],"incorporated":[103],"together":[107],"with":[108,154],"term,":[112,257],"formulate":[114],"reconstruction":[116,142,164],"optimization":[117],"problem.":[118],"This":[119,134,240],"problem":[120],"solved":[122],"by":[123,146],"Chambolle-Pock":[125],"(CP)":[126],"algorithm,":[127],"yielding":[128],"single-loop":[130],"iterative":[131,152],"algorithm":[132],"instance.":[133],"instance":[135],"then":[137],"LDCT,":[145],"replacing":[147],"each":[148],"update":[149],"step":[150],"process":[153],"simple":[156],"sub-network.":[157],"enhance":[160],"quality":[162],"results,":[165],"lightweight":[167],"image":[168],"refinement":[169],"fusion":[171],"module":[172],"employed":[174],"perform":[176],"detail":[177],"enhancement":[178],"remove":[180],"residual":[181],"reconstructed":[185],"images.":[186],"MAIN":[187],"RESULTS:":[188],"Extensive":[189],"experiments":[190],"were":[191],"conducted":[192],"on":[193,249],"\\"Low-Dose":[195],"CT":[196],"Image":[197],"Projection":[199],"Data\\"":[200],"dataset":[201],"\\"Piglet":[204],"Dataset\\".":[205],"experiment":[207],"results":[208],"demonstrated":[209],"CTDNet":[212],"effectively":[213],"removes":[214],"from":[218],"images":[220],"maximally":[222],"preserving":[223],"textural":[224],"structures,":[225],"enables":[227],"it":[228],"outperform":[230],"visual":[234],"effects":[235],"objective":[237],"metrics.":[238],"SIGNIFICANCE:":[239],"work":[241],"CNNs":[248],"basis":[251],"providing":[258],"novel":[260],"strategy":[262],"future":[265],"design":[266],"networks.":[270],".":[271]}